Real-Time Model Predictive Control for Simultaneous Drift and Trajectory Tracking of Autonomous Vehicles

Abstract

Drifting vehicles can accomplish obstacle avoidance at the limits of handling. However, insufficient rear tire lateral force will not suppress the vehicle from spinning if the steering angle and driving torque are unreasonable during trajectory tracking. For this strongly coupled nonlinear system, resolving the contradiction between the real-time performance and the complexity of the controller is challenging. This work proposed a real-time drift trajectory tracking controller (DTTC) for autonomous vehicles under the framework of model predictive control. To improve the computational efficiency and ensure the control performance, DTTC adopts a hierarchical structure, where the upper controller figures out the desired kinematic control inputs and the lower controller swiftly follow these quantities. The multi-objective control law of the upper controller adopts a real-time linearization error model based on a series of nonlinear models, with effective feedback emendation in each control cycle. The high-fidelity simulation results of a typical scenario, figure-8 drift, demonstrate that DTTC has superior transient and steady-state performance to the compared controller, with a 44% reduction in lateral root mean square (RMS) error. The maximum single-step solution time of DTTC is less than 0.035 s.