Artificial intelligence on microcontrollers
Raphael Zingg, Matthias Rosenthal · Zürcher Hochschule für Angewandte Wissenschaften digital collection (Zurich University of Applied Sciences) · 2020
Using artificial intelligence algorithms such as neural networks on microcontrollers offers several possibilities but reveals challenges: limited memory, low computing power and no operating system. In addition, an efficient workflow to port neural networks algorithms to microcontrollers is required. Currently, several frameworks that can be used to port neural networks to microcontrollers are available. This paper evaluates and compares four of them: ”TensorFlow Lite for Microcontrollers” from Google, ”Neural Network on Microcontroller” from Jianjia Ma with the CMSIS-NN backend from ARM, X-CUBE-AI from STMicroelectronics and the e-Ai Solution from Renesas. The frameworks differ considerably in terms of workflow, features and performance. Depending on the application, one framework is better suited than another. Neural networks that are ported to microcontrollers with those frameworks are static. This means that once they are integrated into the firmware they can no longer be adapted. In the context of this work, possibilities of unsupervised domain adaptation learning on microcontrollers was investigated and is discussed.