Semi-active adaptive optimal control of vehicle suspension with a magnetorheological damper based on policy iteration

This article attempts to clarify the question of what the essential difference between the optimal and "clipped-optimal" control is that the Hamilton–Jacobi–Bellman or Hamilton–Jacobi–Isaacs partial differential equation is linear or nonlinear. An adaptive optimal control based on policy iteration to the constrained semi-active vehicle suspension system with a magnetorheological damper is presented. The problem of improving the optimal performance of semi-active control suspension system is converted to L2-gain optimal control problem. A two-player policy iteration scheme is employed to solve the Hamilton–Jacobi–Isaacs equation by use of neural networks to approximate optimal policies and value functions in the admissible global region. Simulation results demonstrate that the adaptive optimal controlled semi-active vehicle suspension system outperforms greater than that of the clipped-LQR controlled.