Gain scheduling model predictive path tracking controller for autonomous vehicle on highway scenario

Abstract

The design of the controller tracking path is one of the important factors in the development of autonomous vehicles. One problem for autonomous vehicle operating on highway road must be able to do a satisfactory path tracking so any accidents do not occur. This paper will discuss designing tracking path controller using combination a model predictive controller (MPC), feed forward (FF) and particle swarm optimization (PSO) based on scenario road courses on the highway with several variations of the vehicle speed. The PSO algorithm used to determine optimal weighting gains on the cost function of the MPC and the FF used to reduce the lateral error of the vehicle to the desired trajectory. The approach solves a single adaptive FF-MPC problem for tracking road trajectories. The vehicle model was developed based on 3 DOF non-linear vehicle model. This controller model was developed based on X, Y global position and yaw rate to get input in the form of front steering to the vehicle dynamic system. For path tracking strategy, comparisons with the Stanley controller are done to analyse MPC reliability as non-linear controller in low and middle speed scenario. Simulation results have found that the FF-gain scheduling MPC controller has the significant performance on tracking trajectory at mid and high of the vehicle speeds. In addition, with the using of feed forward and optimal gain weighting on MPC controller made the actuator lifetime is longer than Stanley controller due to reduce the actuator aggressiveness.