IoT platforms are often quite capable of gathering data about the physical world they inhabit.
Sensors have the potential to collect information about equipment, vehicles, assets, and environments, movements, position, and more. Connectivity can deliver that information to systems capable of monitoring and analyzing the data.
The limitation is that data is only data.
There's a long road from "information about operations" to "insight to improve operations."
That's where the combination of the Internet of Things and Artificial Intelligence becomes an intriguing prospect.
Often termed AIoT, this architecture connects physical objects and processes to AI/ML algorithms to enable operations teams to derive intelligence from their operations.
Let's begin unpacking that a little further.
IoT Provides the Data Layer
An IoT system typically involves connecting sensors and other physical objects to a digital system.
Here's a drastically oversimplified version of that pipeline:
Physical Asset
↓
Sensors
↓
Connectivity
↓
Data Pipeline
↓
AI / ML
↓
Decision
↓
Physical Action
The first four steps in this dramatically oversimplified example are all about getting information about the physical world into a system where that information can be analyzed.
Sensors attached to objects or equipment could provide insight into how that object is being used or the environment in which it is operating. Position-or-connected devices can share information about their whereabouts or the location of a particular process. Environmental sensors could share data about the physical environment surrounding an object or process.
The information itself becomes the data about the physical world.
AI Adds an Intelligence Layer
Having data is one matter; doing something with that data is another.
Especially when that data represents a large body of information, distilling that data down into something meaningful can be a very time consuming and expensive exercise.
That's where artificial intelligence and machine learning come into play.
Depending on the nature of the data and the desired end-use, models can be created to analyze that data, find patterns within it, find outliers, and perhaps make predictions about the future state of that data.
This enables analysts to focus directly on the information from the data that matters most, typically the information that has implications for their operations.
The difference between IoT and AI is simple: IoT is, "What is happening?" while AI is "What does this mean?"
By adding AI into the mix, we've gone from Visibility to Intelligence to Action.
The Data Pipeline Matters
It's easy to consider an AIoT architecture as simply "sensors + AI," but there's a great deal that goes on between those two endpoints. The data pipeline is critically important to the insights that can be extracted from this type of system.
A reliable, high-quality pipeline ensures that the information coming from your sensors is valuable to your system.
A faulty pipeline can result in poor information being delivered to your AI model, ultimately reducing the value of those insights.
There's a great deal that can go wrong with the data pipeline, even before you get to choosing the right AI model.
Sensor accuracy, connection reliability, data formatting, and system relationships are all variables that play a role in an effective AIoT implementation.
From Prediction to Physical Action
One of the interesting aspects of AIoT is that the "output" of the system doesn't have to end up on a dashboard somewhere.
The intelligence discovered by the AI can inform choices made by people or processes in the physical world, bringing us back to that original architecture diagram.
Let's take a look at that diagram again:
Physical World
↓
Data
↓
AI/ML
↓
Intelligence
↓
Decision
↓
Physical World
This is something of a feedback loop, but the key point is that the input into an AIoT system originates in the physical world while the output from the AI can alter the physical world as well.
That's significantly different from many AI systems built purely within the digital world.
Where Can AIoT Be Applied?
That last diagram may as well be a generic blueprint for an AIoT implementation in any industry.
Manufacturing facilities can provide data about equipment and production. Logistics operations can provide data about vehicles and transport. Even construction sites can provide data about equipment, assets, and overall operations.
Almost any industry that involves substantial physical infrastructure can apply some form of this architecture to gain insight about their operations. There might not be the same level of existing automation in every industry, but there are enough commonalities that there is a substantial opportunity for an AIoT application tailored to the needs of any particular site.
The nature of the implementation will vary from organization to organization based on the type of information they need, the data they have available, and what decisions they need to be made.
Building AIoT Around the Problem
Technology should rarely be the starting point for any discussion about solutions.
That's certainly the case with AIoT applications.
There are a number of questions to ask about the operations for which you're seeking an AIoT-based solution:
• What needs to be measured?
• What data is available?
• How reliable is that data?
• What decisions need to be made?
• What action needs to be taken as a result?
A detailed consideration of those questions will help to frame the technology solution around your business needs.
For those seeking to explore that intersection between IoT devices, AI applications, and industrial operations, Aperture Venture Studio is here to help, with a particular focus on AIoT startups and applications within the industrial space.
Essentially, an AIoT application isn't just about connecting devices. It's about connecting devices → data about those devices → AI insight about that data → decisions based on that information → and changes in the physical world.
Top comments (0)