Eligibility traces and forgetting factor in recursive least-squares-based temporal difference

We propose a new reinforcement learning method in the framework of Recursive Least Squares-Temporal Difference (RLS-TD). Instead of using the standard mechanism of eligibility traces (resulting in RLS-TD((Formula presented.))), we propose to use the forgetting factor commonly used in gradient-based or least-square estimation, and we show that...

Relaxing the control-gain assumptions of DSC design for nonlinear MIMO systems

This work focuses on adaptive neural dynamic surface control (DSC) for an extended class of nonlinear MIMO strict-feedback systems whose control gain functions are continuous and possibly unbounded. The method is based on introducing a compact set which is eventually proved to be an invariant set: thanks to this set, the restrictive...