Nowadays embedded systems are designed to support different Artificial Intelligent (AI) techniques (from classical Machine Learning to Deep Learning) becoming intelligent devices. Such devices are capable to locally processing large amounts of data, reducing the overall power consumption, optimizing the bandwidth usage and preserving the data privacy. As a result, many applications benefit from such integration, from autonomous systems such as robots to medical devices capable of generating local patient’s diagnosis. This thesis intends to evaluate existing technology from a point of view of computational power, integration level, supported AI techniques, constraints,…

1.- Literature study of existing solutions and their limitations.
2.- Evaluation of the most interesting tool flows and existing hardware for a realistic use case.
3.- Implementation of AI on embedded device(s) for a realistic use case.

• Interest in AI and embedded systems
• Experience with Python and C/C++.