Observer-based preview control of active vehicle suspensions

Abstract

Feasibility of using preview information in vehicle suspension control is evaluated in this dissertation. Under the assumption of incomplete and noisy measurement condition, we develop design strategies for compensator that gives control action based on available measurement signals. The performance improvement achievable with preview compensators is predicted using an approach based on the optimal control theory. A vehicle is modeled as a system consisting of the rigid body representing main body structure, and the wheels with linear springs, dampers, and actuators. Road inputs and measurement noises are modeled as Gaussian white processes. Using the optimal control theory, the suspension system is optimized with respect to the performance index that trades off between ride quality and handling performance while preserving constraints on suspension rattle space and control input power. Particular emphasis is placed on the optimal use of preview information about the road input in front of the vehicle. Such information is obtained either from look-ahead sensors located in front of a vehicle or by monitoring the motion of the front wheels. The results of numerical simulations for two degrees of freedom and four degrees of freedom vehicle models are presented, which show potential for substantial improvement of performance over the existing designs. While the optimal look-ahead preview compensator improves suspension performance with a moderate computation increase, the optimal wheelbase preview compensator is found to be computationally unaffordable in spite of its control benefit. To relieve its computational burden, the estimation scheme based on a state and input observer is proposed. It is found that although possible performance improvement by the wheelbase preview diminishes a little, the state and input observer is computationally affordable. Optimal preview compensator assumes a constant vehicle speed and a constant preview distance for its derivation. I...