Evolutionary Multi-Objective Optimisation of an Automotive Active Steering Controller

Many real-world engineering design problems involve the satisfaction of multiple conflicting objectives. In this case it is unlikely that a single ideal solution will exist. Instead, the solution of an Multi-Objective Optimisation problem will lead to a family of Pareto optimal points, where any improvement in one objective will result in the degradation of one or more of the other objectives. This paper investigates the use of Evolutionary Multi-objective Optimization (EMO) to optimise the performance of a closed loop feedback Proportional Integral (PI) vehicle yaw controller on a non-linear vehicle. This is done by comparing results against traditional empirical tuning methods relating to rise time, settling time, overshoot and steady-state error. The EMO showed improvement on the original control tuning and also brings light to the difficulty control engineers have with objective interaction for complex problems. I. ACTIVE STEERING FOR PASSENGER VEHICLES Dynamic vehicle control has been an important area of research and development over the past 20 years. This can be seen in the widespread implementation of Electronic Stability Control, Anti-lock Braking Systems and Traction Control Systems onto almost all modern cars. These systems are designed to improve handling and vehicle stability by controlling the vehicle’s dynamic states. Chassis control schemes are a major area of research for control engineers in the automotive sector – primarily motivated by a desire to improve the lateral handling characteristics of vehicles. Many solutions to this lateral stability control problem have been explored in [1]: from simple feed-forward control of yaw rate and lateral acceleration using active front steering to reference model yaw rate feedback control (see