École Nationale Supérieure de
Techniques Avancées

Résumé

In order to improve the performance of the active inference algorithm, a new theory based on neuroscience called deep active inference is recently applied to autonomous driving. To have better control of the algorithm and the car, the algorithm needs to be transplanted to RTMaps, an environment that allows the execution of real-time software based on components, and the vehicle simulator called SiVIC. This report will present the main idea of the active inference algorithm, explain how it is applied in the field of autonomous driving, discuss the modifications made to the algorithm in order to perform the transplantation, and present the result after the transplantation.