AIoT: Analyze Your IoT Data
Internet of Things (IoT) technologies have been around for a long time, and the number of IoT applications continues to grow rapidly, from Smart Homes to Smart Buildings to Smart Cities and beyond. With the tremendous volumes of data generated by the exploding number of IoT devices, IoT needs analytics to analyze all that data. This is called AIoT (Artificial Intelligence of Things).
I got into AIoT a number of years ago, when I worked on an Eldercare project. Our goal was to keep elders in their homes as long as possible and as long as prudent. We installed low data-rate sensors in elders’ apartments and in a floor of a nursing home. We collected the IoT data streams and analyzed them with an AI algorithm. [Specifically, an NLP topic model, turned “sideways”. Contact me for more details]. I stored and organized our data in a large graph data store, but that is the topic for a future post.
From our elders’ low-level sensor data, we were able to identify high-level human behaviors. Interestingly, Suzanne and I installed the sensors in the apartment of a gentleman who lived alone. But, looking at our data, our colleague Sam, said, “hold on, there are two people moving around in there”. That’s when we told Sam, he was seeing the gentleman’s dog !
This experience really whet my appetite for AIoT. Partly because this project so directly improved people’s lives. Our efforts kept one couple in their home an extra 7 months. And partly because AIoT analytics are somewhat different than image/video analytics and somewhat different than natural language analytics. But sometimes, techniques from those other fields can be translated over to AIoT (like we did above with topic modeling). This area fascinates me.
Some people suggest that AIoT only applies to AI running on edge devices. While that is certainly a great application, AIoT also applies when sending IoT data to the cloud and running the AI there.
Now I apply AIoT to my clients’ problems. I’ve applied AIoT to analyze elders’ behavior patterns, analyze traffic in large urban cities, analyze large electrical power equipment, improve worker safety in dangerous environments, and analyze computer network traffic patterns. I’ve built classifiers and regression model of many different types, probabilistic (Bayesian) models, Next-Best Action models, topic models, anomaly detectors, time-series predictors, and more.
AIoT encompasses all of these ideas.
If you’re interested in AIoT, subscribe or contact me.