Embedding intelligence into low-power devices is a difficult task as the lack of computational and memory resources means that only relatively simple solutions can be implemented. Nonetheless, it is possible to build very light reinforcement learning solutions and these can be used to allow a small sensor node to control its connectivity efficiently (e.g. decide when to sleep, or when to transmit more often ).
The students will implement a light weight reinforcement learning algorithm using micro-python on an ESP32 or NORDIC NRF device. The specific algorithm and the problem it will solve can be selected by the students themselves according to their preferences.
10% Theory, 70% Experiments, 20%Writing
Alessandro Chiumento email@example.com