The collaboration between SensiML and Onsemi combines the Analytics Toolkit development software with Onsemi's RSL10 sensor development kit, providing a perfect platform for edge sensing applications, including industrial process control and monitoring. It supports AI capabilities with small memory footprint and advanced sensing and BLE connectivity provided by the RSL10 platform, eliminating cloud analysis of highly dynamic raw sensor data. A complete machine learning solution for autonomous sensor data processing and predictive modeling.
The kit combines the RSL10 radio with a full-spectrum environmental and inertial motion sensor on a tiny shape factor board that can be easily connected to the toolbox, providing the industry's lowest power CONSUMPTION BLE connection. Although developers have extensive experience in data science and ai, with the RSL10-based platform and SensiML software, they can simply add low-latency local AI prediction algorithms to industrial wearables, process control, robotics or predictive maintenance applications. In addition, smart terminals greatly reduce network traffic by communicating data only when providing valuable insights. The resulting auto-generated code allows intellisense to be embedded in endpoints, translating raw sensor data directly into key insight events and taking relevant actions in real time.
"Cloud-based analytics are too slow, too unreliable, and too far away for the most critical industrial processes. "Using local machine learning and remote cloud learning to analyze the difference between critical events is equivalent to keeping production online, equipment without costly failures, and personnel safe and productive." Said Dave Priscak, Vice president of application engineering at Onsemi
Chris Rogers, CEO of SensiML, said, "Other edge automation solutions rely only on neural network classification models, with only the most basic automation provisions, producing suboptimal code for a given application." "Our comprehensive AutoML model search includes not only neural networks, but also a range of classic machine learning algorithms, as well as subdivision, feature selection and digital signal tuning transformations to deliver the most compact model that meets the performance needs of an application."
