A Real-Time and Predictive Trajectory-Generation Motion Planner for Autonomous Ground Vehicles

Abstract

This paper presents a real-time and predictive motion planner for autonomous ground vehicles. The motion planner can generate kinematically feasible and human-driving like trajectories based on an improved state-space trajectory generation method. Meanwhile, the motion planner also considers the future behavior of other participant vehicles through a control-space based Kalman predictor. The experimental results demonstrate that proposed motion planner has an improvement in generating safer and smoother trajectories, compared with Integrated Local Trajectory Planning (ILTP). Our motion planner has the capability to deal with complex traffic environments, especially with interactions of other participant vehicles.