A simple, space-time-symmetric collision avoidance method for autonomous vehicles

Abstract

This paper summarizes a previously developed space-time avoidance method, which applies to autonomous vehicles with temporal uncertainty, and which eliminates a collision threat by shifting vehicle's trajectory laterally as well as temporally. The shift parameters are obtained from simple geometric operations in metric 3D space-time. Though safe and simple, this method lacks efficiency for systems with little temporal uncertainty: the achieved planar collision distance is sometimes much larger than required. Therefore, this paper further proposes a specific, efficient solution for the case of 2 vehicles meeting without temporal uncertainty. Experimental results are presented in which the vehicles collide, or avoid each other mainly laterally or temporally, and by a mixture of both, depending on the choice of the time scaling factor.