Reward inference of discrete-time expert's controllers: A complementary learning approach

Uncovering the reward function of optimal controllers is crucial to determine desired performance that an expert wants inject a certain dynamical system. In this paper, inference algorithm discrete-time expert's proposed. The approach inspired by complementary mechanisms striatum, neocortex, and hippocampus for decision making experience transference. These systems work together infer associated controller using merits data-driven online learning methods. proposed models neocortex system as two independent algorithms given Q-learning gradient identification rule. modelled least-squares update rule extracts relation from states control inputs data. striatum inverse which iteratively finds hidden function. Lyapunov stability theory used show convergence approach. Simulation studies are demonstrate effectiveness algorithm.