Bringing Artificial intelligence to Edge Computing

Artificial intelligence – The IOT has evolved in the last several years. IOT has a fascinating evolution. It started way back when sensors famous referred to as “things” where used in the industries as disparate things. Moreover, then IOT as we know today came into existence, the Internet of things. Connecting sensors created this to the Internet. Hence the name “Internet of Things.” This connection led to the creation of what’s called the IOT platforms.

IOT cloud platform was designed to get data from the sensors. It was also easier to analyze the data in the cloud. From this emerged the term edge computing. In Edge, IOT data could be more efficient if it is processed closer to the source, that is the sensors. So there began the edge evolution, the goal was to make these things smarter, which in turn drives decisions faster. For bringing Realtime input for the data from the sensors, it did not make sense to wait for actions from the cloud. The data processing can occur in the Edge and if any anomalies are spotted actions are driven in Realtime. For examples, collision detection, and prevention, air quality control, environmental control. All these IOT applications can be more effective if the actions from the data sensed are driven in real-time. Technologies like edge analytics and edge AI are the next-generation edge computing evolution.

 The Edge initially was used for short term data storage, data processing, and driving action. Most of the data analysis and data crunching to predict what could happen was still at the cloud. Cloud would have all the data model for Machine learning and Artificial intelligence. As edge computing is maturing the need for data training and modeling closer to the Edge becomes a sensible solution. This process would help in predicting and driving action. For instance, a manufacturing setup have the edge IOT implementation, can make location data decisions without relying on the cloud. Most of the data can still stay in the Edge without the need to have a secondary set up in the cloud. This process could save much money. 

The way AI is implemented at the Edge could slightly vary to how its done cloud. In the cloud, it is a centralized database. So, the cloud has access to an enormous volume of data, and it builds its prediction models based on this most extensive set of data. When it comes to Edge, the data might be limited, and data is distributed. One way is to route the data from different edge gateways to collate and create new models. These models created on the edge gateway can be shared with other gateways. The other way is to leverage the cloud, and the models can be built in the cloud and later deployed in the Edge. Whenever a new anomaly is detected, the original data set can be sent to cloud for a new model which can then used all the edge device. For example, in the case of predictive maintenance of large equipment on a factory floor, the edge devices create models from the data. Then use them for prediction and later send it to cloud for historical data. When a new scenario is observed the new set of data can be sent to the cloud for developing a new model.  

Companies are seeing substantial potential opportunities with edge AI. Microsoft’s Azure IOT has edge AI toolkit, which allows the edge devices to deploy the models from cloud to the Edge. Similarly, AWS IOT has a solution for Edge Artifical intelligence. Goggle has tensor flow optimized for Edge. There are also several startups like Mobodexter are developing AI for the Edge.

It is bringing Artificial intelligence to closer to where the actions are actually taking place and significantly improve the decision-making time for the critical time-driven events. It reduces the cost of the overall implementation by reducing how much data is moved to the cloud. Another important aspect is the privacy of the data. By having AI at the Edge the data need not be sent to the cloud, it means one more hole of the data leak is closed.

Footnotes:

  • Mobodexter, Inc based in Redmond, WA builds Internet of things Edge solutions for enterprises applications on Kubernetes Edge Clusters. 
  • Check our Edge Marketplace. Special 6th Anniversary Discount sale of 20% on all products. 
  • IoT Developers can try Edge Platform here
  • Mobodexter’s IoT Edge Blogs: Read
%d bloggers like this: