M4E #31: Edge Machine Learning in Manufacturing Automation
Today I’m with some colleagues at the second day of the kickoff meeting of the European Project AI REDGIO 5.0 where our R&D center Gradiant is one of the partners. One of the key points of this project is to evolve cloud AI technologies to AI-at-the-edge during the next 3 years. So, I thought that a good concept to introduce today would be Edge Machine Learning for manufacturing. See the next video from engineering.com to find more about it!
Time for Insights
One thing you may be wondering is: what’s the difference between Cloud, Fog and Edge Computing? One good and easy way to differentiate them is to consider parameters such as distance from the data source to processing servers. Near is edge, fog is in the middle and cloud is far. This also applies to latency. Other things to take into account are computation power or scalability (higher at the cloud, lower in the other cases). You have to take also into account that some people consider fog and edge to be the same. You can read more about it here.
When capturing data from the factory and sending it to the cloud the data cleaning procedure is performed after all this data has been pushed to the servers. One of the advantages of edge AI is that data cleaning and quality improvement algorithms can be deployed nearer to the machines, easing data transmission and processing in the cloud.
Typically, edge analytics makes sense for those use cases where there are near real time requirements. Some examples can be anomaly detection algorithms, computer vision models, cobot and AGV intelligence…