<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: YUANZHI ROBOTICS</title>
    <description>The latest articles on DEV Community by YUANZHI ROBOTICS (@yuanzhirobotics).</description>
    <link>https://dev.to/yuanzhirobotics</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4208144%2F51312e12-24bc-4997-9e06-9b574482f069.jpg</url>
      <title>DEV Community: YUANZHI ROBOTICS</title>
      <link>https://dev.to/yuanzhirobotics</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/yuanzhirobotics"/>
    <language>en</language>
    <item>
      <title>Robot dog or drone for construction and plant site inspection: how to decide</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Sat, 14 Nov 2026 01:13:27 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/robot-dog-or-drone-for-construction-and-plant-site-inspection-how-to-decide-2kh5</link>
      <guid>https://dev.to/yuanzhirobotics/robot-dog-or-drone-for-construction-and-plant-site-inspection-how-to-decide-2kh5</guid>
      <description>&lt;p&gt;&lt;em&gt;By the engineering team at YUANZHI ROBOTICS. We design and build sensor payloads for robot dogs, UGVs and drones. Originally published on our site; AI-assisted draft, reviewed by our engineers.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; use a drone when the job is to see a large area, a roof, a facade or a stockpile from above, quickly and often. Use a robot dog when the job is inside or underneath things: plant rooms, cable trenches, corridors, stairs, and any check that needs a sensor held close to an asset for a fixed time, such as reading a gauge, measuring a hot joint or sniffing for gas. Many sites end up with both, because the two machines answer different questions. The sections below turn that rule into a checklist you can apply to your own site.&lt;/p&gt;

&lt;h2&gt;
  
  
  Start from the question, not the machine
&lt;/h2&gt;

&lt;p&gt;Write down what an inspection round must produce. "Has the excavation moved since last week?" is an aerial question: it needs coverage, a consistent viewpoint and a georeferenced output. "Is pump 3 running hotter than yesterday?" or "Is the pressure on gauge PI-204 inside its band?" is a ground question: it needs the same close viewpoint every time, often under a roof, and a sensor that can dwell on the target. If most items on your list look like the second type, a drone will struggle no matter how good its camera is.&lt;/p&gt;

&lt;h2&gt;
  
  
  Side-by-side comparison
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Factor&lt;/th&gt;
&lt;th&gt;Inspection drone&lt;/th&gt;
&lt;th&gt;Robot dog&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Best viewpoint&lt;/td&gt;
&lt;td&gt;From above and from a distance&lt;/td&gt;
&lt;td&gt;At human eye level, close to the asset&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Indoors and under structures&lt;/td&gt;
&lt;td&gt;Limited: GNSS and line of sight are lost, flying near people and equipment is restricted&lt;/td&gt;
&lt;td&gt;Native use case: corridors, plant rooms, trenches, mezzanines&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Stairs and gratings&lt;/td&gt;
&lt;td&gt;Not relevant&lt;/td&gt;
&lt;td&gt;Climbs stairs; check the specific platform's limits&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Area covered per hour&lt;/td&gt;
&lt;td&gt;Very high&lt;/td&gt;
&lt;td&gt;Walking pace, route-based&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Dwell time on one target&lt;/td&gt;
&lt;td&gt;Short; hovering consumes the battery&lt;/td&gt;
&lt;td&gt;Long; standing still costs little energy&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payload headroom (examples)&lt;/td&gt;
&lt;td&gt;Matrice 4T about 0.2 kg; Matrice 30T about 0.3 kg; Matrice 350 RTK about 2.7 kg&lt;/td&gt;
&lt;td&gt;Unitree Go2 EDU about 8 kg; Unitree B2 about 40 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Endurance (listed)&lt;/td&gt;
&lt;td&gt;Matrice 4TD up to 54 min; Matrice 350 RTK 55 min without payload&lt;/td&gt;
&lt;td&gt;Unitree B2 listed at 4 to 6 hours&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Typical outputs&lt;/td&gt;
&lt;td&gt;Orthophotos, 3D models, roof and facade thermal surveys, stockpile volume&lt;/td&gt;
&lt;td&gt;Gauge readings, close thermal readings, gas readings, 3D LiDAR maps of interiors, anomaly photos per checkpoint&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Main regulatory work&lt;/td&gt;
&lt;td&gt;Airspace and flight permissions, operator licensing&lt;/td&gt;
&lt;td&gt;Site safety rules for a moving machine among people&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Platform figures are the manufacturers' listed values and change with configuration; payload numbers are what the airframe or body can carry, not what is left after mounts, cables and computers.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxy83k4jbesxj2p4071a5.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxy83k4jbesxj2p4071a5.jpg" alt="3D render of a DJI Matrice 350 RTK carrying a YUANZHI ROBOTICS sensor payload" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a DJI Matrice 350 RTK carrying a YUANZHI ROBOTICS sensor payload&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  When a drone is the better tool
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Construction progress and earthworks.&lt;/strong&gt; Repeated aerial surveys give comparable models and volumes. A ground robot cannot see the top of a stockpile or a roof slab.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Roofs, facades, tall structures and long linear assets.&lt;/strong&gt; Solar arrays on roofs, pipe racks seen from above, perimeter fences and access roads are covered in minutes from the air.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Automated daily flights from a dock.&lt;/strong&gt; Docked aircraft such as the Matrice 4TD (listed at about 1.85 kg, IP55, up to 54 minutes) can fly a fixed route without a pilot walking out. The trade-off is that you mostly keep DJI's integrated sensors; the small aircraft leave little room for a third-party payload.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Heavier custom sensors outdoors.&lt;/strong&gt; A Matrice 350 RTK can carry a gimbal such as the Zenmuse H30T (about 920 g) plus a light extra pod. For more detail see &lt;a href="https://www.yuanzhirobot.com/en/guides/dji-matrice-4t-vs-payload-drone.html" rel="noopener noreferrer"&gt;Matrice 4T vs a custom payload drone&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4536sctp701tjhvm0fd.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4536sctp701tjhvm0fd.jpg" alt="3D render of a Unitree Go2 EDU robot dog carrying a YUANZHI ROBOTICS sensor payload" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a Unitree Go2 EDU robot dog carrying a YUANZHI ROBOTICS sensor payload&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  When a robot dog is the better tool
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Indoor and covered areas.&lt;/strong&gt; Substations, pump houses, boiler rooms, cable tunnels and basements. There is no GNSS, so the robot localizes against a LiDAR map recorded on the first walk.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Repeated close-range readings.&lt;/strong&gt; A pan-tilt camera on a legged robot can stop at the same checkpoint, at the same angle, every round, which is what automatic gauge reading and trend-based thermal alarms need.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Multi-level plants.&lt;/strong&gt; Legged robots handle stairs, kerbs and grating floors that stop wheeled robots, and they do not need a clear flight path.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Heavier or power-hungry sensor stacks.&lt;/strong&gt; Even a mid-size robot dog has more payload margin than a compact drone. A Livox Mid-360 LiDAR (265 g, about 6.5 W), a RealSense D435i depth camera (75 g, about 2.5 W) and a four-gas module (about 50 g) together weigh well under 1 kg before the computer and mount.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F76lxu4y9hacekl4a9azs.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F76lxu4y9hacekl4a9azs.jpg" alt="A robot dog's pan-tilt camera stopping at a checkpoint to read analog pressure gauges" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;A robot dog's pan-tilt camera stopping at a checkpoint to read analog pressure gauges&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Questions that decide it on a real site
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;What share of checkpoints are under a roof?&lt;/strong&gt; If more than about half, start with a ground robot.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Do you need numbers or pictures?&lt;/strong&gt; Numbers from a fixed asset (pressure, temperature trend, gas concentration) favour the robot dog. Area pictures and models favour the drone.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Who is around during rounds?&lt;/strong&gt; Flying over workers usually needs extra permissions; a walking robot needs a speed limit and stop zones agreed with site safety.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;How often, and with how much staff?&lt;/strong&gt; A docked drone or a robot dog with a charging station can run unattended rounds. Without either, someone has to deploy the machine every time.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Does the sensor exist for the platform?&lt;/strong&gt; A compact drone's sensor set is mostly fixed. A robot dog or a larger drone can take a custom payload, which matters for gas detection, acoustic imaging or a specific LiDAR.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Using both
&lt;/h2&gt;

&lt;p&gt;A common split on larger sites: the drone does a scheduled outdoor sweep and the perimeter, the robot dog does the indoor route and the close readings, and both feed the same dashboard with alarms tied to asset IDs. The integration work is mostly software: one asset list, one alarm format, and maps that share a site coordinate frame. Plan that early, because retrofitting a common data model after two separate deployments is slower than building it once. The same reasoning applies when comparing legged and wheeled ground robots; see &lt;a href="https://www.yuanzhirobot.com/en/guides/ugv-vs-robot-dog-inspection.html" rel="noopener noreferrer"&gt;UGV vs robot dog for inspection&lt;/a&gt;.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" alt="LiDAR-inertial mapping trajectory recorded by a robot dog during a walk-through" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;LiDAR-inertial mapping trajectory recorded by a robot dog during a walk-through&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Next step
&lt;/h2&gt;

&lt;p&gt;If you already have a platform in mind, build a sensor stack and check its weight and power in the &lt;a href="https://www.yuanzhirobot.com/?lang=en#config" rel="noopener noreferrer"&gt;payload configurator&lt;/a&gt;. Platform pages with dimensions and payload notes: &lt;a href="https://www.yuanzhirobot.com/en/platform/unitree-go2-edu.html" rel="noopener noreferrer"&gt;Unitree Go2 EDU&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/platform/unitree-b2.html" rel="noopener noreferrer"&gt;Unitree B2&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/platform/dji-matrice-350-rtk.html" rel="noopener noreferrer"&gt;DJI Matrice 350 RTK&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/platform/dji-matrice-4td.html" rel="noopener noreferrer"&gt;DJI Matrice 4TD&lt;/a&gt;. YUANZHI ROBOTICS designs and integrates payloads and software for both kinds of machine: see &lt;a href="https://www.yuanzhirobot.com/en/service/inspection-robot.html" rel="noopener noreferrer"&gt;inspection robots&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/service/robot-payload.html" rel="noopener noreferrer"&gt;robot dog payloads&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/service/drone-payload.html" rel="noopener noreferrer"&gt;drone payloads&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/service/robot-autonomy.html" rel="noopener noreferrer"&gt;autonomy software&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Is a robot dog or a drone better for construction site inspection?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;For progress tracking, earthworks, roofs and facades a drone is better because it sees large areas from above quickly. For enclosed floors, plant rooms, stairs and close checks of installed equipment a robot dog is better because it works without GNSS and can stop at the same checkpoint every round.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Can a drone inspect indoors instead of a robot dog?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Only in limited cases. Indoors a drone loses GNSS, has to fly close to people and equipment, and cannot hover for long at one target. Repeated indoor rounds with fixed checkpoints are usually done by a ground robot.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;How much payload can a robot dog carry compared with an inspection drone?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Listed figures: Unitree Go2 EDU about 8 kg and Unitree B2 about 40 kg, against about 0.2 kg for a DJI Matrice 4T, 0.3 kg for a Matrice 30T and 2.7 kg for a Matrice 350 RTK. Mounts, cables and computers come out of that budget.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Can a robot dog and a drone share the same inspection software?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Yes, if the integration is planned: one asset list, one alarm format and maps in a shared site coordinate frame. The robots run their own navigation, and a common dashboard collects readings and alarms from both.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Original article: &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-dog-vs-drone-for-site-inspection.html" rel="noopener noreferrer"&gt;Robot dog or drone for construction and plant site inspection: how to decide&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>drones</category>
      <category>iot</category>
      <category>hardware</category>
    </item>
    <item>
      <title>DJI Matrice 350 RTK vs Matrice 400 for third-party custom payloads</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Fri, 13 Nov 2026 01:13:20 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/dji-matrice-350-rtk-vs-matrice-400-for-third-party-custom-payloads-4bl9</link>
      <guid>https://dev.to/yuanzhirobotics/dji-matrice-350-rtk-vs-matrice-400-for-third-party-custom-payloads-4bl9</guid>
      <description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Pick the Matrice 400 when your non-DJI payload, its bracket and its onboard computer add up to more than about 1.5 kg, or when it has to fly together with a Zenmuse gimbal. The Matrice 400 is listed with about 6 kg of payload; the Matrice 350 RTK with about 2.7 kg. Pick the Matrice 350 RTK when a single light sensor pod is the whole mission, when you already own the aircraft, batteries and trained pilots, or when a smaller airframe matters for transport. The rest of this guide shows how to make that decision with numbers instead of headlines.&lt;/p&gt;

&lt;h2&gt;
  
  
  The two aircraft side by side
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Matrice 350 RTK&lt;/th&gt;
&lt;th&gt;Matrice 400&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Aircraft weight&lt;/td&gt;
&lt;td&gt;about 6.47 kg&lt;/td&gt;
&lt;td&gt;about 9.74 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Rated payload&lt;/td&gt;
&lt;td&gt;about 2.7 kg&lt;/td&gt;
&lt;td&gt;about 6 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payload as share of aircraft weight&lt;/td&gt;
&lt;td&gt;about 42%&lt;/td&gt;
&lt;td&gt;about 62%&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Listed flight time (no payload)&lt;/td&gt;
&lt;td&gt;55 min&lt;/td&gt;
&lt;td&gt;59 min&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Size (mm)&lt;/td&gt;
&lt;td&gt;810 x 670 x 430&lt;/td&gt;
&lt;td&gt;980 x 1020 x 470&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Built in&lt;/td&gt;
&lt;td&gt;RTK, six-direction obstacle sensing, IP55 listed&lt;/td&gt;
&lt;td&gt;RTK, obstacle-sensing radar&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Best fit for third-party work&lt;/td&gt;
&lt;td&gt;one light pod, or one Zenmuse plus a very light add-on&lt;/td&gt;
&lt;td&gt;Zenmuse plus a real second payload, or one heavy combined pod&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Figures are manufacturer listings as recorded in our payload catalog. Check DJI's current specification sheet before you freeze a design.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxy83k4jbesxj2p4071a5.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxy83k4jbesxj2p4071a5.jpg" alt="3D render of a DJI Matrice 350 RTK carrying a YUANZHI ROBOTICS sensor payload" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a DJI Matrice 350 RTK carrying a YUANZHI ROBOTICS sensor payload&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Work out the real payload budget first
&lt;/h2&gt;

&lt;p&gt;The rated payload is not what your sensor may weigh. It has to cover the sensor, the bracket and dampers, the enclosure, the onboard computer with its carrier board and heatsink, cables, connectors and any Zenmuse gimbal you keep on the aircraft. The mechanical and cabling overhead is never zero, and for a small sensor it can weigh as much as the sensor itself.&lt;/p&gt;

&lt;p&gt;Two worked examples with catalog weights:&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Configuration&lt;/th&gt;
&lt;th&gt;Listed sensor weight&lt;/th&gt;
&lt;th&gt;Matrice 350 RTK (2.7 kg)&lt;/th&gt;
&lt;th&gt;Matrice 400 (6 kg)&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Zenmuse H30T + Livox Mid-360 pod&lt;/td&gt;
&lt;td&gt;920 g + 265 g = 1,185 g&lt;/td&gt;
&lt;td&gt;about 1.5 kg left for bracket, computer and cables: workable, little margin&lt;/td&gt;
&lt;td&gt;about 4.8 kg left: comfortable&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Zenmuse L2 + Zenmuse S1 spotlight + Zenmuse V1 speaker&lt;/td&gt;
&lt;td&gt;905 g + 760 g + 690 g = 2,355 g&lt;/td&gt;
&lt;td&gt;about 0.35 kg left: no room for anything custom&lt;/td&gt;
&lt;td&gt;about 3.6 kg left for a third-party pod&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Hesai XT32M2X + Jetson Orin NX 16GB computer + mesh radio, no Zenmuse&lt;/td&gt;
&lt;td&gt;490 g + 34 g, plus computer assembly&lt;/td&gt;
&lt;td&gt;fits with good margin&lt;/td&gt;
&lt;td&gt;fits; aircraft is larger than the job needs&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The pattern is consistent: the Matrice 350 RTK carries &lt;em&gt;one&lt;/em&gt; thing well. As soon as the mission needs a Zenmuse camera for the pilot and a separate measuring payload, the 350 RTK is flown close to its limit, and wind margin and endurance suffer. For a method to build this budget line by line, see our guide on &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-payload-weight-budget.html" rel="noopener noreferrer"&gt;calculating a payload weight and power budget&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Power: the second limit people forget
&lt;/h2&gt;

&lt;p&gt;Third-party payloads normally draw power from the aircraft through DJI's payload interfaces, and the available power is limited per port. The loads add up faster than the weights do. From our catalog: Zenmuse H30T about 28 W, Zenmuse L2 about 28 W, Zenmuse S1 spotlight about 68 W, Zenmuse V1 speaker about 25 W, Livox Mid-360 6.5 W, Hesai XT32M2X 10 W, a Jetson Orin NX 16GB module up to about 40 W, a Doodle Labs Mesh Rider Mini about 5 W.&lt;/p&gt;

&lt;p&gt;A LiDAR plus an Orin NX and a radio is therefore a 50 to 60 W custom load before any DJI payload is counted. On either aircraft we check three things before choosing the supply: the continuous power the chosen port allows according to DJI's current Payload SDK documentation, the inrush current when the computer boots, and whether a brown-out on the payload rail could disturb the aircraft. When the load is too high or too noisy for the port, the answer is a small isolated payload battery, which costs weight. That weight is easy to accept on a Matrice 400 and painful on a 350 RTK.&lt;/p&gt;

&lt;h2&gt;
  
  
  Mounting and interfaces
&lt;/h2&gt;

&lt;p&gt;Both aircraft accept third-party payloads through DJI's Payload SDK (PSDK) route: a gimbal-port adapter for payloads that should behave like a gimbal camera, and an expansion port for payloads fixed to the airframe. Through PSDK a custom payload can read aircraft position, attitude and time, show its own widgets in the pilot app and send a data stream down the DJI link. The details differ between the two generations: connector type, the number of positions that can be used at once and the power available at each. A bracket and cable set designed for the Matrice 350 RTK should not be assumed to fit the Matrice 400.&lt;/p&gt;

&lt;p&gt;What does carry over is the expensive part: the sensor, the computer, the time synchronisation scheme and the software. If you expect to move from a 350 RTK to a 400 later, ask your integrator to keep the pod self-contained with a single power and data connector, so only the bracket and the harness change.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fitx2eo0dd56augoxxfhu.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fitx2eo0dd56augoxxfhu.jpg" alt="3D render of a DJI Matrice 400 with a YUANZHI ROBOTICS under-body sensor pod and a gimbal camera at the front" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a DJI Matrice 400 with a YUANZHI ROBOTICS under-body sensor pod and a gimbal camera at the front&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  When the Matrice 350 RTK is still the right answer
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;You already own it.&lt;/strong&gt; Aircraft, batteries, charging station and pilot training are sunk cost. A pod of up to roughly 1.5 kg all-in does not justify a new airframe.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;The payload replaces the Zenmuse instead of joining it.&lt;/strong&gt; A mapping pod with its own camera, a gas-sensing pod or a compact LiDAR with onboard computer can each fly alone.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Transport and site access.&lt;/strong&gt; At 810 x 670 x 430 mm and about 6.47 kg the 350 RTK is noticeably easier to carry to a tower base or into a plant than a 9.74 kg aircraft of roughly one metre span.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Wet sites.&lt;/strong&gt; The 350 RTK is listed with IP55. Remember that the rating covers the aircraft, not your pod; a custom enclosure needs its own sealing design, as explained in our guide on &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-payload-ip-rating.html" rel="noopener noreferrer"&gt;IP ratings for robot payloads&lt;/a&gt;.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  When to go straight to the Matrice 400
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Camera plus measurement in one flight.&lt;/strong&gt; Pilot view on a Zenmuse H30T while a LiDAR, gas pod or multi-sensor box records underneath.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Onboard AI with a large computer.&lt;/strong&gt; Carrier board, heatsink, fan and enclosure for an Orin-class computer are a real mass, and the 400 absorbs it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;An independent data link.&lt;/strong&gt; Mesh or 5G radios with their antennas and ground plane need space away from the aircraft's own antennas.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Payloads still growing.&lt;/strong&gt; Research and pilot projects almost always add a sensor in the second phase. Starting with 6 kg of headroom avoids a second integration.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  A five-question decision check
&lt;/h2&gt;

&lt;ol&gt;
&lt;li&gt;What is the all-in weight: sensor, bracket, enclosure, computer, cables, and any Zenmuse that stays on?&lt;/li&gt;
&lt;li&gt;What is the peak power of the custom load, and does the chosen port supply it?&lt;/li&gt;
&lt;li&gt;Does the pilot need a Zenmuse view during the same flight?&lt;/li&gt;
&lt;li&gt;What endurance do you need with the payload fitted, measured rather than listed?&lt;/li&gt;
&lt;li&gt;Will another sensor be added within a year?&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;If the answer to question 1 is under about 1.5 kg and questions 3 and 5 are "no", the Matrice 350 RTK is sufficient. Otherwise the Matrice 400 is the safer base, and our guide to &lt;a href="https://www.yuanzhirobot.com/en/guides/dji-matrice-400-custom-payload.html" rel="noopener noreferrer"&gt;custom payloads for the Matrice 400&lt;/a&gt; lists what is commonly integrated on it.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fpelgfsiabd59shtl7kqj.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fpelgfsiabd59shtl7kqj.jpg" alt="Payload design renders from YUANZHI ROBOTICS across robot dogs, UGVs and drones (3D renders)" width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Payload design renders from YUANZHI ROBOTICS across robot dogs, UGVs and drones (3D renders)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Next step
&lt;/h2&gt;

&lt;p&gt;You can compare both aircraft with real component weights in our &lt;a href="https://www.yuanzhirobot.com/?lang=en#config" rel="noopener noreferrer"&gt;online payload configurator&lt;/a&gt;: select the &lt;a href="https://www.yuanzhirobot.com/en/platform/dji-matrice-350-rtk.html" rel="noopener noreferrer"&gt;DJI Matrice 350 RTK&lt;/a&gt; or the &lt;a href="https://www.yuanzhirobot.com/en/platform/dji-matrice-400.html" rel="noopener noreferrer"&gt;DJI Matrice 400&lt;/a&gt;, add sensors and compute, and read the total weight and peak power. YUANZHI ROBOTICS designs and builds &lt;a href="https://www.yuanzhirobot.com/en/service/drone-payload.html" rel="noopener noreferrer"&gt;custom drone payloads&lt;/a&gt; for both, including the bracket, power and data harness, onboard software and the ground-station display. Send us the sensor list and the mission, and we will tell you which airframe it fits and what the budget looks like.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Original guide: &lt;a href="https://www.yuanzhirobot.com/en/guides/dji-matrice-350-vs-matrice-400-custom-payload.html" rel="noopener noreferrer"&gt;https://www.yuanzhirobot.com/en/guides/dji-matrice-350-vs-matrice-400-custom-payload.html&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>drones</category>
      <category>robotics</category>
      <category>hardware</category>
      <category>iot</category>
    </item>
    <item>
      <title>DEEP Robotics M20 wheel-legged robot payloads: what fits and when to choose it</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Thu, 12 Nov 2026 01:12:58 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/deep-robotics-m20-wheel-legged-robot-payloads-what-fits-and-when-to-choose-it-1d3k</link>
      <guid>https://dev.to/yuanzhirobotics/deep-robotics-m20-wheel-legged-robot-payloads-what-fits-and-when-to-choose-it-1d3k</guid>
      <description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; the DEEP Robotics M20 is a mid-size wheel-legged quadruped of about 33 kg with a rated payload of about 15 kg (OEM reference figure). A full inspection set of a Livox Mid-360, a dual-sensor thermal gimbal, a depth camera and a Jetson Orin NX computer uses only around 2.3 kg of sensors plus the compute box and brackets, so weight is rarely the limit. Choose the M20 when the route mixes long flat corridors or roads with stairs and kerbs and you need to cover distance quickly; choose a larger quadruped when the payload itself is heavy or tall.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the M20 is as a payload carrier
&lt;/h2&gt;

&lt;p&gt;The M20 has a driven wheel at the end of each leg. On flat ground it rolls like a small UGV, and at stairs, kerbs and cable trenches it uses its legs. That combination changes what a payload has to cope with: rolling gives a smoother ride for cameras and LiDAR than a trot, but the body still pitches and steps on obstacles, so mounts must be as stiff as on any quadruped.&lt;/p&gt;

&lt;p&gt;The figures we work from in our configurator are about 33 kg robot weight, about 15 kg rated payload and outer dimensions of 820 x 430 x 570 mm. The robot ships with front and rear 96-channel LiDAR for its own obstacle avoidance and navigation. Runtime, speed, sealing and stair limits depend on version and load, so confirm them on the current official datasheet before you size a mission.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fiwljjbprekumu8hrdcjm.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fiwljjbprekumu8hrdcjm.jpg" alt="DEEP Robotics M20 with a YUANZHI ROBOTICS sensor payload (3D render)" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;DEEP Robotics M20 with a YUANZHI ROBOTICS sensor payload (3D render)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What fits: weight and power of typical payload sets
&lt;/h2&gt;

&lt;p&gt;Because the rated payload is generous for the robot's size, the practical limits are deck area, height above the back and electrical power, not kilograms. The table uses component weights and power from manufacturer datasheets; brackets, cables and the compute enclosure come on top.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Payload set&lt;/th&gt;
&lt;th&gt;Main components&lt;/th&gt;
&lt;th&gt;Sensor weight&lt;/th&gt;
&lt;th&gt;Listed power&lt;/th&gt;
&lt;th&gt;Typical use&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Minimal mapping&lt;/td&gt;
&lt;td&gt;Livox Mid-360 + Jetson Orin Nano-class computer&lt;/td&gt;
&lt;td&gt;0.27 kg (LiDAR)&lt;/td&gt;
&lt;td&gt;6.5 W LiDAR + compute&lt;/td&gt;
&lt;td&gt;Own SLAM map and navigation stack, research&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Inspection standard&lt;/td&gt;
&lt;td&gt;Livox Mid-360, Hikvision HM-TD5737T-4/W dual-sensor gimbal, Intel RealSense D435i, Jetson Orin NX 16GB&lt;/td&gt;
&lt;td&gt;About 2.34 kg&lt;/td&gt;
&lt;td&gt;About 29 W sensors + up to 40 W compute&lt;/td&gt;
&lt;td&gt;Thermal and visual inspection points, gauge reading&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Security patrol&lt;/td&gt;
&lt;td&gt;Dual-sensor gimbal, loudspeaker and light, mesh or 5G radio, Mid-360&lt;/td&gt;
&lt;td&gt;Gimbal 2 kg + LiDAR 0.27 kg + audio and radio&lt;/td&gt;
&lt;td&gt;Depends on light and radio&lt;/td&gt;
&lt;td&gt;Perimeter and campus patrol&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Gas detection&lt;/td&gt;
&lt;td&gt;Gas sensor module, small thermal core, Mid-360, 5G module&lt;/td&gt;
&lt;td&gt;Well under 1 kg of sensors&lt;/td&gt;
&lt;td&gt;Low; compute dominates&lt;/td&gt;
&lt;td&gt;Leak patrol in plants and utility corridors&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;For the inspection standard set the arithmetic is simple: 265 g for the Mid-360, 2 kg for the gimbal and 75 g for the D435i give about 2.34 kg, roughly 16% of the 15 kg rating. Even after adding a machined base plate, a sealed compute box and harness you normally stay below half of the rating, which leaves margin for stairs. See our &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-payload-weight-budget.html" rel="noopener noreferrer"&gt;weight and power budget guide&lt;/a&gt; for the full method.&lt;/p&gt;

&lt;h2&gt;
  
  
  Reuse the built-in LiDAR or add your own?
&lt;/h2&gt;

&lt;p&gt;The M20 already carries front and rear 96-channel LiDAR, so the first question is whether to add another one. If the robot's own navigation covers your route, a second LiDAR may not be needed and the payload can be limited to the inspection sensors. A separate Mid-360 on the payload still makes sense in three cases: you want your own SLAM and navigation stack with full access to raw points and timing, you need a survey-quality point cloud of the site as a deliverable, or the gimbal and mast would block part of the built-in sensors' view. Whichever you choose, keep added parts out of the built-in LiDAR field of view; a tall mast directly in front of a sensor creates permanent false obstacles.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" alt="LiDAR-inertial mapping trajectory recorded by a robot dog" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;LiDAR-inertial mapping trajectory recorded by a robot dog&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Mounting rules specific to a wheel-legged body
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Keep it low and centred.&lt;/strong&gt; When rolling fast and braking, a high payload shifts load onto the front or rear wheels. Put the compute box and any battery low on the back, and the gimbal only as high as its view requires.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Check leg clearance in the stair posture.&lt;/strong&gt; Legs fold differently when climbing than when rolling. Side-mounted antennas and overhanging plates that clear the legs on flat ground can collide on stairs.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Design for both vibration types.&lt;/strong&gt; Rolling produces high-frequency vibration from the surface; stepping produces low-frequency shocks. A gimbal needs a stiff bracket; small boards and connectors need locking and strain relief.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Protect the underside and wheels from cables.&lt;/strong&gt; Any harness that hangs below the deck edge will eventually meet a wheel.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Plan power early.&lt;/strong&gt; Confirm the voltage and current available at the robot's payload port, then size DC-DC converters for the gimbal, computer and radio. Our &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-payload-power-distribution.html" rel="noopener noreferrer"&gt;power distribution guide&lt;/a&gt; covers conversion and protection.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" alt="Robot dog with a payload climbing stairs" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Robot dog with a payload climbing stairs&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  When to choose the M20, and when not
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Situation&lt;/th&gt;
&lt;th&gt;Better fit&lt;/th&gt;
&lt;th&gt;Why&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Long corridors, roads or halls with some stairs and kerbs&lt;/td&gt;
&lt;td&gt;M20&lt;/td&gt;
&lt;td&gt;Rolls on the flat sections, steps over the obstacles&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payload above roughly 10 kg, or a robot arm&lt;/td&gt;
&lt;td&gt;Unitree B2-W (about 40 kg rated payload) or a tracked UGV&lt;/td&gt;
&lt;td&gt;More payload margin and a larger deck&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Rough natural terrain, deep mud, loose rubble&lt;/td&gt;
&lt;td&gt;Legged X30 (about 20 kg rated payload) or B2&lt;/td&gt;
&lt;td&gt;Wheels at the feet help little off hard ground&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Flat floors only, no stairs&lt;/td&gt;
&lt;td&gt;Wheeled UGV such as AgileX Scout Mini or Tracer&lt;/td&gt;
&lt;td&gt;Simpler and easier to carry large sensors&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Tight indoor spaces and light sensors only&lt;/td&gt;
&lt;td&gt;Smaller quadruped such as Go2 EDU or Lite3&lt;/td&gt;
&lt;td&gt;Lighter robot, easier to transport and deploy&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;The M20's position is therefore clear: at about 33 kg it is far lighter than a B2-W at about 85 kg, so it is much easier to transport, yet it carries enough to hold a complete inspection payload. It is the natural choice for utility corridors, campuses, logistics sites and plants where most of the route is flat but not all of it.&lt;/p&gt;

&lt;h2&gt;
  
  
  Software work to expect
&lt;/h2&gt;

&lt;p&gt;Hardware is the smaller part of an M20 project. Typical software items are: drivers and time synchronisation for the added sensors, a decision on whose navigation runs the robot (the built-in stack or your own), inspection-point logic that stops the robot and points the gimbal, gauge or thermal analysis, and a link to your operator console over 5G or site Wi-Fi. Ask the supplier which control interfaces the M20 version you are buying exposes for secondary development, because that decides how deep custom navigation can go.&lt;/p&gt;

&lt;h2&gt;
  
  
  Next steps
&lt;/h2&gt;

&lt;p&gt;You can assemble an M20 payload and see its weight and peak power in the &lt;a href="https://www.yuanzhirobot.com/?lang=en#config" rel="noopener noreferrer"&gt;payload configurator&lt;/a&gt;, and check the platform data on our &lt;a href="https://www.yuanzhirobot.com/en/platform/deep-robotics-m20.html" rel="noopener noreferrer"&gt;DEEP Robotics M20&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/platform/deep-robotics-x30.html" rel="noopener noreferrer"&gt;DEEP Robotics X30&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/platform/unitree-b2-w.html" rel="noopener noreferrer"&gt;Unitree B2-W&lt;/a&gt; pages. YUANZHI ROBOTICS designs the mounting structure, harness and software for these robots; see &lt;a href="https://www.yuanzhirobot.com/en/service/robot-payload.html" rel="noopener noreferrer"&gt;robot payload integration&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/service/quadruped-development.html" rel="noopener noreferrer"&gt;quadruped development&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/service/robot-autonomy.html" rel="noopener noreferrer"&gt;robot autonomy&lt;/a&gt;. For the larger wheel-legged option, read &lt;a href="https://www.yuanzhirobot.com/en/guides/unitree-b2w-wheeled-quadruped-payload.html" rel="noopener noreferrer"&gt;Unitree B2-W wheeled quadruped payloads&lt;/a&gt;.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Original article: &lt;a href="https://www.yuanzhirobot.com/en/guides/deep-robotics-m20-wheeled-payload.html" rel="noopener noreferrer"&gt;https://www.yuanzhirobot.com/en/guides/deep-robotics-m20-wheeled-payload.html&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>ros</category>
      <category>hardware</category>
      <category>iot</category>
    </item>
    <item>
      <title>Fleet software for mixed robot dog, UGV and drone inspection teams</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Wed, 11 Nov 2026 05:10:30 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/fleet-software-for-mixed-robot-dog-ugv-and-drone-inspection-teams-1i8j</link>
      <guid>https://dev.to/yuanzhirobotics/fleet-software-for-mixed-robot-dog-ugv-and-drone-inspection-teams-1i8j</guid>
      <description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; there is rarely one off-the-shelf product that runs a Unitree or DEEP Robotics quadruped, a Clearpath or AgileX UGV and a DJI drone as one team. What works in practice is a thin common layer: an edge agent on each robot that translates its vendor SDK into a shared mission and telemetry format, one site map with a common coordinate frame, a server that assigns inspection points and stores results, and a single operator view. Most of the effort goes into map alignment, data formats and network behaviour, not into the dashboard.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why mixed fleets need their own software layer
&lt;/h2&gt;

&lt;p&gt;Each platform family ships with a different control interface. Unitree quadrupeds such as the Go2 EDU and B2 are programmed through unitree_sdk2 over DDS. Clearpath UGVs (Husky, Jackal, Warthog) run ROS 2 natively. AgileX bases (Scout, Hunter, Bunker, Tracer) expose a CAN interface with ROS drivers on top. DJI enterprise drones are reached through the Payload SDK on the aircraft, the Mobile SDK on the controller or the Cloud API used by docks. Boston Dynamics Spot uses its own gRPC-based SDK. Vendor fleet tools usually manage only their own robots, so a site that mixes brands ends up with three apps, three maps and three report formats unless something sits above them.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4536sctp701tjhvm0fd.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4536sctp701tjhvm0fd.jpg" alt="Unitree Go2 EDU with a YUANZHI ROBOTICS sensor payload (3D render)" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Unitree Go2 EDU with a YUANZHI ROBOTICS sensor payload (3D render)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What the common layer actually has to do
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;One mission model.&lt;/strong&gt; An inspection point is defined once (position, heading and a sensor action such as a thermal snapshot of a valve or a gauge reading) and each robot type converts it into its own motion: a walking goal for a dog, a Nav2 goal for a UGV, a waypoint with gimbal angles for a drone.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;A shared coordinate frame.&lt;/strong&gt; Ground robots normally localise in a LiDAR map with a local origin; drones fly in GNSS coordinates. You need a surveyed transform between the two, or RTK GNSS on the ground robots outdoors, otherwise "point 14" means a different place for each robot.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Task allocation.&lt;/strong&gt; Stairs and narrow walkways go to the quadruped, long flat routes and heavy sensors to the UGV, roofs, tanks and stacks to the drone. Simple rule-based assignment by capability and battery level covers most sites before any optimisation is needed.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Uniform results.&lt;/strong&gt; Every image, thermal frame, gas reading and gauge value should be stored with the same fields: point ID, robot ID, time, pose, sensor and value. That is what makes trend reports and alarms possible across robots.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Operator handover.&lt;/strong&gt; One screen to see positions, video and alarms, plus a clear rule for who has control when a robot needs manual teleoperation.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Integration layer by platform type
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Platform type&lt;/th&gt;
&lt;th&gt;Typical native interface&lt;/th&gt;
&lt;th&gt;Localisation&lt;/th&gt;
&lt;th&gt;What the edge agent adds&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Quadruped (Go2 EDU, B2, AS2, X30, Lite3)&lt;/td&gt;
&lt;td&gt;Vendor SDK (DDS for Unitree), ROS 2 bridges&lt;/td&gt;
&lt;td&gt;LiDAR map, e.g. SLAM with a Livox Mid-360&lt;/td&gt;
&lt;td&gt;Goal and gait commands, stair and docking states, battery mapping&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Wheeled or tracked UGV (Husky, Jackal, Scout, Bunker)&lt;/td&gt;
&lt;td&gt;ROS 2, or CAN with a ROS driver&lt;/td&gt;
&lt;td&gt;LiDAR map, RTK GNSS outdoors&lt;/td&gt;
&lt;td&gt;Nav2 goals, arm or mast actions, charging dock logic&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Enterprise drone (Matrice 350 RTK, Matrice 400, Matrice 4T)&lt;/td&gt;
&lt;td&gt;Payload SDK, Mobile SDK or Cloud API&lt;/td&gt;
&lt;td&gt;GNSS / RTK&lt;/td&gt;
&lt;td&gt;Waypoint and gimbal translation, geofence checks, media download&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F48b5phf0v3k469yu4az2.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F48b5phf0v3k469yu4az2.jpg" alt="Four-wheel UGV payload design with a mast-mounted LiDAR and gimbal camera (3D render)" width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Four-wheel UGV payload design with a mast-mounted LiDAR and gimbal camera (3D render)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Architecture that holds up on real sites
&lt;/h2&gt;

&lt;p&gt;Keep ROS 2 and DDS traffic inside each robot. DDS discovery over 4G/5G or site Wi-Fi is fragile, so robots talk to the server through a small agent using MQTT or HTTPS for commands and telemetry and WebRTC or RTSP for video. The agent runs on the robot's payload computer; a Jetson Orin NX 16GB module (listed at 157 TOPS) leaves room for SLAM, detection and video encoding at the same time. Results are buffered locally and uploaded when the link returns, so a dead spot in a tunnel does not lose data. For remote sites a 5G module such as the Quectel RM520N-GL (3GPP Release 16) gives a usable uplink, but the mission must still finish safely if the link drops. Our &lt;a href="https://www.yuanzhirobot.com/en/guides/5g-module-for-robot-dog-remote-control.html" rel="noopener noreferrer"&gt;5G remote control guide&lt;/a&gt; covers that part in more detail.&lt;/p&gt;

&lt;h2&gt;
  
  
  Buy, adapt or build
&lt;/h2&gt;

&lt;p&gt;If all robots come from one vendor, start with that vendor's tool. If the fleet is mixed, check whether a commercial fleet platform already has adapters for your exact models and SDK versions; adapters for quadrupeds are less common than for UGVs and drones. Building is justified when you have site-specific tasks (gauge reading on your panel types, gas alarm rules, integration with your maintenance system or SCADA) or when data must stay on-premises. A realistic build keeps the shared layer small and puts robot-specific logic in the agents, so adding a new platform means writing one adapter rather than touching the server.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxy83k4jbesxj2p4071a5.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fxy83k4jbesxj2p4071a5.jpg" alt="DJI Matrice 350 RTK with a YUANZHI ROBOTICS sensor payload (3D render)" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;DJI Matrice 350 RTK with a YUANZHI ROBOTICS sensor payload (3D render)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Common mistakes
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Defining inspection points separately in each vendor app, which makes cross-robot reports impossible later.&lt;/li&gt;
&lt;li&gt;Skipping the map-to-GNSS transform, then finding that drone images and ground images of the same asset never line up.&lt;/li&gt;
&lt;li&gt;Treating charging as an afterthought; docking behaviour differs per platform and decides how many missions per day you really get.&lt;/li&gt;
&lt;li&gt;Streaming every camera at full quality to the server instead of sending events and keeping raw data on the robot until it is needed.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Where we fit
&lt;/h2&gt;

&lt;p&gt;YUANZHI ROBOTICS builds payloads and software for quadrupeds, UGVs and drones, including edge agents, mapping and navigation, inspection-point logic and operator consoles. You can sketch the hardware for each robot in the &lt;a href="https://www.yuanzhirobot.com/?lang=en#config" rel="noopener noreferrer"&gt;payload configurator&lt;/a&gt;, review platform pages such as &lt;a href="https://www.yuanzhirobot.com/en/platform/unitree-go2-edu.html" rel="noopener noreferrer"&gt;Unitree Go2 EDU&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/platform/clearpath-husky-a200.html" rel="noopener noreferrer"&gt;Clearpath Husky A200&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/platform/dji-matrice-350-rtk.html" rel="noopener noreferrer"&gt;DJI Matrice 350 RTK&lt;/a&gt;, and read about our &lt;a href="https://www.yuanzhirobot.com/en/service/robot-autonomy.html" rel="noopener noreferrer"&gt;robot autonomy&lt;/a&gt;, &lt;a href="https://www.yuanzhirobot.com/en/service/inspection-robot.html" rel="noopener noreferrer"&gt;inspection robot&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/service/drone-payload.html" rel="noopener noreferrer"&gt;drone payload&lt;/a&gt; services. For deciding which robot should handle which job, see &lt;a href="https://www.yuanzhirobot.com/en/guides/ugv-vs-robot-dog-inspection.html" rel="noopener noreferrer"&gt;UGV vs robot dog for inspection&lt;/a&gt;.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Originally published at &lt;a href="https://www.yuanzhirobot.com/en/guides/multi-robot-fleet-software-inspection.html" rel="noopener noreferrer"&gt;yuanzhirobot.com&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>ros</category>
      <category>drones</category>
      <category>automation</category>
    </item>
    <item>
      <title>Clearpath Jackal vs AgileX Scout Mini for indoor research and inspection</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Tue, 10 Nov 2026 09:12:32 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/clearpath-jackal-vs-agilex-scout-mini-for-indoor-research-and-inspection-2045</link>
      <guid>https://dev.to/yuanzhirobotics/clearpath-jackal-vs-agilex-scout-mini-for-indoor-research-and-inspection-2045</guid>
      <description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; pick the &lt;strong&gt;Clearpath Jackal&lt;/strong&gt; when you need more payload headroom (about 20 kg rated), a smaller 508 × 430 mm footprint and IP62 sealing for occasional outdoor runs. Pick the &lt;strong&gt;AgileX Scout Mini&lt;/strong&gt; when your sensor stack is light (its rating is about 10 kg) and you value the optional mecanum-wheel version that can slide sideways between shelves or equipment racks. For a typical LiDAR, depth camera and Jetson payload, both platforms have enough capacity; the decision usually comes down to footprint, sealing, drive type and how much you expect the payload to grow.&lt;/p&gt;

&lt;h2&gt;
  
  
  Specifications side by side
&lt;/h2&gt;

&lt;p&gt;The figures below are the manufacturer reference values we use in our payload configurator. Always confirm against the latest OEM datasheet before ordering, because battery and drive options change between production batches.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Clearpath Jackal&lt;/th&gt;
&lt;th&gt;AgileX Scout Mini&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Dimensions (L × W × H)&lt;/td&gt;
&lt;td&gt;508 × 430 × 250 mm&lt;/td&gt;
&lt;td&gt;612 × 580 × 245 mm&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Robot weight&lt;/td&gt;
&lt;td&gt;about 17 kg&lt;/td&gt;
&lt;td&gt;about 23 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Rated payload&lt;/td&gt;
&lt;td&gt;about 20 kg&lt;/td&gt;
&lt;td&gt;about 10 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payload-to-weight ratio&lt;/td&gt;
&lt;td&gt;about 1.2&lt;/td&gt;
&lt;td&gt;about 0.43&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Protection&lt;/td&gt;
&lt;td&gt;IP62 (OEM listed)&lt;/td&gt;
&lt;td&gt;check current OEM spec&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Drive options&lt;/td&gt;
&lt;td&gt;4-wheel skid steer&lt;/td&gt;
&lt;td&gt;4-wheel skid steer, optional mecanum wheels&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Typical use&lt;/td&gt;
&lt;td&gt;research, mixed indoor and light outdoor&lt;/td&gt;
&lt;td&gt;indoor research, education, flat-floor inspection&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Two numbers stand out. The Jackal is about 100 mm narrower and lighter, yet rated to carry twice as much. The Scout Mini is wider, which helps stability with a tall mast, but its payload rating leaves less margin once you add a battery extension, an arm or a heavy enclosure.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhzaa6p1am1g4438i4oqr.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fhzaa6p1am1g4438i4oqr.jpg" alt="Clearpath Jackal with a YUANZHI ROBOTICS sensor payload (3D render)" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Clearpath Jackal with a YUANZHI ROBOTICS sensor payload (3D render)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What a realistic indoor payload weighs
&lt;/h2&gt;

&lt;p&gt;Most indoor research and inspection builds use the same core parts. Using the component values from our catalog:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Livox Mid-360 LiDAR&lt;/strong&gt;: 265 g, about 6.5 W, 360° × 59° field of view. A good default for LiDAR-inertial SLAM in rooms and corridors.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Intel RealSense D435i&lt;/strong&gt;: 75 g, about 2.5 W, 87° × 58° depth field of view with an IMU. Useful for low obstacles that a LiDAR mounted on top can miss.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;NVIDIA Jetson Orin NX 16GB&lt;/strong&gt; (157 TOPS class module): budget up to about 40 W peak for the module. A packaged unit such as a Seeed reComputer J4012 is a common way to deploy it.&lt;/li&gt;
&lt;li&gt;Alternative LiDARs if you need more vertical coverage or a classic spinning sensor: Ouster OS0 (430 g, about 20 W, 360° × 90°) or Velodyne VLP-16 (830 g, about 8 W).&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Sensors and compute alone are well under 2 kg. The weight that actually decides the platform is the structure: base plate, mast, enclosure, DC-DC converters, cabling and any extra battery. A clean aluminium build typically keeps the whole payload in the low single-digit kilograms, which fits either robot. If you plan to add a small arm, a second camera head or a swappable battery later, the Jackal's 20 kg rating gives you room to grow; on the Scout Mini you should keep a written weight budget from day one. Our guide on &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-payload-weight-budget.html" rel="noopener noreferrer"&gt;calculating payload weight and power budget&lt;/a&gt; shows the method.&lt;/p&gt;

&lt;h2&gt;
  
  
  Footprint and manoeuvring in tight spaces
&lt;/h2&gt;

&lt;p&gt;Indoor inspection often means server-room aisles, plant rooms and lab corridors. The Jackal's 430 mm width passes most doorways and aisles with margin and turns on the spot as a skid-steer robot. The Scout Mini is 580 mm wide, so measure your narrowest aisle first. Its advantage is the mecanum option: the robot can translate sideways to line up a camera with a gauge or rack label without a three-point turn. That is valuable for repeated close-range inspection points, but mecanum wheels prefer clean, flat floors and add more wheel odometry drift, so LiDAR-based localization becomes more important.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fzw8phubc3f21hwe305v9.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fzw8phubc3f21hwe305v9.jpg" alt="AgileX Scout Mini with a YUANZHI ROBOTICS sensor payload (3D render)" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;AgileX Scout Mini with a YUANZHI ROBOTICS sensor payload (3D render)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Software and integration effort
&lt;/h2&gt;

&lt;p&gt;Both manufacturers provide ROS drivers, and both robots are widely used with ROS 2. The integration work is similar on either platform:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Mounting and calibration&lt;/strong&gt;: rigid LiDAR and camera mounts, LiDAR-to-camera and LiDAR-to-base extrinsics, and time synchronization between sensors.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Mapping and localization&lt;/strong&gt;: LiDAR-inertial odometry, map saving and relocalization so the robot can restart a route on an existing map.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Navigation&lt;/strong&gt;: Nav2 or a custom planner, with the controller tuned for skid-steer slip or for holonomic mecanum motion.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Inspection tasks&lt;/strong&gt;: waypoint routes, photo or thermal capture at fixed poses, gauge reading and a simple web or ground-station interface.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;The Scout Mini talks to the onboard computer over CAN, so plan a USB-CAN adapter or a carrier board with CAN. The checklist in our &lt;a href="https://www.yuanzhirobot.com/en/guides/ros2-driver-checklist-robot-sensors.html" rel="noopener noreferrer"&gt;ROS 2 driver checklist for robot sensors&lt;/a&gt; applies to both builds.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" alt="LiDAR-inertial mapping trajectory recorded by a robot dog; the same mapping approach applies to wheeled bases" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;LiDAR-inertial mapping trajectory recorded by a robot dog; the same mapping approach applies to wheeled bases&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Which one to choose
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Choose the Jackal&lt;/strong&gt; if your payload may grow beyond about 5 kg, you need IP62 sealing, you work in narrow aisles under 550 mm, or you need occasional outdoor runs on paths and grass.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Choose the Scout Mini&lt;/strong&gt; if your payload is a compact LiDAR, camera and compute stack, your floors are flat and clean, and sideways motion with mecanum wheels would save time at inspection points.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Consider a larger base&lt;/strong&gt; such as a Clearpath Husky A200 or AgileX Hunter 2.0 if you need longer outdoor routes, a PTZ camera on a mast or a heavier battery.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fv5r2ozayhedk1u3qkgjh.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fv5r2ozayhedk1u3qkgjh.jpg" alt="UGV payload kit with a robot arm, an example of how payloads grow beyond sensors alone (3D render)" width="800" height="571"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;UGV payload kit with a robot arm, an example of how payloads grow beyond sensors alone (3D render)&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Next step
&lt;/h2&gt;

&lt;p&gt;You can lay out sensors, compute and comms on either robot and see the total weight and power in our &lt;a href="https://www.yuanzhirobot.com/?lang=en#config" rel="noopener noreferrer"&gt;payload configurator&lt;/a&gt;. Platform details are on the &lt;a href="https://www.yuanzhirobot.com/en/platform/clearpath-jackal.html" rel="noopener noreferrer"&gt;Clearpath Jackal&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/platform/agilex-scout-mini.html" rel="noopener noreferrer"&gt;AgileX Scout Mini&lt;/a&gt; pages. For a complete build including mounting, wiring and mapping and navigation software, see our &lt;a href="https://www.yuanzhirobot.com/en/service/ugv-payload.html" rel="noopener noreferrer"&gt;UGV payload&lt;/a&gt; and &lt;a href="https://www.yuanzhirobot.com/en/service/robot-autonomy.html" rel="noopener noreferrer"&gt;robot autonomy&lt;/a&gt; services; for an indoor build on the Scout Mini specifically, read &lt;a href="https://www.yuanzhirobot.com/en/guides/agilex-scout-mini-indoor-mapping-inspection-payload.html" rel="noopener noreferrer"&gt;Scout Mini payload for indoor mapping and inspection&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Is the Clearpath Jackal or the AgileX Scout Mini better for indoor inspection?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Both work indoors. The Jackal has the higher rated payload (about 20 kg vs about 10 kg) and the smaller footprint, so it suits heavier sensor stacks. The Scout Mini suits lighter LiDAR-plus-camera payloads and offers an optional mecanum-wheel version for omnidirectional moves in narrow aisles.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;How much payload can a Scout Mini carry compared with a Jackal?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;AgileX rates the Scout Mini at about 10 kg and Clearpath rates the Jackal at about 20 kg. A typical mapping and inspection payload with a Livox Mid-360, a RealSense D435i, a Jetson Orin NX computer, mounting plate and enclosure usually stays within a few kilograms, so it fits either platform.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Can both robots run ROS 2 navigation?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Yes. Both platforms have ROS drivers from their manufacturers, so LiDAR SLAM, localization and Nav2 can be built on either. The integration work is mostly in sensor mounting, time synchronization, extrinsic calibration and tuning the controller for skid-steer or mecanum motion.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Which one is easier to take outdoors?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The Jackal is listed with IP62 protection and is often used on light outdoor terrain. The Scout Mini is mainly used on indoor floors and paved surfaces; check the current OEM specification for its protection rating before planning outdoor work.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Who can build a sensor payload for a Jackal or Scout Mini?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;A payload integrator such as YUANZHI ROBOTICS can design the mounting structure, power and wiring, integrate LiDAR, cameras and compute, and deliver mapping, navigation and inspection software for either platform.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Originally published at &lt;a href="https://www.yuanzhirobot.com/en/guides/clearpath-jackal-vs-agilex-scout-mini.html" rel="noopener noreferrer"&gt;https://www.yuanzhirobot.com/en/guides/clearpath-jackal-vs-agilex-scout-mini.html&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>ros</category>
      <category>hardware</category>
      <category>iot</category>
    </item>
    <item>
      <title>Unitree B2-W wheeled quadruped payloads: when wheels-on-legs beat a UGV</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Tue, 10 Nov 2026 05:11:10 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/unitree-b2-w-wheeled-quadruped-payloads-when-wheels-on-legs-beat-a-ugv-4nlj</link>
      <guid>https://dev.to/yuanzhirobotics/unitree-b2-w-wheeled-quadruped-payloads-when-wheels-on-legs-beat-a-ugv-4nlj</guid>
      <description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; The Unitree B2-W is the B2 chassis with powered wheels instead of feet. On hard floors and paved roads it rolls instead of walking, which means higher cruising speed, far lower energy per kilometre, much less vibration into the payload and longer rounds per charge; when it meets stairs, kerbs or rubble it locks the wheels and walks like a B2. That combination is why it competes less with the walking B2 than with wheeled and tracked UGVs: it covers UGV-type distances on a plant road and still gets into the building. The payload rail, power connector and SDK are the same as the B2, so a B2 payload bolts on without redesign; what changes is the mission profile you should build for: longer routes, fewer stops, more emphasis on mapping and gimbal stabilisation at speed, and a battery budget that favours the robot's own endurance.&lt;/p&gt;

&lt;p&gt;A wheel-legged robot is not a compromise between a dog and a car; it is a dog that can stop walking when walking is wasteful. The Unitree B2-W puts powered wheels where the B2 has feet. On a plant road, a warehouse floor or a corridor it rolls, fast and smooth; at a staircase, a kerb or a pile of debris it stands up on its legs and climbs. For a sensor payload this changes less about the mounting and more about the mission you should design for. This guide covers what is the same as the B2, what is different, where the B2-W beats a UGV and where it does not, and which payloads suit it.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fi2rwabz2a22f2xuggsz6.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fi2rwabz2a22f2xuggsz6.jpg" alt="3D render of a YUANZHI inspection payload on a Unitree B2-W: same rail and power interface as the B2, lower vibration on hard floors" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a YUANZHI inspection payload on a Unitree B2-W: same rail and power interface as the B2, lower vibration on hard floors&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Same as the B2: interface, power, software
&lt;/h2&gt;

&lt;p&gt;The B2-W shares the B2 body, battery system, back payload rail, payload connector and SDK. A payload designed for the B2 bolts on without a new base plate, the power budget is the same, and the navigation and inspection software talks to the same motion interface with additional wheel and hybrid modes. If you already run a B2 payload, the B2-W is a drop-in platform; if you are choosing between the two, the choice is about terrain and distance, not about integration effort.&lt;/p&gt;

&lt;h2&gt;
  
  
  Different: speed, energy, vibration, stairs
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;B2 (walking)&lt;/th&gt;
&lt;th&gt;B2-W (rolling, walking when needed)&lt;/th&gt;
&lt;th&gt;Tracked or wheeled UGV&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Cruise on hard floor&lt;/td&gt;
&lt;td&gt;Walking pace, higher energy per km&lt;/td&gt;
&lt;td&gt;Rolling, several times faster, far lower energy per km&lt;/td&gt;
&lt;td&gt;Rolling, similar or faster&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Stairs and kerbs&lt;/td&gt;
&lt;td&gt;Yes&lt;/td&gt;
&lt;td&gt;Yes, by switching to legs&lt;/td&gt;
&lt;td&gt;No (kerbs sometimes)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Vibration into payload&lt;/td&gt;
&lt;td&gt;Gait-induced, periodic&lt;/td&gt;
&lt;td&gt;Low when rolling; gait vibration only on legs&lt;/td&gt;
&lt;td&gt;Low on floor, high on tracks over rubble&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Route length per charge&lt;/td&gt;
&lt;td&gt;Shortest&lt;/td&gt;
&lt;td&gt;Long on hard surface&lt;/td&gt;
&lt;td&gt;Long&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payload mass class&lt;/td&gt;
&lt;td&gt;Tens of kg&lt;/td&gt;
&lt;td&gt;Same as B2&lt;/td&gt;
&lt;td&gt;Tens to hundreds of kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Price class&lt;/td&gt;
&lt;td&gt;Industrial quadruped&lt;/td&gt;
&lt;td&gt;Above B2&lt;/td&gt;
&lt;td&gt;Wide range, often lower&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Two consequences matter for payload design. First, rolling on a smooth floor is quieter mechanically than any gait, so LiDAR scans and gimbal video are cleaner and localization drift per kilometre is lower; you can afford a lighter gimbal stabiliser than on a walking dog. Second, the robot spends less of its battery on moving, so a long inspection route that would need two B2 charges may fit in one B2-W charge. Both push the design towards longer routes with more inspection points rather than towards more sensors.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fg2l8ljmu6j2lh7jmegyr.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fg2l8ljmu6j2lh7jmegyr.jpg" alt="The walking B2 with the same payload: identical interface, different mission profile" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;The walking B2 with the same payload: identical interface, different mission profile&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  When the B2-W beats a UGV
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Mixed sites&lt;/strong&gt;: a plant with roads between buildings and stairs inside them. A UGV covers the roads and stops at the first staircase; a walking dog does everything slowly; the B2-W rolls the roads and climbs the stairs.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Thresholds and small obstacles&lt;/strong&gt;: door sills, cable trays, hoses, a kerb at the loading bay. These end a UGV round or need ramps; the B2-W steps over them.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Long patrol loops&lt;/strong&gt; with a few hard segments, such as a perimeter road with a gravel bank or a drainage channel to cross.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Delivery inside a campus&lt;/strong&gt;: speed on the paths, stairs at the destination, and a smooth ride for the box.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  When a UGV is the better buy
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;The site is flat and the robot never needs to leave one level. A Scout, Hunter, Husky or Bunker is cheaper, carries more and is simpler to maintain.&lt;/li&gt;
&lt;li&gt;The payload is heavy or needs a stable base for manipulation, such as a robot arm or a large sampling system.&lt;/li&gt;
&lt;li&gt;The environment is deep mud, snow or loose sand where tracks win over both wheels and legs.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fv5r2ozayhedk1u3qkgjh.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fv5r2ozayhedk1u3qkgjh.jpg" alt="The UGV alternative: more payload and lower cost on flat ground, but no stairs" width="800" height="571"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;The UGV alternative: more payload and lower cost on flat ground, but no stairs&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Payloads that suit the B2-W
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Long-route inspection kit&lt;/strong&gt;: dual-spectrum PTZ gimbal, Mid-360 LiDAR for mapping and localization, gas module, lights, 4G or 5G router with an external antenna, payload computer, and a dock at the start of the loop. This is our standard B2 kit; on the B2-W we usually add a longer-range LiDAR option for faster mapping of outdoor roads.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Mapping payload&lt;/strong&gt;: a Mid-360 or a longer-range sensor plus RTK, taking advantage of the smooth rolling motion for cleaner point clouds over large areas.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Delivery box&lt;/strong&gt;: a locked compartment with door sensors and a tablet, where the smooth ride and speed are the point.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Security patrol&lt;/strong&gt;: lights, speaker, thermal camera and a long loop per charge.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Design notes specific to wheels
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Ground clearance and splash&lt;/strong&gt;: rolling throws more water and dust at the underside than walking. Keep connectors on the top of the payload and route cables away from the wheel arcs.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Speed and camera exposure&lt;/strong&gt;: at rolling speed a gimbal needs faster shutter settings for gauge reading; set the inspection points to be reached at walking speed or stopped, and let the robot roll between them.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Antenna height&lt;/strong&gt;: the same rules as on the B2, with RTK antennas at least 30 cm apart and clear of the gimbal; the wheels do not change the geometry.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Mode switching in software&lt;/strong&gt;: the navigation stack has to decide where to roll and where to walk. We mark walking zones (stairs, thresholds, rough patches) on the map so the switch is deterministic rather than reactive.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" alt="Stairs are where the B2-W earns its price over a UGV: it switches to legs and climbs with the payload on board" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Stairs are where the B2-W earns its price over a UGV: it switches to legs and climbs with the payload on board&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What we deliver
&lt;/h2&gt;

&lt;p&gt;We build and fit payloads for the B2 and B2-W on the same interface, install our mapping, navigation and inspection software, mark the roll and walk zones on your site map, and commission the first rounds with your operator. If you are deciding between a B2, a B2-W and a UGV, send us a floor plan or a site photo set with the route you have in mind; the decision is usually obvious once the stairs, thresholds and distances are on one drawing.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Originally published at &lt;a href="https://www.yuanzhirobot.com/en/guides/unitree-b2w-wheeled-quadruped-payload.html" rel="noopener noreferrer"&gt;https://www.yuanzhirobot.com/en/guides/unitree-b2w-wheeled-quadruped-payload.html&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>ros</category>
      <category>hardware</category>
      <category>iot</category>
    </item>
    <item>
      <title>Custom payloads for the DJI Matrice 400: what can be integrated and how</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Tue, 10 Nov 2026 01:14:39 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/custom-payloads-for-the-dji-matrice-400-what-can-be-integrated-and-how-5hl3</link>
      <guid>https://dev.to/yuanzhirobotics/custom-payloads-for-the-dji-matrice-400-what-can-be-integrated-and-how-5hl3</guid>
      <description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; The DJI Matrice 400 is listed at about 9.74 kg with roughly 6 kg of payload and up to 59 minutes of flight time, which is enough to carry a Zenmuse gimbal and a second, custom payload at the same time: a 3D LiDAR, a gas-sensing pod, a loudspeaker, an onboard AI computer or a long-range mesh radio. What decides success is not the headline payload but four things: a balanced and vibration-isolated mount, a power and data path the drone accepts, time-synchronised data, and software that turns raw sensor output into a result your team can use.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the Matrice 400 already gives you
&lt;/h2&gt;

&lt;p&gt;Before adding anything, list what is built in. The Matrice 400 has its own RTK positioning and obstacle-sensing radar, and it carries DJI Zenmuse gimbals natively. If a Zenmuse H30T, L2 or P1 already answers your question, buy that and stop: DJI payloads are calibrated, supported in DJI Pilot and FlightHub, and need no custom work. Custom integration is worth it when you need a sensor DJI does not make, an onboard computer, your own data path, or several sensors flying together.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Matrice 350 RTK&lt;/th&gt;
&lt;th&gt;Matrice 400&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Aircraft weight&lt;/td&gt;
&lt;td&gt;about 6.47 kg&lt;/td&gt;
&lt;td&gt;about 9.74 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Rated payload&lt;/td&gt;
&lt;td&gt;about 2.7 kg&lt;/td&gt;
&lt;td&gt;about 6 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Size (unfolded, mm)&lt;/td&gt;
&lt;td&gt;810 x 670 x 430&lt;/td&gt;
&lt;td&gt;980 x 1020 x 470&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Typical custom use&lt;/td&gt;
&lt;td&gt;one gimbal or one light sensor pod&lt;/td&gt;
&lt;td&gt;gimbal plus a second payload, or a heavier combined pod&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Values are manufacturer listings as recorded in our payload catalog; check DJI's current specification before final design.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fitx2eo0dd56augoxxfhu.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fitx2eo0dd56augoxxfhu.jpg" alt="3D render of a DJI Matrice 400 with a YUANZHI ROBOTICS under-body sensor pod and a gimbal camera at the front" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a DJI Matrice 400 with a YUANZHI ROBOTICS under-body sensor pod and a gimbal camera at the front&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  What people actually integrate
&lt;/h2&gt;

&lt;p&gt;These are the payload classes we are asked about most often, with real weights and power from our component catalog. Weights are for the sensor itself and exclude bracket, cables and enclosure.&lt;/p&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Payload class&lt;/th&gt;
&lt;th&gt;Example parts&lt;/th&gt;
&lt;th&gt;Weight&lt;/th&gt;
&lt;th&gt;Power&lt;/th&gt;
&lt;th&gt;Why on a drone&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Compact 3D LiDAR&lt;/td&gt;
&lt;td&gt;Livox Mid-360 / Hesai JT16 / Hesai XT32M2X&lt;/td&gt;
&lt;td&gt;265 g / 200 g / 490 g&lt;/td&gt;
&lt;td&gt;6.5 W / 4.3 W / 10 W&lt;/td&gt;
&lt;td&gt;low-altitude mapping under canopy, in tunnels or around structures where a survey LiDAR is overkill&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Gas sensing&lt;/td&gt;
&lt;td&gt;ION Science MiniPID 2 (VOC) + Winsen ZCE04B (4-in-1)&lt;/td&gt;
&lt;td&gt;40 g + 50 g&lt;/td&gt;
&lt;td&gt;low (sensor level)&lt;/td&gt;
&lt;td&gt;leak screening along pipelines, landfills and plant fences&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Thermal core for custom pods&lt;/td&gt;
&lt;td&gt;Teledyne FLIR Boson 640&lt;/td&gt;
&lt;td&gt;7.5 g&lt;/td&gt;
&lt;td&gt;0.5 W&lt;/td&gt;
&lt;td&gt;radiometric or AI thermal channel in your own housing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Broadcast and light&lt;/td&gt;
&lt;td&gt;DJI Zenmuse V1 / S1&lt;/td&gt;
&lt;td&gt;690 g / 760 g&lt;/td&gt;
&lt;td&gt;–&lt;/td&gt;
&lt;td&gt;security, rescue and crowd guidance&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Onboard compute&lt;/td&gt;
&lt;td&gt;NVIDIA Jetson Orin Nano 8GB / Orin NX 16GB / AGX Orin 64GB module&lt;/td&gt;
&lt;td&gt;module only; add carrier and heatsink&lt;/td&gt;
&lt;td&gt;about 25 W / 40 W / 60 W&lt;/td&gt;
&lt;td&gt;detection, mapping or tracking in flight&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Long-range data link&lt;/td&gt;
&lt;td&gt;Doodle Labs Mini / Silvus SC4200 OEM / Quectel RM520N-GL 5G&lt;/td&gt;
&lt;td&gt;34 g / 116 g / module&lt;/td&gt;
&lt;td&gt;about 5 W / – / –&lt;/td&gt;
&lt;td&gt;raw data off the aircraft, independent of the DJI link&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;As a worked example, a Zenmuse H30T (920 g) plus a Livox Mid-360 (265 g) is about 1.19 kg of sensors, roughly 20% of the rated payload. That leaves real room for a compute box, bracket and cabling, which on a drone often weigh as much as the sensors themselves.&lt;/p&gt;

&lt;h2&gt;
  
  
  Step 1: mount, balance and vibration
&lt;/h2&gt;

&lt;p&gt;A drone payload is a mechanical design problem first. The centre of gravity of the whole payload should sit close to the aircraft's own, or the flight controller spends thrust correcting a constant tilt and flight time drops. LiDARs and IMU-dependent sensors need vibration isolation tuned to the propeller frequencies, but not so soft that the sensor wobbles relative to the RTK antenna, because every millimetre of unmodelled motion becomes error in the point cloud. We design form-fitting brackets around the aircraft's mounting interfaces and keep the sensor's field of view clear of the landing gear and propellers.&lt;/p&gt;

&lt;h2&gt;
  
  
  Step 2: power and data path
&lt;/h2&gt;

&lt;p&gt;There are two routes. The first is DJI's Payload SDK (PSDK) path: the payload talks to the drone through DJI's payload interface, shows up in DJI Pilot, can receive aircraft position and time, and can send small amounts of data back to the pilot. This is the cleanest option for anything the pilot needs to see or control. The second route is an independent power supply and radio for high-bandwidth or power-hungry parts, for example a Jetson AGX Orin or a full-rate LiDAR stream, which would exceed what a payload port is meant to provide. Many real payloads combine both: PSDK for control and position, an isolated rail and a mesh or 5G link for the heavy data.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F0amjx7gd6jt75prusx5j.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F0amjx7gd6jt75prusx5j.jpg" alt="3D render of a payload bay with a battery, power distribution board and fanless compute box, the same layout principles we use for drone payload pods" width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a payload bay with a battery, power distribution board and fanless compute box, the same layout principles we use for drone payload pods&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Step 3: time and position
&lt;/h2&gt;

&lt;p&gt;A custom sensor is only as useful as its timestamps. LiDAR scans, thermal frames and gas readings must share a clock with the aircraft's RTK position and attitude, through PPS and time messages or a hardware trigger. Without that, the map smears and a gas peak cannot be tied to a location. If you need heading from a separate GNSS receiver, a dual-antenna module such as the Unicore UM982 weighs under 2 g on its board; the antennas and their separation are the real design constraint.&lt;/p&gt;

&lt;h2&gt;
  
  
  Step 4: software and ground station
&lt;/h2&gt;

&lt;p&gt;Most of the engineering hours go here. Typical work includes a ROS 2 driver set for the added sensors, onboard detection on the gimbal video, LiDAR-inertial mapping, and a ground-station view that overlays results on a map and exports reports. For pipeline or tunnel work we usually treat the drone as one asset alongside robot dogs and UGVs, which is covered in our guide on &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-dog-tunnel-inspection-payload.html" rel="noopener noreferrer"&gt;tunnel inspection payloads&lt;/a&gt;. For a choice between an integrated drone and a payload drone, see &lt;a href="https://www.yuanzhirobot.com/en/guides/dji-matrice-4t-vs-payload-drone.html" rel="noopener noreferrer"&gt;Matrice 4T or a custom payload drone&lt;/a&gt;.&lt;/p&gt;

&lt;h2&gt;
  
  
  Common mistakes
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Budgeting only sensor weight and forgetting bracket, cables, enclosure and cooling.&lt;/li&gt;
&lt;li&gt;Mounting a LiDAR where the landing gear or a propeller crosses its field of view.&lt;/li&gt;
&lt;li&gt;Running a high-power computer from the payload port instead of a sized, fused supply.&lt;/li&gt;
&lt;li&gt;Logging sensor data with the computer's clock rather than the aircraft's RTK time.&lt;/li&gt;
&lt;li&gt;Choosing a custom sensor when a calibrated Zenmuse payload would have done the job.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Next step
&lt;/h2&gt;

&lt;p&gt;To see how sensors fit on the airframe, open the &lt;a href="https://www.yuanzhirobot.com/?lang=en#config" rel="noopener noreferrer"&gt;3D payload configurator&lt;/a&gt; and pick the Matrice 400, or read the &lt;a href="https://www.yuanzhirobot.com/en/platform/dji-matrice-400.html" rel="noopener noreferrer"&gt;Matrice 400 platform page&lt;/a&gt; and compare it with the &lt;a href="https://www.yuanzhirobot.com/en/platform/dji-matrice-350-rtk.html" rel="noopener noreferrer"&gt;Matrice 350 RTK&lt;/a&gt;. Our &lt;a href="https://www.yuanzhirobot.com/en/service/drone-payload.html" rel="noopener noreferrer"&gt;drone payload integration&lt;/a&gt; service covers bracket design, wiring, PSDK work and the ground station, and &lt;a href="https://www.yuanzhirobot.com/en/service/robot-autonomy.html" rel="noopener noreferrer"&gt;robot autonomy&lt;/a&gt; covers mapping and detection software. Send us your sensor list and mission and we will reply with a feasibility check and a function list.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Original article: &lt;a href="https://www.yuanzhirobot.com/en/guides/dji-matrice-400-custom-payload.html" rel="noopener noreferrer"&gt;Custom payloads for the DJI Matrice 400: what can be integrated and how&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>drones</category>
      <category>robotics</category>
      <category>hardware</category>
      <category>iot</category>
    </item>
    <item>
      <title>DEEP Robotics X30 vs Unitree B2 for industrial inspection payloads</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Mon, 09 Nov 2026 05:11:40 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/deep-robotics-x30-vs-unitree-b2-for-industrial-inspection-payloads-56nj</link>
      <guid>https://dev.to/yuanzhirobotics/deep-robotics-x30-vs-unitree-b2-for-industrial-inspection-payloads-56nj</guid>
      <description>&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; Both robots are IP67, rated for -20 to 55 degrees C and weigh about 56-60 kg, so the choice comes down to payload margin, payload power and software access. The Unitree B2 publishes more than 40 kg walking payload, over 4 hours with 20 kg, four 24 V and four 12 V outputs, four gigabit Ethernet ports and a documented SDK (unitree_sdk2, unitree_ros2). The DEEP Robotics X30 is rated for 20 kg continuous payload and about 2.5 hours loaded, but ships with autonomous docking and recharging, open-grate stair handling and, on the X30 Pro, 5/12/24 V and USB outputs. Pick the B2 for heavy or power-hungry payloads and custom software; pick the X30 Pro for unattended rounds where docking matters more than payload weight.&lt;/p&gt;

&lt;p&gt;The DEEP Robotics X30 and the Unitree B2 are two industrial quadrupeds often compared as the base for a heavy inspection payload: a pan-tilt with thermal and zoom cameras, a gas detector, an extra LiDAR and a payload computer. On paper they look similar. Both weigh about 60 kg, both are IP67 and both are rated from -20 to 55 degrees C. The differences that matter to an integrator are payload margin, runtime with load, the power and data ports on the back, and how much of the robot you can reach from your own software. This guide puts the manufacturers' public figures side by side and says plainly where a figure is not published.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fg2l8ljmu6j2lh7jmegyr.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fg2l8ljmu6j2lh7jmegyr.jpg" alt="3D render of a Unitree B2 carrying an inspection payload with pan-tilt camera, LiDAR and payload computer on the back rail" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a Unitree B2 carrying an inspection payload with pan-tilt camera, LiDAR and payload computer on the back rail&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Specifications side by side
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;DEEP Robotics X30 / X30 Pro&lt;/th&gt;
&lt;th&gt;Unitree B2&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Weight with battery&lt;/td&gt;
&lt;td&gt;About 56 kg (X30), about 59 kg (X30 Pro)&lt;/td&gt;
&lt;td&gt;About 60 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Standing size&lt;/td&gt;
&lt;td&gt;About 1000 x 470 x 695 mm (X30), 1000 x 470 x 715 mm (X30 Pro)&lt;/td&gt;
&lt;td&gt;About 1098 x 450 x 645 mm&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payload&lt;/td&gt;
&lt;td&gt;20 kg continuous; separate maximum not published&lt;/td&gt;
&lt;td&gt;More than 40 kg walking, 120 kg or more standing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Runtime&lt;/td&gt;
&lt;td&gt;About 4 h unloaded, about 2.5 h loaded&lt;/td&gt;
&lt;td&gt;Over 5 h / 20 km unloaded, over 4 h / 15 km with 20 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Battery&lt;/td&gt;
&lt;td&gt;Capacity not published; field swap; about 1.5 h charge&lt;/td&gt;
&lt;td&gt;58 V, 45 Ah (2250 Wh), quick battery change&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Docking and recharge&lt;/td&gt;
&lt;td&gt;Autonomous return and recharge listed&lt;/td&gt;
&lt;td&gt;Autonomous charging listed as optional&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Ingress protection&lt;/td&gt;
&lt;td&gt;IP67&lt;/td&gt;
&lt;td&gt;IP67&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Operating temperature&lt;/td&gt;
&lt;td&gt;-20 to 55 degrees C&lt;/td&gt;
&lt;td&gt;-20 to 55 degrees C&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Stairs and steps&lt;/td&gt;
&lt;td&gt;Max step 20 cm in the spec table; stairs up to 45 degrees and open-grate industrial stairs in the feature text&lt;/td&gt;
&lt;td&gt;Continuous 20 to 25 cm steps; 40 cm step forward; obstacles up to 40 cm&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Slope&lt;/td&gt;
&lt;td&gt;30 degrees in the spec table&lt;/td&gt;
&lt;td&gt;Above 45 degrees&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Speed&lt;/td&gt;
&lt;td&gt;2.5 m/s maximum operating speed (4.5 m/s top tested speed)&lt;/td&gt;
&lt;td&gt;Over 6 m/s in special configurations, speed limited in normal use&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Power to payloads&lt;/td&gt;
&lt;td&gt;X30: 72 V external interface. X30 Pro: 24 V, 12 V, 5 V (USB)&lt;/td&gt;
&lt;td&gt;4 x 12 V, 4 x 24 V, 1 x 5 V, 1 x battery output&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Data to payloads&lt;/td&gt;
&lt;td&gt;X30: Ethernet. X30 Pro: Ethernet, USB 2.0, USB 3.0&lt;/td&gt;
&lt;td&gt;4 x gigabit Ethernet, 4 x USB 3.0&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Onboard computing&lt;/td&gt;
&lt;td&gt;Two (X30) or three (X30 Pro) octa-core industrial processors&lt;/td&gt;
&lt;td&gt;Intel Core i5 for the platform, Core i7 for user development; Orin NX and others optional&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Built-in sensors&lt;/td&gt;
&lt;td&gt;4 LiDARs, 1 wide-angle camera&lt;/td&gt;
&lt;td&gt;1 3D LiDAR, 2 depth cameras, 2 optical cameras (configuration dependent)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Software access&lt;/td&gt;
&lt;td&gt;x30_motion_sdk (C++, Python) on GitHub; official ROS package not found&lt;/td&gt;
&lt;td&gt;unitree_sdk2 (CycloneDDS based, with B2 examples) and unitree_ros2 (lists B2)&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Price&lt;/td&gt;
&lt;td&gt;Not published; quoted by distributors&lt;/td&gt;
&lt;td&gt;Not published; quoted by distributors&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;All figures are from the manufacturers' public product pages and repositories. Real performance depends on configuration, payload and site conditions. Neither publishes current or wattage limits per payload output.&lt;/p&gt;

&lt;h2&gt;
  
  
  Payload weight and runtime
&lt;/h2&gt;

&lt;p&gt;Add up the full payload before choosing: pan-tilt head, thermal and zoom cameras, gas module, back-mounted LiDAR, computer, brackets, cabling and any weather cover. If that total is, for example, 15 kg, on the X30 it uses most of the published 20 kg, and the published loaded runtime of about 2.5 hours sets the length of a round. On the B2 the same payload sits well under the 40 kg walking figure, and Unitree publishes over 4 hours with 20 kg on board. If the payload will grow over the project, for example by adding a manipulator or a second camera head, the B2 leaves more room.&lt;/p&gt;

&lt;p&gt;The X30 compensates with autonomous docking and recharging as a listed feature and a charge time of about 1.5 hours. For a fixed route that runs several short rounds a day without staff, a shorter round plus reliable docking can be worth more than a longer battery.&lt;/p&gt;

&lt;h2&gt;
  
  
  Power and data on the back
&lt;/h2&gt;

&lt;p&gt;For the payload builder, the B2's published list is the more convenient one: regulated 12 V and 24 V outputs in quantity, four gigabit Ethernet ports and four USB 3.0 ports. A pan-tilt head, LiDAR, computer and gas module can each take a separate rail and port without an extra power board.&lt;/p&gt;

&lt;p&gt;On the X30 family, pick the version carefully. The base X30 lists only a 72 V external interface and Ethernet, so the payload needs its own DC-DC stage and switch. The X30 Pro lists 24 V, 12 V and 5 V outputs plus USB 2.0 and USB 3.0, which is closer to what most payloads expect. In both cases ask the manufacturer for the current limit on each output, because it is not on the public page.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm5pzovr1lto2r09x1vk2.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm5pzovr1lto2r09x1vk2.jpg" alt="3D render of a DEEP Robotics X30 with an inspection payload; the payload power board converts the robot supply to the rails the sensors need" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a DEEP Robotics X30 with an inspection payload; the payload power board converts the robot supply to the rails the sensors need&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Terrain, stairs and weather
&lt;/h2&gt;

&lt;p&gt;Both robots are IP67 and rated from -20 to 55 degrees C, so rain, dust and normal outdoor temperatures are not the deciding factor. Your payload usually is: a pan-tilt head, gas sensor or connector rated lower than IP67 sets the real limit for the whole system.&lt;/p&gt;

&lt;p&gt;For stairs, DEEP Robotics highlights open-grate industrial stairs and stairs up to 45 degrees in its feature text, while its spec table lists a 20 cm step and a 30 degree slope. Unitree lists continuous 20 to 25 cm steps, 40 cm single steps and slopes above 45 degrees. Plants with steep metal stairs and grating should test both robots on their own stairs with the payload fitted, since a high payload moves the centre of mass.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" alt="Stair climbing with an inspection payload fitted; test on the site's own stairs before choosing a base" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Stair climbing with an inspection payload fitted; test on the site's own stairs before choosing a base&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Software and secondary development
&lt;/h2&gt;

&lt;p&gt;Unitree documents unitree_sdk2, a CycloneDDS-based SDK with B2 examples in the repository, and unitree_ros2, which lists Go2 and B2 with low-level control and state examples. The B2 also carries a user-development computer. That makes it straightforward to run your own navigation, inspection tasks and data upload.&lt;/p&gt;

&lt;p&gt;DEEP Robotics publishes x30_motion_sdk on GitHub, a motion-control SDK with C++ and a Python version, and offers mapping, navigation and docking in its own software on the X30 Pro. We did not find an official X30 ROS package, so a ROS 2 based payload usually needs a small bridge on the payload computer. If your project depends on the manufacturer's navigation and docking stack rather than your own, that is less of an issue.&lt;/p&gt;

&lt;h2&gt;
  
  
  Price and availability
&lt;/h2&gt;

&lt;p&gt;Neither manufacturer lists a price on its product page. Both robots are sold through international distributors and quoted per configuration, and the price changes with sensors, docking station and software options. Request quotes for the exact configuration, including the X30 Pro rather than the base X30 if you need low-voltage payload outputs.&lt;/p&gt;

&lt;h2&gt;
  
  
  Decision table by site type
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Site&lt;/th&gt;
&lt;th&gt;Suggested base&lt;/th&gt;
&lt;th&gt;Reason&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Substation&lt;/td&gt;
&lt;td&gt;Either; X30 Pro for docked unattended rounds, B2 for heavy pan-tilt and long rounds&lt;/td&gt;
&lt;td&gt;Flat yards, fixed routes, gauge and thermal reading; docking and runtime decide&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Tunnel or utility corridor&lt;/td&gt;
&lt;td&gt;B2&lt;/td&gt;
&lt;td&gt;Long linear routes favour runtime and payload margin for lighting, LiDAR and gas sensing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Chemical plant&lt;/td&gt;
&lt;td&gt;Test both on site; X30 Pro if grated stairs dominate&lt;/td&gt;
&lt;td&gt;Metal stairs and grating; note that neither public page lists an explosion-proof rating&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Outdoor yard or storage area&lt;/td&gt;
&lt;td&gt;B2&lt;/td&gt;
&lt;td&gt;Slopes, kerbs and longer distances; payload room for RTK and weather protection&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;For either robot we supply the payload side: brackets, power conversion, sensors, the payload computer and the inspection software. Send us the robot model, the inspection task, the site type and the delivery country, and we will tell you which base fits and what the payload will weigh before you order.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Originally published at &lt;a href="https://www.yuanzhirobot.com/en/guides/deep-robotics-x30-vs-unitree-b2.html" rel="noopener noreferrer"&gt;https://www.yuanzhirobot.com/en/guides/deep-robotics-x30-vs-unitree-b2.html&lt;/a&gt;.&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>hardware</category>
      <category>ros</category>
      <category>iot</category>
    </item>
    <item>
      <title>Multi-floor navigation for robot dogs: stairs, elevators and map switching</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Sun, 08 Nov 2026 06:01:11 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/multi-floor-navigation-for-robot-dogs-stairs-elevators-and-map-switching-11ha</link>
      <guid>https://dev.to/yuanzhirobotics/multi-floor-navigation-for-robot-dogs-stairs-elevators-and-map-switching-11ha</guid>
      <description>&lt;p&gt;&lt;em&gt;By the engineering team at YUANZHI ROBOTICS. We design and build sensor payloads for robot dogs, UGVs and drones. Originally published on our site; AI-assisted draft, reviewed by our engineers.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;A robot dog that patrols one floor needs one map. A robot dog that patrols a building needs to know which floor it is on, how to get to the next one, and how to pick up its position again after it arrives. This guide covers map structure, floor detection, stair and lift transitions, map switching, failure handling and what the building owner must provide.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyjwi1p3jd9jgfkxq2i79.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fyjwi1p3jd9jgfkxq2i79.jpg" alt="A robot dog at a lift lobby: the elevator is a transition between two floor maps, not part of either map" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;A robot dog at a lift lobby: the elevator is a transition between two floor maps, not part of either map&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Per-floor maps or one 3D map
&lt;/h2&gt;

&lt;p&gt;There are two common ways to describe a building to the navigation stack.&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;One 2D map per floor, linked by transitions.&lt;/strong&gt; Each floor has its own occupancy grid and its own coordinate frame. A small graph records where the floors connect: the foot and head of each staircase, each lift door. The planner finds a route through that graph, then plans normally inside each floor. It is light on computing and easy to edit.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;One 3D map of the whole building.&lt;/strong&gt; A single LiDAR point cloud covers every floor and the stairwells between them, and the robot localises in 3D. Stairs become ordinary terrain and there is no explicit switch. The cost is more memory and computing and harder editing, and a moving lift car still cannot be represented as static geometry.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;In practice most deployments end up with a hybrid: a 3D map is built during the survey, then sliced into per-floor layers for planning, with the staircases kept as 3D segments. Lifts are always handled as transitions, because the robot is stationary relative to the car while the car moves.&lt;/p&gt;

&lt;h2&gt;
  
  
  Knowing which floor the robot is on
&lt;/h2&gt;

&lt;p&gt;A planar localiser can match equally well on two floors with the same layout, so the floor is tracked as explicit state, not inferred from scan matching alone. Sources, used together:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Transition bookkeeping.&lt;/strong&gt; The robot knows it entered a staircase on floor 2 heading up, so it expects floor 3 at the top.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Height from LiDAR-inertial odometry.&lt;/strong&gt; The vertical component of the odometry gives the change in height across a staircase and is dependable over one flight.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Barometer.&lt;/strong&gt; Pressure change gives relative height over the few minutes of a lift ride. It drifts with weather and ventilation, and lift shafts and stairwells can be pressurised, so it confirms a floor change rather than defining the floor.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Map matching with distinctive features.&lt;/strong&gt; After arrival, the robot scores its scan against each candidate floor map. Floors that look identical need distinguishing features near the arrival point, or a fiducial marker at each landing and lift lobby.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Stairs as a transition
&lt;/h2&gt;

&lt;p&gt;Stairs are handled as a separate mode, and manufacturer limits are the first check. Boston Dynamics, for example, publishes a maximum step height of 300 mm and slopes of plus or minus 30 degrees for Spot, and states stair compatibility as stairs that meet US building code, typically a 7 in rise for a 10 to 11 in run. For other models, use the figures in the manufacturer's own datasheet; where a stair specification is not published, test on the actual staircase before committing to a route.&lt;/p&gt;

&lt;p&gt;A stair transition usually runs in four steps:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Approach.&lt;/strong&gt; The robot drives to a recorded pose at the foot or head of the flight, square to the first step.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Detection and alignment.&lt;/strong&gt; The perception stack confirms the staircase in the depth or LiDAR data, measures the first step edge and aligns the heading to the stair axis. A robot that reaches the staircase but does not detect it should stop, not guess.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Gait mode.&lt;/strong&gt; The robot switches to its stair or climbing gait and reduces speed, and the navigation layer stops planning freely: the robot follows the stair axis. Obstacle checks stay active for people on the stairs.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Landing handover.&lt;/strong&gt; At the top or bottom the robot steps onto the landing, returns to its normal gait, switches the active map to the new floor and relocalises from the known landing pose.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Spiral stairs, open risers, wet steps and gratings are checked on site; unsuitable staircases are marked as blocked transitions.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" alt="Stair transition: approach pose, step detection, stair gait, then map switch on the landing" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Stair transition: approach pose, step detection, stair gait, then map switch on the landing&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Elevators
&lt;/h2&gt;

&lt;p&gt;A lift poses three problems the robot cannot solve by walking: calling the car, choosing the floor, and knowing that the car has arrived at the right floor with the doors open.&lt;/p&gt;

&lt;h3&gt;
  
  
  Calling the lift
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Through the lift controller.&lt;/strong&gt; The robot, or a server it talks to, sends a call and destination request to the elevator system and receives status back: car position, door state, arrival. Some lift manufacturers offer this as a product; KONE, for example, publicly describes a Service Robot API for connecting cleaning, delivery and security robots to its elevators, alongside an Elevator Call API. Access terms and interface details are agreed with the lift company and are not public in full. For older lifts, an IoT interface board wired to the call buttons and floor signals by the lift maintainer gives a similar result.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;With an arm or button pusher.&lt;/strong&gt; A robot dog fitted with an arm, or a dedicated pushing mechanism, presses the physical buttons and reads the floor indicator with a camera. It needs no work on the lift, but depends on button layout, lighting and crowds, and gives no reliable door or arrival signal.&lt;/li&gt;
&lt;/ul&gt;

&lt;h3&gt;
  
  
  Entering, riding and leaving
&lt;/h3&gt;

&lt;ul&gt;
&lt;li&gt;Wait at a recorded lobby pose clear of the doors.&lt;/li&gt;
&lt;li&gt;Confirm the doors are open from the controller signal where available, and from LiDAR or depth data showing free space inside the car.&lt;/li&gt;
&lt;li&gt;If the car is too full, let it go and call again.&lt;/li&gt;
&lt;li&gt;Enter, turn to face the door and hold a fixed position. While the car moves, the localiser is paused or set to ignore motion, since the scan shows a static box.&lt;/li&gt;
&lt;li&gt;On arrival, confirm the floor from the controller status, the barometer height change and, after leaving the car, map matching in the lobby. Only then switch the active map and continue.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Map switching and relocalisation
&lt;/h2&gt;

&lt;p&gt;Each transition stores a pair of poses: where it leaves one floor and where it arrives on the other. On arrival the navigation stack loads the new floor map, seeds the localiser with the arrival pose, lets it converge over a few scans, and checks the match score against a threshold set during commissioning. If the score is low, the robot holds position, turns slowly to see more structure, or reports for help. Starting a route on the wrong map is the error to design against.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" alt="A per-floor LiDAR map with transition points marked at the staircase and lift lobby" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;A per-floor LiDAR map with transition points marked at the staircase and lift lobby&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Failure handling
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Situation&lt;/th&gt;
&lt;th&gt;Expected behaviour&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Lift does not arrive or call is rejected&lt;/td&gt;
&lt;td&gt;Retry after a set time, then reroute by stairs if allowed, otherwise report and wait&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Doors close while entering or leaving&lt;/td&gt;
&lt;td&gt;Stop, back out to the lobby pose, call again&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Arrived floor does not match the request&lt;/td&gt;
&lt;td&gt;Stay inside, request the correct floor, do not leave the car&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Wireless link lost in shaft or stairwell&lt;/td&gt;
&lt;td&gt;Finish the current transition on board, resume telemetry on the landing&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Low localisation score after map switch&lt;/td&gt;
&lt;td&gt;Hold position, try relocalisation, then ask an operator&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Staircase not detected at approach pose&lt;/td&gt;
&lt;td&gt;Do not climb; mark the transition unavailable and report&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Low battery before a transition&lt;/td&gt;
&lt;td&gt;Do not start stairs or lift; return to a charger on the current floor&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;h2&gt;
  
  
  What the building owner must provide
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Lift interface.&lt;/strong&gt; Agreement with the lift owner and maintainer for an API connection or an interface board, including who installs it and who is responsible for it.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Network coverage.&lt;/strong&gt; Wi-Fi or private cellular in lobbies, stairwells and, where possible, inside the lift car and shaft. Without it the robot must complete every transition on board, offline.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Permitted routes.&lt;/strong&gt; Which staircases and lifts the robot may use, at what hours, and whether it may share a car with people.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Access.&lt;/strong&gt; Stairwell and lobby doors that the robot can pass, either held open or connected to the access control system.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Survey time.&lt;/strong&gt; Time on every floor to build and check maps and to record and test each transition.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;YUANZHI ROBOTICS builds the sensor payloads and software that make this work on robot dogs and UGVs: LiDAR mapping, floor tracking, stair and lift transitions and the operator console. Our entry mapping kit starts from US$2,000, the complete inspection kit from US$8,690, and the Jetson Orin NX 16 GB computer is US$2,290. Send us the robot model, the task, the building or site and the delivery country, and we will tell you what the lift and network side will need before you commit.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Should a multi-floor site use one 3D map or one map per floor?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Most deployments use per-floor 2D maps for planning, linked by a graph of stair and lift transitions, often sliced from a 3D survey map. A single 3D map handles stairs naturally but costs more computing and is harder to edit, and lifts still have to be handled as transitions.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;How does the robot know which floor it is on?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;It keeps the floor as explicit state, updated at each transition. Height change from LiDAR-inertial odometry and a barometer confirms the move, and matching the scan against the new floor map confirms the arrival. Floors with identical layouts need distinctive features or markers at landings and lobbies.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Can a robot dog use a lift without any change to the lift?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Yes, with an arm or button pusher and a camera reading the floor indicator, but it depends on button layout, lighting and crowds and gives no reliable door or arrival signal. For routine operation, an API or interface board connected to the lift controller is the dependable option.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What stairs can a robot dog climb?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;It depends on the model; check the manufacturer's datasheet. Boston Dynamics publishes a 300 mm maximum step height for Spot and stair compatibility with US building code stairs, typically 7 in rise for 10 to 11 in run. Where a model's stair limits are not published, test on the actual staircase.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What happens if Wi-Fi drops in the lift or stairwell?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The robot should finish the current transition on board without the network and resume telemetry once it reaches the landing or lobby. Coverage in lobbies and stairwells is still needed for monitoring, and for the lift API if the lift is called from a server.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;More guides: &lt;a href="https://www.yuanzhirobot.com/en/guides/" rel="noopener noreferrer"&gt;Robot payload guides&lt;/a&gt;. Questions about a specific robot or site: &lt;a href="mailto:lizhongpan@yuanzhirobot.com"&gt;lizhongpan@yuanzhirobot.com&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>iot</category>
      <category>automation</category>
    </item>
    <item>
      <title>Best robot dog for night security patrol: Go2, B2, X30 or Spot compared</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Fri, 06 Nov 2026 18:11:44 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/best-robot-dog-for-night-security-patrol-go2-b2-x30-or-spot-compared-4m8j</link>
      <guid>https://dev.to/yuanzhirobotics/best-robot-dog-for-night-security-patrol-go2-b2-x30-or-spot-compared-4m8j</guid>
      <description>&lt;p&gt;&lt;em&gt;By the engineering team at YUANZHI ROBOTICS. We design and build sensor payloads for robot dogs, UGVs and drones. Originally published on our site; AI-assisted draft, reviewed by our engineers.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Night security patrol asks more of a robot dog than a daytime demonstration does. The robot has to run for most of a shift, keep walking through rain and cold, climb the stairs on the route, find its way in poor light and carry a thermal camera, a light and a speaker without running out of payload. This guide compares the Unitree Go2 EDU, Unitree B2, DEEP Robotics X30 and Boston Dynamics Spot on those points, using only the manufacturers' public specifications, and says where a figure is not published.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm5pzovr1lto2r09x1vk2.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fm5pzovr1lto2r09x1vk2.jpg" alt="3D render of a DEEP Robotics X30 carrying one of our patrol payloads on the back rail" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a DEEP Robotics X30 carrying one of our patrol payloads on the back rail&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  The four robots side by side
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Unitree Go2 EDU&lt;/th&gt;
&lt;th&gt;Unitree B2&lt;/th&gt;
&lt;th&gt;DEEP Robotics X30&lt;/th&gt;
&lt;th&gt;Boston Dynamics Spot&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Weight with battery&lt;/td&gt;
&lt;td&gt;About 15 kg&lt;/td&gt;
&lt;td&gt;About 60 kg&lt;/td&gt;
&lt;td&gt;About 56 kg&lt;/td&gt;
&lt;td&gt;33.8 kg&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Runtime&lt;/td&gt;
&lt;td&gt;About 2 to 4 h (15,000 mAh battery)&lt;/td&gt;
&lt;td&gt;4 to 6 h; over 5 h walking unloaded, over 4 h with 20 kg&lt;/td&gt;
&lt;td&gt;About 4 h unloaded, about 2.5 h loaded&lt;/td&gt;
&lt;td&gt;90 min average&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;IP rating&lt;/td&gt;
&lt;td&gt;Not published&lt;/td&gt;
&lt;td&gt;IP67&lt;/td&gt;
&lt;td&gt;IP67&lt;/td&gt;
&lt;td&gt;IP54&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Operating temperature&lt;/td&gt;
&lt;td&gt;Not published&lt;/td&gt;
&lt;td&gt;-20 to 55 °C&lt;/td&gt;
&lt;td&gt;-20 to 55 °C&lt;/td&gt;
&lt;td&gt;-20 to 55 °C&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Payload&lt;/td&gt;
&lt;td&gt;About 8 kg (maximum about 12 kg)&lt;/td&gt;
&lt;td&gt;Over 40 kg walking&lt;/td&gt;
&lt;td&gt;20 kg continuous&lt;/td&gt;
&lt;td&gt;14 kg maximum&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Stairs and steps&lt;/td&gt;
&lt;td&gt;Stair height not published; climb drop height about 16 cm&lt;/td&gt;
&lt;td&gt;Continuous stairs 20 to 25 cm; 40 cm steps forward&lt;/td&gt;
&lt;td&gt;Stairs up to 20 cm&lt;/td&gt;
&lt;td&gt;Maximum step height 300 mm&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Slope&lt;/td&gt;
&lt;td&gt;Climb angle 40°&lt;/td&gt;
&lt;td&gt;Over 45°&lt;/td&gt;
&lt;td&gt;30°&lt;/td&gt;
&lt;td&gt;±30°&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Low light&lt;/td&gt;
&lt;td&gt;Not published&lt;/td&gt;
&lt;td&gt;Not published&lt;/td&gt;
&lt;td&gt;Manufacturer states operation in zero-light environments, no numeric figure&lt;/td&gt;
&lt;td&gt;Terrain sensing needs more than 2 lux&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Price class&lt;/td&gt;
&lt;td&gt;Lowest of the four; EDU price on request&lt;/td&gt;
&lt;td&gt;Industrial; price on request&lt;/td&gt;
&lt;td&gt;Industrial; price on request&lt;/td&gt;
&lt;td&gt;Highest class; price on request&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Figures are taken from each manufacturer's product page. DEEP Robotics notes that its figures are laboratory results for reference only, and Boston Dynamics notes that runtime varies with payload and environment. Treat every runtime as a best case and plan the shift around the loaded figure.&lt;/p&gt;

&lt;h2&gt;
  
  
  Runtime and shift planning
&lt;/h2&gt;

&lt;p&gt;No robot in this group walks a full eight to twelve hour night on one battery. The B2 has the largest published battery (2,250 Wh) and the longest loaded endurance, so it can cover a long round with only one battery change. The X30 sits next, with about 2.5 hours under load and a published charging time of about 1.5 hours. The Go2 EDU and Spot are better planned as several short rounds with charging in between, or with a spare battery and a docking point on the route. For a patrol, ask what a charging station or battery swap costs before comparing robot prices.&lt;/p&gt;

&lt;h2&gt;
  
  
  Weather and temperature
&lt;/h2&gt;

&lt;p&gt;Outdoor night patrol means dew, rain and cold. The B2 and X30 both publish IP67 and -20 to 55 °C, which is the strongest rating in this comparison. Spot publishes IP54 and the same temperature range, adequate for light rain and dust. Unitree does not publish an IP rating or operating temperature for the Go2, so we treat it as an indoor or covered-area robot unless the site accepts the risk. The payload must match the rating of the robot: an IP67 robot carrying an unsealed camera and computer is only as good as the weakest box.&lt;/p&gt;

&lt;h2&gt;
  
  
  Stairs and route geometry
&lt;/h2&gt;

&lt;p&gt;Walk the route before choosing. The B2 publishes continuous stair climbing at 20 to 25 cm and a 40 cm forward step, and Spot a 300 mm maximum step height. The X30 publishes stairs up to 20 cm, which covers most building stairs but not every older or industrial one. Unitree does not publish a stair height for the Go2. Heavier robots also raise practical questions on a patrol: lift loading, floor loading on mezzanines, and whether one person can recover the robot if it falls.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" alt="Stair climbing on a patrol route: check the real step height and landing size before choosing the robot" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Stair climbing on a patrol route: check the real step height and landing size before choosing the robot&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Navigation in poor light
&lt;/h2&gt;

&lt;p&gt;Most robot dogs localise with LiDAR, which does not need light, but obstacle detection and stair perception often rely on depth or stereo cameras, which do. Boston Dynamics publishes a lighting requirement of more than 2 lux for Spot's terrain sensing, so an unlit route still needs lighting for Spot. DEEP Robotics claims zero-light operation for the X30 and gives no numeric figure. Unitree does not publish a lighting threshold for the Go2 or B2. In practice we plan night patrols on a LiDAR map built offline in daylight, add a back-mounted LiDAR where the built-in sensing is weak, and fit forward lighting for the cameras. Ask the manufacturer for the lowest light level at which stair and obstacle perception is supported before signing.&lt;/p&gt;

&lt;h2&gt;
  
  
  Patrol payloads
&lt;/h2&gt;

&lt;p&gt;A useful night patrol payload usually has four parts:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Thermal camera&lt;/strong&gt; to detect people and hot spots in the dark. At night it is the main sensor, not an extra.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Spotlight or LED illumination&lt;/strong&gt; for the visible camera and for deterrence. Boston Dynamics' Spot Cam 2 includes eight LED lights, a 25x optical zoom PTZ camera and a 640 by 480 thermal camera.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Speaker and microphone&lt;/strong&gt; for two-way talk with an operator. The Spot Cam 2 page does not list audio, so plan a separate audio unit if it is needed.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;PTZ camera&lt;/strong&gt; for zooming on fences, gates and number plates from the route.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;Add a computer for detection and logging, an LTE or mesh radio link, and the mounting, power and cable sealing. With a full patrol payload, a robot with 8 kg capacity has little margin left, while the B2 and X30 have room for heavier, sealed equipment. Unitree, DEEP Robotics and Boston Dynamics do not publish a complete night patrol payload for the Go2 EDU, B2 or X30 on the pages we checked, so these are normally integrated.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fij593bdp4se2dzwltwps.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fij593bdp4se2dzwltwps.jpg" alt="Thermal camera view on a night round: people and warm equipment stand out where the visible camera sees little" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Thermal camera view on a night round: people and warm equipment stand out where the visible camera sees little&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Price class
&lt;/h2&gt;

&lt;p&gt;Only qualitative ranges can be given from public information. Unitree publishes a starting price for the Go2 range, but not for the EDU version, which is sold on request. The B2, X30 and Spot are sold through sales enquiries without list prices on their product pages. The Go2 EDU is the lowest-cost platform of the four by a wide margin; the B2 and X30 are industrial platforms; Spot is in the highest class. On our side, the entry mapping kit starts from US$2,000, the complete inspection kit from US$8,690, and the Jetson Orin NX 16 GB computer is US$2,290.&lt;/p&gt;

&lt;h2&gt;
  
  
  Decision table
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Site and task&lt;/th&gt;
&lt;th&gt;Our usual recommendation&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Indoor building, short rounds, tight budget&lt;/td&gt;
&lt;td&gt;Go2 EDU with thermal, light and back LiDAR&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Outdoor yard, rain and cold, long rounds&lt;/td&gt;
&lt;td&gt;B2 or X30 with a sealed patrol payload&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Heavy payload, steep or rough ground&lt;/td&gt;
&lt;td&gt;B2&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Industrial site with stairs up to 20 cm and dark areas&lt;/td&gt;
&lt;td&gt;X30&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Existing Spot fleet or established Spot software workflow&lt;/td&gt;
&lt;td&gt;Spot with Spot Cam 2 and a separate audio unit&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Full-night coverage on one robot&lt;/td&gt;
&lt;td&gt;Any of the four plus docking or battery swap; none covers a full shift on one battery&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;If you send us the robot model you have or plan to buy, the patrol task, the site and the delivery country, we will tell you which payload and software fit, what the route needs in charging and lighting, and where the published specs leave a question to put to the manufacturer.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Which robot dog lasts a full night patrol?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;None of the four on one battery. Published figures are 4 to 6 hours for the B2, about 2.5 hours loaded for the X30, about 2 to 4 hours for the Go2 EDU and 90 minutes average for Spot. Plan rounds with a docking station or battery swaps.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Which robot dogs are rated for rain?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The Unitree B2 and DEEP Robotics X30 publish IP67, and Boston Dynamics Spot publishes IP54. Unitree does not publish an IP rating for the Go2. The payload must be sealed to the same level as the robot.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Can these robots navigate in the dark?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;LiDAR localisation works without light, but camera-based obstacle and stair perception needs some light. Boston Dynamics lists a terrain-sensing lighting requirement of more than 2 lux for Spot, and DEEP Robotics claims zero-light operation for the X30 without a numeric figure. Unitree does not publish a lighting threshold for the Go2 or B2. Fit lighting and ask the manufacturer.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What payload does a night patrol robot dog need?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;A thermal camera, a spotlight or LED illumination, a speaker and microphone, and usually a PTZ camera, plus a computer and a radio link. Spot Cam 2 combines PTZ, thermal and eight LEDs, but its page does not list audio.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;How much do these robots cost?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The manufacturers do not publish list prices for the Go2 EDU, B2, X30 or Spot on their product pages; all are sold on request. The Go2 EDU is the lowest price class, the B2 and X30 are industrial, and Spot is the highest class.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;More guides: &lt;a href="https://www.yuanzhirobot.com/en/guides/" rel="noopener noreferrer"&gt;Robot payload guides&lt;/a&gt;. Questions about a specific robot or site: &lt;a href="mailto:lizhongpan@yuanzhirobot.com"&gt;lizhongpan@yuanzhirobot.com&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>iot</category>
      <category>automation</category>
    </item>
    <item>
      <title>Livox Mid-360 vs the built-in Unitree L1 LiDAR for mapping on a robot dog</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Thu, 05 Nov 2026 16:50:51 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/livox-mid-360-vs-the-built-in-unitree-l1-lidar-for-mapping-on-a-robot-dog-41mh</link>
      <guid>https://dev.to/yuanzhirobotics/livox-mid-360-vs-the-built-in-unitree-l1-lidar-for-mapping-on-a-robot-dog-41mh</guid>
      <description>&lt;p&gt;&lt;em&gt;By the engineering team at YUANZHI ROBOTICS. We design and build sensor payloads for robot dogs, UGVs and drones. Originally published on our site; AI-assisted draft, reviewed by our engineers.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;Every Unitree Go2 ships with a LiDAR in its head, so the first question on a mapping or inspection project is whether a second LiDAR is needed at all. The short answer is that the built-in Unitree L1 was designed for obstacle avoidance and the Livox Mid-360 for mapping and localization, and most serious projects end up running both. This guide compares the two on the numbers that matter for SLAM on a walking robot, explains where each one fails, and gives a decision table by type of project.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4536sctp701tjhvm0fd.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ff4536sctp701tjhvm0fd.jpg" alt="3D render of a Unitree Go2 with a back-mounted payload; the Livox Mid-360 sits on the rail behind the head, the built-in L1 stays in the head" width="800" height="533"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;3D render of a Unitree Go2 with a back-mounted payload; the Livox Mid-360 sits on the rail behind the head, the built-in L1 stays in the head&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  The two sensors side by side
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;&lt;/th&gt;
&lt;th&gt;Unitree L1 (in the Go2 head)&lt;/th&gt;
&lt;th&gt;Livox Mid-360 (on the back)&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Design purpose&lt;/td&gt;
&lt;td&gt;Obstacle avoidance, terrain right in front of the robot&lt;/td&gt;
&lt;td&gt;Mapping, localization, navigation&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Point rate&lt;/td&gt;
&lt;td&gt;About 21,600 effective points per second&lt;/td&gt;
&lt;td&gt;About 200,000 points per second&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Range&lt;/td&gt;
&lt;td&gt;About 30 m on bright targets, minimum 5 cm&lt;/td&gt;
&lt;td&gt;40 m at 10 percent reflectivity, 70 m at 80 percent, minimum 0.1 m&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Field of view&lt;/td&gt;
&lt;td&gt;360 by 90 degrees, hemispherical, sees the ground close to the robot&lt;/td&gt;
&lt;td&gt;360 by 59 degrees (minus 7 to plus 52), little ground right at the feet&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;IMU&lt;/td&gt;
&lt;td&gt;Built in&lt;/td&gt;
&lt;td&gt;Built in, used by LiDAR-inertial SLAM&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Mounting&lt;/td&gt;
&lt;td&gt;In the head, moves with every step&lt;/td&gt;
&lt;td&gt;Rigid bracket on the back rail&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Weight and power&lt;/td&gt;
&lt;td&gt;Part of the robot&lt;/td&gt;
&lt;td&gt;265 g, about 6.5 W, 9 to 27 V, Ethernet&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Figures are from the manufacturers' public datasheets; real performance depends on surfaces, mounting and software.&lt;/p&gt;

&lt;h2&gt;
  
  
  Why the L1 alone struggles with maps
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;Sparse points.&lt;/strong&gt; With roughly a tenth of the Mid-360's point rate, each scan carries little structure. In long corridors, car parks and open plant halls the scan matcher has too few features to hold its position, and the map stretches or bends.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Head motion.&lt;/strong&gt; The head pitches and rolls with the gait. The robot's state estimate compensates partly, but every error in that compensation lands in the map.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Range.&lt;/strong&gt; Large rooms and outdoor areas put walls and structures beyond what the L1 reliably returns on dark or angled surfaces.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;For a demonstration in an office or a small lab, these limits are acceptable and the L1 builds a usable map. They become a problem as soon as the map must be repeatable across days, the area grows past a few hundred square metres, or inspection points must be revisited within a few centimetres.&lt;/p&gt;

&lt;h2&gt;
  
  
  What the Mid-360 changes
&lt;/h2&gt;

&lt;p&gt;Mounted rigidly on the back, the Mid-360 gives dense scans with its own IMU in the same housing, which is exactly what LiDAR-inertial odometry such as FAST-LIO2 or Point-LIO expects. In our builds the difference shows up as straight corridors staying straight, loops closing without manual help, and localization on a saved map that stays stable through doors and turns. Its range also makes outdoor yards, substations and car parks practical.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" alt="A LiDAR map built with a back-mounted Mid-360; walls and corridors stay straight across the whole site" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;A LiDAR map built with a back-mounted Mid-360; walls and corridors stay straight across the whole site&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Why we usually keep both
&lt;/h2&gt;

&lt;p&gt;The Mid-360's field of view starts at minus 7 degrees, so the floor directly in front of and around the robot is in its blind cone. The L1 sees that area well. The practical split is: Mid-360 for mapping, localization and medium-range obstacles; L1 for steps, kerbs and low obstacles right at the feet. Both feed the navigation costmap.&lt;/p&gt;

&lt;h2&gt;
  
  
  Mounting and integration notes
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Use a machined aluminium bracket with a measured offset to the body frame; vibration from a flexible mount is visible in every map.&lt;/li&gt;
&lt;li&gt;Keep the sensor's upper field clear of antennas and the gimbal, or mask those angles in software.&lt;/li&gt;
&lt;li&gt;Power it from a regulated rail on the payload board, not directly from the robot battery, and give it its own Ethernet port or a small switch shared with the computer.&lt;/li&gt;
&lt;li&gt;Synchronise time between the Mid-360, the payload computer and the robot state, or the IMU fusion degrades.&lt;/li&gt;
&lt;li&gt;Build the final map offline with loop closure and dynamic object removal rather than exporting the live map.&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" alt="Stairs and kerbs: the head L1 covers the floor right in front of the robot, the Mid-360 covers the rest" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Stairs and kerbs: the head L1 covers the floor right in front of the robot, the Mid-360 covers the rest&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Decision table
&lt;/h2&gt;

&lt;div class="table-wrapper-paragraph"&gt;&lt;table&gt;
&lt;thead&gt;
&lt;tr&gt;
&lt;th&gt;Project&lt;/th&gt;
&lt;th&gt;Recommendation&lt;/th&gt;
&lt;/tr&gt;
&lt;/thead&gt;
&lt;tbody&gt;
&lt;tr&gt;
&lt;td&gt;Demo, teaching, small lab&lt;/td&gt;
&lt;td&gt;L1 only is acceptable&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Indoor inspection rounds, repeated daily&lt;/td&gt;
&lt;td&gt;Add Mid-360, keep L1 for close obstacles&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Site or building map as a deliverable&lt;/td&gt;
&lt;td&gt;Mid-360 with offline rebuild&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;Outdoor yards, substations, car parks&lt;/td&gt;
&lt;td&gt;Mid-360; consider RTK for outdoor absolute position&lt;/td&gt;
&lt;/tr&gt;
&lt;tr&gt;
&lt;td&gt;SLAM research&lt;/td&gt;
&lt;td&gt;Mid-360, record raw data from both sensors&lt;/td&gt;
&lt;/tr&gt;
&lt;/tbody&gt;
&lt;/table&gt;&lt;/div&gt;

&lt;p&gt;Our entry mapping kit for the Go2 adds the Mid-360 with a machined bracket, regulated power and cabling, and a ROS 2 mapping setup; the full inspection kit adds cameras, computing and the inspection application. If you tell us the robot, the site size and what the map is for, we will tell you whether the built-in sensor is enough before you buy anything.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Can I build a map with only the Go2's built-in L1 LiDAR?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Yes, for small indoor areas and demonstrations. The point cloud is sparse, the sensor sits in the head and moves with the gait, and long corridors or open halls give it little to lock onto, so maps drift and localization is less repeatable than with a back-mounted Mid-360.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What does the Mid-360 add in numbers?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;From the manufacturers' datasheets: about 200,000 points per second against roughly 21,600 effective points per second for the L1; up to 40 m range at 10 percent reflectivity against about 30 m for bright targets on the L1; a 360 by 59 degree field of view with a built-in IMU; 265 g and about 6.5 W.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Do I lose anything by adding the Mid-360?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Weight and power (around 0.5 kg with mount and cable), a blind cone directly around and below the sensor, and one more device on the Ethernet and power rails. Its lower field of view starts at minus 7 degrees, so it does not see the floor right at the robot's feet: that is what the L1 is good at.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Which SLAM works best with the Mid-360 on a quadruped?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;LiDAR-inertial odometry such as FAST-LIO2 or Point-LIO using the Mid-360's own IMU, with an offline pass for loop closure and removal of moving objects to produce the final map. Pure scan matching without the IMU struggles with the vibration of walking.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Where should the Mid-360 be mounted?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;On the back rail, as high and as far forward as the payload allows, tilted slightly forward if the work is mostly indoors, on a rigid aluminium bracket with a known offset to the robot body. A wobbly or 3D-printed bracket shows up directly as map noise.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;More guides: &lt;a href="https://www.yuanzhirobot.com/en/guides/" rel="noopener noreferrer"&gt;Robot payload guides&lt;/a&gt;. Questions about a specific robot or site: &lt;a href="mailto:lizhongpan@yuanzhirobot.com"&gt;lizhongpan@yuanzhirobot.com&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>iot</category>
      <category>automation</category>
    </item>
    <item>
      <title>Mapping payload for robot dogs: LiDAR SLAM hardware setup from sensor to map</title>
      <dc:creator>YUANZHI ROBOTICS</dc:creator>
      <pubDate>Thu, 05 Nov 2026 01:16:20 +0000</pubDate>
      <link>https://dev.to/yuanzhirobotics/mapping-payload-for-robot-dogs-lidar-slam-hardware-setup-from-sensor-to-map-5b10</link>
      <guid>https://dev.to/yuanzhirobotics/mapping-payload-for-robot-dogs-lidar-slam-hardware-setup-from-sensor-to-map-5b10</guid>
      <description>&lt;p&gt;&lt;em&gt;By the engineering team at YUANZHI ROBOTICS. We design and build sensor payloads for robot dogs, UGVs and drones. Originally published on our site; AI-assisted draft, reviewed by our engineers.&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Short answer:&lt;/strong&gt; A mapping payload for a robot dog is a Livox Mid-360 or similar solid-state LiDAR on a short mast at the rear of the back plate, its built-in or an external IMU, a depth camera at the front, and a Jetson Orin NX class computer in a low shell, all on one adapter plate. The LiDAR and IMU must share a clock (PTP or the LiDAR's own IMU) and have a measured extrinsic; the SLAM stack (FAST-LIO2 or LIO-SAM class) runs on the Jetson in real time for localization and the final map is rebuilt offline from the recording with loop closure and dynamic-object removal. The payload weighs 1.5 to 2.5 kg, draws 35 to 45 W and runs from the robot's payload port on Go2, Lite3, AS2 and larger robots alike.&lt;/p&gt;

&lt;p&gt;Building a map is the first thing every autonomous robot dog does at a new site, and the quality of that map decides how well it localizes and navigates afterwards. The software is well known: FAST-LIO2, LIO-SAM and their descendants are open source and run on a Jetson. What decides whether the map comes out clean is the hardware setup around them: which LiDAR and IMU, where they sit, how they share a clock, how the extrinsics are measured, and how the recording is turned into a final map. This guide covers that hardware side.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Ftusl6syiw0bg8e4erq0m.jpg" alt="A LiDAR-inertial map and trajectory recorded by a YUANZHI mapping payload on a quadruped, rebuilt offline with loop closure" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;A LiDAR-inertial map and trajectory recorded by a YUANZHI mapping payload on a quadruped, rebuilt offline with loop closure&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Sensors
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;
&lt;strong&gt;LiDAR&lt;/strong&gt;: a Livox Mid-360 for most work. Its 360° horizontal and 59° vertical field of view covers walls, floor and ceiling from a mast on the robot's back, it has a built-in IMU, it weighs 265 g and draws about 7 W. For long outdoor ranges a 32-line spinning LiDAR sees further at a higher mass and power cost; our comparison guide covers the trade-off.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;IMU&lt;/strong&gt;: the Mid-360's built-in IMU is enough for LiDAR-inertial odometry on a walking robot. An external IMU (a tactical-grade MEMS unit) is added when runs are long and featureless or when the accuracy requirement is tight.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;Depth camera&lt;/strong&gt;: an Intel RealSense D435i or D455 at the front, looking forward and down. It is not part of SLAM; it fills the LiDAR's blind cone close to the robot for obstacle detection and stairs, and gives colour for the map.&lt;/li&gt;
&lt;li&gt;
&lt;strong&gt;GNSS&lt;/strong&gt;: a dual-antenna RTK receiver when mapping outdoors over hundreds of metres, to anchor the map to real coordinates and bound drift.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  Placement
&lt;/h2&gt;

&lt;p&gt;The LiDAR goes on a short mast at the rear of the back plate, high enough that its lower rays clear the robot's body and the payload shell, which on a Go2 class robot means about 10 to 15 cm above the plate. Rear placement keeps the forward view clear and lets the LiDAR see the floor behind and beside the robot. The mast is stiff, not tall: every millimetre of flex is noise in the point cloud. The depth camera sits at the front edge of the shell, pitched down 10 to 20°. The compute module and any external IMU sit inside the shell, the IMU rigidly bolted to the same plate as the LiDAR mast with no isolator between them. Antennas go on the rear corners outside the LiDAR's field of view.&lt;/p&gt;

&lt;p&gt;Vibration isolation is a trade-off. The whole payload sits on isolators between the adapter plate and the robot's back, which protects the sensors from the footfall shock. The LiDAR and IMU are never isolated from each other, because the SLAM stack needs them to move as one body.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fpelgfsiabd59shtl7kqj.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fpelgfsiabd59shtl7kqj.jpg" alt="The same mapping payload family on small and large quadrupeds; the LiDAR sits at the rear on a short mast, the depth camera at the front" width="800" height="500"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;The same mapping payload family on small and large quadrupeds; the LiDAR sits at the rear on a short mast, the depth camera at the front&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Time synchronization
&lt;/h2&gt;

&lt;p&gt;A LiDAR-inertial stack is only as good as the timestamps. With the Mid-360's internal IMU, both streams share the LiDAR's clock and nothing more is needed; the Jetson's clock is used only for recording. With an external IMU, the IMU and LiDAR must be on one time base: the Mid-360 accepts PTP, so the Jetson runs a PTP master and the IMU is timestamped on arrival with a hardware trigger or on a microcontroller synchronized to the same master. A camera that will be fused later is triggered by the LiDAR's PPS or timestamped the same way. We check synchronization on every payload by recording a sharp rotation and confirming that the IMU and LiDAR see it at the same instant to within a millisecond.&lt;/p&gt;

&lt;h2&gt;
  
  
  Calibration
&lt;/h2&gt;

&lt;p&gt;The extrinsic between the LiDAR and the IMU comes from the CAD when the IMU is the LiDAR's own, and is refined with an open-source LiDAR-IMU calibration on a recorded figure-of-eight run when it is external. The LiDAR-to-camera extrinsic is measured with a target board. The LiDAR-to-robot-body transform, which the navigation stack needs, is measured against the robot's own odometry. Calibration is repeated after any mechanical change to the mast and checked before each site deployment by looking at a flat wall in the map: a double surface means the extrinsic or the synchronization has drifted.&lt;/p&gt;

&lt;h2&gt;
  
  
  Compute
&lt;/h2&gt;

&lt;p&gt;A Jetson Orin NX 16 GB on an industrial carrier runs the LiDAR driver, FAST-LIO2 or LIO-SAM at 10 Hz, the depth camera, the navigation stack and recording, with headroom left. A light stack alone (LiDAR driver and odometry at reduced map resolution) runs on the robot's own computer on Go2 EDU and Lite3, which is how trials often start. The recording goes to an NVMe drive in the payload; a mapping run of an hour at 10 Hz is tens of gigabytes of point clouds and IMU data. The Jetson is bolted through a thermal pad to the aluminium shell so that it holds its clock speed in a sealed enclosure; our guide on onboard compute covers the alternatives.&lt;/p&gt;

&lt;h2&gt;
  
  
  From recording to map
&lt;/h2&gt;

&lt;p&gt;The real-time stack gives the operator a live map and a position during the run, which is what the operator needs to see that the site is covered. The final map is not that live map. After the run, the recording is rebuilt offline on the payload computer or on a workstation: loop closure and global optimization remove the accumulated drift, dynamic objects (people, vehicles, doors that were open) are filtered out, the floor is identified, and the result is exported as the point cloud map for localization, a 2D occupancy grid for path planning and a mesh for the operator's view. On our payloads this rebuild runs automatically when the operator stops the recording, so that the robot is ready to localize and navigate against the clean map a few minutes after the mapping walk, without a laptop.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F3xp5g0fx24c4iavfzkpk.jpg" alt="Stairs in the point cloud map; the depth camera fills the LiDAR's close-range blind cone so that stair edges are seen before the robot steps" width="799" height="375"&gt;&lt;/a&gt;&lt;br&gt;
&lt;em&gt;Stairs in the point cloud map; the depth camera fills the LiDAR's close-range blind cone so that stair edges are seen before the robot steps&lt;/em&gt;&lt;/p&gt;

&lt;h2&gt;
  
  
  Before the first mapping run
&lt;/h2&gt;

&lt;ul&gt;
&lt;li&gt;Confirm time synchronization with a sharp rotation and check the IMU and LiDAR timestamps agree.&lt;/li&gt;
&lt;li&gt;Look at a flat wall in the live map from two positions; a double surface means the calibration is off.&lt;/li&gt;
&lt;li&gt;Check the recording drive has space and that recording survives a power cycle.&lt;/li&gt;
&lt;li&gt;Walk a loop that returns to the start and confirm the start point lines up; that is the loop-closure check.&lt;/li&gt;
&lt;li&gt;Walk slowly through doorways and on stairs the first time; a quadruped's pitch on stairs is where odometry is most stressed.&lt;/li&gt;
&lt;li&gt;Keep the LiDAR window clean and check the mast bolts; both are found loose more often than any software fault.&lt;/li&gt;
&lt;/ul&gt;

&lt;h2&gt;
  
  
  What the payload weighs and draws
&lt;/h2&gt;

&lt;p&gt;The complete mapping payload (LiDAR, depth camera, Jetson, GNSS, Wi-Fi, structure and cables) is 1.5 to 2.5 kg and draws 35 to 45 W, which runs from the robot's payload port on Go2, Lite3, AS2, B2 and X30 alike. It is the same payload family across those platforms, so a map built on one robot localizes another.&lt;/p&gt;

&lt;p&gt;If you are planning a mapping payload for a quadruped, tell us the robot, the site (indoor, outdoor or both) and the accuracy you need; we will tell you which LiDAR and IMU we would fit and what the mapping and rebuild workflow looks like for that site.&lt;/p&gt;

&lt;h2&gt;
  
  
  FAQ
&lt;/h2&gt;

&lt;p&gt;&lt;strong&gt;Which LiDAR is right for a mapping payload on a quadruped?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;A Livox Mid-360 for most indoor and near-outdoor work: 360° horizontal field of view, 59° vertical, 40 m range at 10 percent reflectivity, 265 g, about 7 W, and a built-in IMU. For long outdoor ranges a 32-line spinning LiDAR (Ouster OS1, RoboSense Helios) sees further at 2 to 3 times the mass and power. Our LiDAR comparison guide covers the alternatives.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Does the LiDAR need an external IMU?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;The Mid-360's built-in IMU is sufficient for FAST-LIO2 class odometry on a walking robot. An external tactical-grade IMU improves robustness on long featureless corridors and during fast turns, and is what we fit when the customer's accuracy requirement is tight.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Can the robot's own computer run the SLAM stack?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;On Go2 EDU and Lite3 a light stack (LiDAR driver and FAST-LIO2 at reduced map resolution) runs on the robot's computer. Anything that also records, runs a camera or navigates gets a Jetson Orin NX in the payload, which also keeps the robot's controller untouched.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;Why rebuild the map offline?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Real-time odometry drifts over a long run and has no loop closure or dynamic-object removal. Rebuilding from the recorded point clouds and IMU data with loop closure, global optimization and removal of people and vehicles that walked through gives a clean map that the robot then localizes against on every later round.&lt;/p&gt;

&lt;p&gt;&lt;strong&gt;What accuracy does a quadruped mapping payload achieve?&lt;/strong&gt;&lt;/p&gt;

&lt;p&gt;Indoors, with loop closure, a few centimetres of consistency across a building floor. Outdoors over hundreds of metres, decimetre-level drift without GNSS and centimetre-level with RTK GNSS fused in. The walking motion itself is not the limit; the IMU-LiDAR synchronization and extrinsic calibration are.&lt;/p&gt;




&lt;p&gt;&lt;em&gt;More guides: &lt;a href="https://www.yuanzhirobot.com/en/guides/" rel="noopener noreferrer"&gt;Robot payload guides&lt;/a&gt;. Questions about a specific robot or site: &lt;a href="mailto:lizhongpan@yuanzhirobot.com"&gt;lizhongpan@yuanzhirobot.com&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

&lt;p&gt;&lt;em&gt;Written by the engineering team at YUANZHI ROBOTICS. Original article: &lt;a href="https://www.yuanzhirobot.com/en/guides/robot-dog-mapping-lidar-slam-payload.html" rel="noopener noreferrer"&gt;Mapping payload for robot dogs: LiDAR SLAM hardware setup from sensor to map&lt;/a&gt;&lt;/em&gt;&lt;/p&gt;

</description>
      <category>robotics</category>
      <category>ros</category>
      <category>hardware</category>
      <category>iot</category>
    </item>
  </channel>
</rss>
