Active suspension control with state estimation using finite Markov chains

Abstract

State estimation and optimal control of a nonlinear stochastic MacPherson suspension system using finite state and action Markov chains is considered. A system model and optimal controller are iteratively constructed based on k-means clustering of closed loop data and re-discretisation of the continuous system state space. Bayesian estimation of measured and unmeasured states using a cell filter is considered, and the unscented Kalman filter is considered as an alternative implementation. The main contribution is the introduction of the finite state and action Markov chains to the optimal control design and state estimation in active suspension systems. The application for active suspension control is illustrated and discussed via simulations using a simplified nonlinear model of a MacPherson system including stochastic road and measurement noise.