Cooperative Differential Game-based Optimal Tracking Control of Modular Reconfigurable Robot with Actor-Critic-Identifier Structure

Based on the cooperative differential game, this paper proposes an optimal control for modular reconfigurable robot (MRR) systems. Describe the dynamical model of MRR in the form of
$n$
joint subsystems, including interconnected dynamic coupling (IDC). Considering
$n$
joint subsystems as
$n$
players in a cooperative game, uncertainties such as IDC and unknown disturbances contained in the dynamic model are approximated by neural network (NN) identifier. An n-player cooperative game theory can be utilized to convert the tracking control issue of MRR into an optimal control task. Solving the Hamilton-Jacobi-Bellman (HJB) equations for optimal control by adaptive dynamic programming (ADP) using actor-critic-identifier (ACI) structure. It is proved that the MRR system tracking error is uniformly ultimately bounded (UUB) by the Lyapunov theory. Finally, simulation experiments are conducted to verify the effectiveness of this method.