A Real-time Motion Planner for Intelligent Vehicles Based on a Fast SSTG Method

Abstract

This paper proposes a motion planner for intelligent vehicles in real time. The motion planner can adaptively sample a number of states according to the initial state of the intelligent vehicle and road geometry. Then a novel method based on the state-space trajectory-generation method generates kinematically feasible trajectories which connected the initial state and the sampled states in a fast manner. At last, the final trajectories will be selected by a defined performance function. The experimental results demonstrate that proposed motion planner has an improvement in generating feasible trajectories. The average runtime of our method is approximately 5% of that of model-based predictive state-space trajectory-generation method. Moreover, our motion planner has the capability to deal with complex traffic environments in real time.