Occlusion-aware on-road autonomous driving: A trajectory planning method considering occlusions of Lidars

Abstract

On-road motion planning concerns about how to find a local trajectory that is free from collisions or violations of traffic laws, easy to be tracked by the low-level controllers, comfortable for the passengers, and consistent with the common practice of human drivers. Safety is an essential issue when a vehicle travels on an urban road, especially when blind regions appear in the way ahead. Absent-minded pedestrians may appear suddenly from the blind region at the roadside. Following the predominant first-sampling-then-optimization hierarchical planning framework, this work is focused on how to make the derived trajectory safer w.r.t. the underlying risks in the occlusion areas. In searching for a coarse trajectory via sampling in the time-s-l state space and searching via dynamic programming (DP), we formulate a cost function that can handle occlusion-related risks quantitatively. Through this way, the occluded regions would be treated seriously but not overcautiously. According to the derived simulation results, the ego vehicle tends to go far away from the occluded regions or to slow down, which is intuitively following human drivers’ common practice.