A unified approach to local trajectory planning and control for autonomous driving along a reference path

This paper presents a unified approach to local trajectory planning and control for the autonomous ground vehicle driving along a rough predefined path. In order to cope with the unpredictably changing environment reactively and reason about the global guidance, we develop an efficient sampling-based model predictive local path generation approach to generate a set of kinematically-feasible trajectories aligning with the reference path. A discrete optimization scheme is developed to select the best path based on a specified objective function, then followed by the velocity profile generation. As for the low-level control, to achieve high performance of control, two degree of freedom control architecture is employed by combining the feedforward control with the feedback control. The simulation results demonstrate the capability of the proposed approach to track the curvature-discontinuous reference path robustly, while avoiding collisions with static obstacles.