Robust Vehicle Acceleration Control

Abstract

This thesis investigates robust acceleration control for heavy-duty vehicles to improve longitudinal control and driver-assist functions. Despite advances in longitudinal vehicle control, important limitations remain. Existing control strategies may be sensitive to measurement delays, perform poorly for small acceleration errors, and insufficiently reject oscillatory disturbances from driveline windup and tire dynamics. These challenges motivate a model-based control approach that explicitly accounts for the relevant system dynamics to improve tracking performance and disturbance rejection. A third-order state-space driveline model was developed to capture the system’s dominant dynamics, including flexible driveshaft behavior modeled as a springdamper system. The model was tuned and validated against real-world measurements from a battery electric vehicle (BEV), demonstrating close agreement in both speed and acceleration responses across multiple driving scenarios and gear configurations. A Smith predictor was subsequently designed and evaluated to compensate for the system’s inherent dead-time delays. Simulation testing showed that the Smith predictor maintains stability under model and delay mismatches of up to ±50%, while also enabling higher controller gains and faster response times. To further improve performance, the Smith predictor was extended to include measurement updates. An LQG controller with integral action, focusing on robustness properties, was proposed to address driveline oscillation damping and steady-state tracking accuracy. Simulation results show that, given an accurate model, the proposed controller reduces oscillations and improvesmproved steady-state tracking compared to the current controller. Using logged vehicle data, the simulation validated the proposed controllers, showing that model-based control strategies can improve acceleration-tracking performance and robustness compared to the existing controller.