Real-time Obstacles Avoidance for Vehicles in the Urban Grand Challenge

Abstract

In this paper, the real-time trajectory planning problem is considered for vehicles in the urban grand challenge. Typically, reference paths generated by a high level path planner are not feasible for vehicles and need improvement. In addition, moving obstacles in the environment are generally not known apriori, which also requires the paths to be able to be replanned in real-time. The proposed method presents three novel features to satisfy those requirements. First, all the paths satisfying boundary conditions and the vehicle’s kinematic constraints are parameterized in terms of polynomials of sufficient order. Then, a collisionfree criterion is developed for avoiding “hard” and “soft” obstacles detected. In the third step, a L2-norm performance index is introduced to find the best path among the class of collisionfree paths. The performance index is chosen such that a path analogous to the shortest path can be analytically solved. The proposed method provides a systematic and analytical solution to find the feasible path while addressing obstacle avoidance and guaranteeing performance. Simulation results illustrate the proposed algorithm.