An interaction-aware predictive motion planner for unmanned ground vehicles in dynamic street scenarios

Abstract

An interaction-aware predictive motion planning method for unmanned ground vehicles is presented in dynamic street scenarios. Although trajectory prediction in motion planners is widely covered in the past few years, most of them only consider the physical model of the vehicles and ignore the interaction among vehicles. Our motion planner predicts the future trajectories of surrounding participant vehicles taking the traffic interaction and manoeuvres into consideration. Furthermore, the motion planner exploits an improved trajectory generation method. The kinematically feasible trajectories are generated, which prevents a long-term collision using the predicted results in a probabilistic manner. The results show that our motion planner improves the safety and smoothness of driving trajectories in interactive scenarios.