Bias-Policy Iteration Based Adaptive Dynamic Programming for First-Order Fully Actuated Systems

In this paper, the bias-policy iteration (bias-PI) based adaptive dynamic programming for first-order fully actuated systems is considered. By exploiting the fully actuated property, the bias-PI method for first-order fully actuated systems is revisited, with a more streamlined formulation and a more concise convergence proof provided. The data-driven implementation for the proposed algorithm is introduced by neural networks accordingly. A numerical example verifies the effectiveness of the proposed results.