Collision risk assessment of occluded vehicle based on the motion predictions using the precise road map

A perception system based on vehicle detection sensors, which are mounted on an ego vehicle, has restricted visibility because of blockage by obstacles. Estimating the risk of collision with moving vehicles in an occluded area is difficult because their locations and speeds cannot be detected. To address the occlusion problem, this paper proposes a probabilistic collision risk assessment method for a potential collision vehicle in an occluded area. The proposed method estimates the collision risk in three steps: occlusion boundary modeling of perception, motion prediction of the potential collision vehicles, and probabilistic collision risk assessment. The first step models the occlusion boundary to classify the free space and the unknown region. In the second step, the moving path of each potential collision vehicle is predicted considering its future behavior. The final step estimates the collision probability with a potential collision vehicle based on the speed distribution of the vehicles on the road. We evaluate the proposed probabilistic collision risk assessment method in several occlusion scenarios with real traffic, including an alleyway, a merging lane, and blockage by a bulky vehicle.