UC Berkeley

Model uncertainty is one of the most challenging areas of control theory. It is the key reason for using feedback control in order to achieve safety and performance. High-gain feedback is the simplest solution to improve performance. However, it comes at the price of instability and constraints violation. In particular, in sliding mode control, it manifests itself as an undesirable chattering. This dissertation addresses this issue by focusing on the development of advanced sliding mode control and demonstrating its effectiveness for autonomous vehicles under the limits of sensing and driving capabilities. First, a new adaptive sliding mode control (ASMC) strategy is proposed to reduce the control action to its minimum possible value while guaranteeing robust stability. Then, integral sliding model predictive control is introduced by merging the concept of ASMC with a robust model predictive control formulation for nonlinear constrained systems. Motivated by the ever-growing interests in autonomous vehicles, we apply the proposed control techniques to control an autonomous vehicle at the limits of handling. Different extreme driving scenarios such as large side slip angle estimation, path tracking with a large model mismatch and drifting maneuvers are considered and solved with the proposed control algorithms. Successful experimental results support the effectiveness of the developed control method.