Zero-sum Neuro-optimal Control of Modular and Reconfigurable Robots with External Collisions via Critic Only Policy Iteration

In the cause of deal with this problem of optimal position and velocity tracking regarding modular reconfigurable robots (MRRs) with external collisions, a zero-sum neural-optimal control method ground on critical only policy iteration (COPI) scheme is proposed by adaptive dynamic programming (ADP). The MRR dynamics model uncertainty is compensated by adaptive fuzzy control method. Unique the critic neural network (NN) ground on policy iteration (PI) and ADP method is used as work out hamilton-Jacobi-Issacs (HJI) equation, then the approximated optimal control is obtained. Then Lyapunov's theory proved that the MRR system was stable. At length, simulation results make clear the virtue of the proposed method.