Integrated Design of Local Trajectory Planning and Tracking Control for Autonomous Driving

Abstract

This paper describes a real-time local trajectory planning and tracking control algorithm, based on the constrained optimization framework, applicable to autonomous vehicles operating in unstructured environment. The primary novelty is in use of closed-loop information prediction along the given global path in the framework of trajectory planning, i.e. reference velocities and trajectory points used in trajectory planning are generated by low-level tracking control. A continuous-curvature path-smoothing algorithm based on parametric cubic Bezier curves is firstly adopted to smooth the given global path, presented by an ordered sequence of waypoints. An obstacle-free dynamic-feasible trajectory planning problem is carefully formulated by a sequential convex optimization procedure. The solution trajectory points are used for feed-forward and feed-back tracking control. The proposed algorithm was at the core of the planning and control modules for an eight-wheel independent drive vehicle, which demonstrated the obstacle-avoidance ability in a task-based offroad autonomous driving test.