Abstract

This thesis intends to design new tracking schemes to enhance the performance and stability of general autonomous vehicles (AVs). Three main types of controllers used for tracking control are investigated.

The geometric controller cannot meet high tracking requirements, and control parameters significantly affect its performance. Therefore, an observer-based nonlinear control combined with a particle swarm optimization (PSO) algorithm is developed for low-speed vehicles to track the pre-determined trajectory accurately. A control law featured with self-tuning gains is designed using the backstepping control technique, for which global asymptotic stability is validated. The PSO evaluates tracking performance through the proposed fitness function and generates optimized tuning parameters with fewer iterations, reducing tuning efforts. Velocity and steering tracking could also be rapidly realized by modifying the error weights of the performance evaluation criterion. Based on the proposed yaw error observer (YEO), the problem of the angle measurements being temporarily inaccurate or unavailable is tackled effectively with the given information.

Further, existing methods can suffer from complex control algorithms and a lack of tracking stability at high speed. The vehicle's motion is decoupled by considering the Frenet frame. A lateral control law based on the linear-quadratic-regulator (LQR) imposes the tracking errors to converge to zero stably and quickly, providing the optimal solution in real-time due to adaptive gains. Regarding the steady-state errors, they are eliminated through the correction of the feedforward term. Besides, the designed double proportional-integral-derivative (PID) controller realizes not only the longitudinal control but also the velocity tracking.