Predicting the consequence of action in digital control state spaces

Emmanuel Daucé

The objective of this dissertation is to shed light on some fundamental impediments in learning control laws in continuous state spaces. In particular, if one wants to build artifi…