Application of Model-based Control in Autonomous Vehicle Chassis Control and Motion Planning

Automatic control systems are found ubiquitously in today’s automobiles. This dissertation focuses on the development of control systems for autonomous vehicle applications using model-based control design approaches. We start with the control system development for chassis actuators. Firstly, we develop a closed loop torque overlay control system to improve the steering feel of an electric power-assisted steering (EPAS) system. This system has a reference model, a rack force estimator, and a tracking controller. A target steering feel is generated from a reference model with an rack force estimate to reflect the actual vehicle operating conditions. The performance of the proposed control system is evaluated through simulation and a hardware-in-the-loop test. Secondly, we present an EMB clamping force control system is developed that addresses several major challenges in practical implementation. A nonlinear EMB model including a novel clamping force model is introduced. A clamping force estimation and contact detection algorithm is proposed that requires only the existing measurements. Furthermore, a unified architecture is proposed to realize a smooth transition between gap closing and clamping force tracking. The performance of the control system is evaluated based on simulation. In the second part, we developed a hierarchical nonlinear model predictive control (NMPC) framework for autonomous vehicle motion planning and control. At low level, a trajectory tracking NMPC is used to track the reference trajectory given by the highlevel motion planner. A frequency shaped objective function is used to incorporate lower level actuator dynamics. Real-time implementation is realized by a fast NMPC algorithm based on RTI with condensing and control parameterization. Simulation results show that the control system yields good performance under various road conditions when the vehicle is operating at its handling limit. A high-level motion planner is formulated in the NMPC framework incorporating high order control barrier function (HOCBF) based collision avoidance constraint. The motion planner is able to generate dynamically feasible trajectories with respect to various constraints. Simulations of the motion planner and trajectory tracking controller demonstrate the effectiveness of the proposed control framework.Ph.D.College of Engineering & Computer ScienceUniversity of Michigan-Dearbornhttp://deepblue.lib.umich.edu/bitstream/2027.42/170078/1/Yijun Li final dissertation.pdfDescription of Yijun Li final dissertation.pdf : Dissertatio