Construction is rapidly evolving into an information-heavy industry.
Data can be collected about a commercial jobsite from equipment telematics, RFID tags, BLE beacons, UWB positioning systems, GPS, sensors, inspections, schedules, deliveries, and labour activity.
However, data collection is just the first step.
It takes much more effort to turn physical world signals into actionable intelligence.
This is where AIoT, or Artificial Intelligence of Things, becomes relevant.
While IoT and AI are considered two different technological domains, AIoT can connect physical-world data collection with analytics, machine learning, edge computing, and enterprise applications.
In the construction sector, it may help build a connection between physical processes on a construction site and data seen by project teams in their digital systems.
Construction jobsite as a distributed data environment
Construction projects are constantly evolving.
Laborers go in and out of zones. Equipment is deployed, moved around, and even put into idleness. Materials are delivered to the site, staged, and gradually used up in the process of building the final product.
Meanwhile, project scheduling, inspection, procurement, and work package data are constantly updated.
Such dynamics create a distributed data environment.
One system typically does not have all the data.
Example:
GPS will help determine location of equipment.
RFID will help identify tagged equipment/materials.
BLE will enable flexible asset tracking.
UWB will help with precise positioning under right conditions.
Telematics will help understand operation of equipment.
IoT sensors will provide additional information about environment/operational status.
Each data source on its own is valuable.
Value appears when they are connected.
From sensor readings to operational context
Sensor readings do not automatically give you an answer.
Let’s say that your equipment tracking system informs that some equipment has been stationary for some hours now.
What's the problem?
Not really.
Equipment may be positioned this way because work package for the equipment hasn't started yet.
Alternatively, equipment may be idling because crew is waiting for materials.
Or this means that equipment has been allocated poorly.
This is where context plays a role.
AIoT solutions will help integrate equipment data with all other operational data to give you a broader picture.
The same applies to workforce and materials.
Position of a worker is more valuable when combined with work zones and project schedule.
A material inventory record is much more valuable when compared with future installation actions.
A milestone is more useful when associated with real field actions.
Telemetry is not just about gathering information.
It is about contextualization.
*Why edge computing is important
*
Construction sites can also generate difficulties for data processing.
The transmission of each sensor event to the cloud platform can generate certain connectivity dependency, latency and data transmission inefficiency.
Edge computing can be an alternative.
The transmission of all data to the cloud platform is not required and a part of it can be processed closer to its origin.
The edge device can filter sensor events, recognize certain conditions, aggregate the information and do selected analytics prior to transmission of the results to centralized systems.
It can be especially important in case when construction sites have intermittent connectivity or the reaction speed is an issue.
Architecture is not required to be either edge or cloud.
Hybrid architecture can let edge systems deal with time-critical tasks and cloud or enterprise platforms process analytics and aggregation.
*Where AI comes into play in the architecture
*
IoT creates the data layer.
AI adds another layer of interpretation.
Think about construction project progress.
Historical production metrics, labour activities, equipment utilization, material inventory levels, inspection results, and scheduling data can all be potentially combined in order to find patterns associated with scheduling or production problems.
The point is not to build a model in place of a project manager.
It can bring insights that may require further investigation by a human.
Like this:
Data: Equipment utilization is decreasing.
Context: A critical work package is getting near.
Extra signal: Required materials have not been delivered to the work zone yet.
Possible insight: Project managers should examine if material delivery could impact future production.
This is one of the key concepts of AIoT:
The more connected operational context AI has access to, the more valuable it gets.
*Integration is the hard part
*
Connecting sensors is generally not as challenging as integrating whatever they produce.
Construction organizations have numerous software tools already, from BIM, through ERP, scheduling, and project controls tools, to procurement, labour management, and safety management systems.
Thus, AIoT architecture has to consider how physical world data would get into existing digital processes.
One such conceptual architecture can look as follows:
Physical layer ➔ Connectivity ➔ Edge processing ➔ Data platform ➔ AI/analytics ➔ Enterprise applications
Each layer has its own purpose.
The physical layer collects the events.
Connectivity moves the information.
Edge computing is able to analyse the selected events on the spot.
The data platform manages and structures the information.
AI and analytics discover the patterns.
Enterprise applications implement the patterns in business processes.
This model is better than AIoT as just "adding AI to IoT."
Data quality is essential anyway.
AI cannot resolve data quality issues automatically.
For example, if sensors were poorly deployed, devices were incorrectly configured, there are duplicate identifiers, or data formats are not compatible with each other, machine learning models could return incorrect results.
As a result, construction AIoT implies dealing with basics:
Reliable sensing
Correct asset identifiers
Timestamps
Appropriate connectivity
Data validation
Secure APIs
Interoperability
Governance
Operational goals
The right technology must be chosen depending on the problem.
Ultra-wideband may be the right choice for some kind of positioning.
GPS may be more appropriate for outdoor tracking of a fleet of vehicles.
RFID may suit better for identification of assets.
LoRaWAN may be helpful in low power sensing across the large territory.
A single sensor technology is not suitable for all construction scenarios.
The next step is connected intelligence, not automation alone
While automation might not be the most promising development within construction technology,
it could well be the connection of previously disconnected operational signals.
Workforce data can help contextualize productivity.
Equipment data can help contextualize utilization.
Material data can help contextualize procurement.
Access data can help contextualize security.
Progress data can help contextualize schedule performance.
With the connection of all of these signals, there would be more data points for AI to analyze and detect possible issues.
However, it does not necessarily mean that each construction site should have thousands of sensors and an elaborate AI system.
What it does mean is that technologies have to be deployed according to specific needs and expected outcomes.
AIoT can create such a connection between the physical construction environment and its management system.
This means that the real question here is not:
"How can we add AI to construction?"
But rather:
"How can we connect the data collected from the physical world with the information necessary for better construction decision making?"
.
Top comments (0)