Comparison between mixed-integer and second order cone programming for autonomous overtaking

Paper in proceeding, 2018

This paper concerns optimally controlling an autonomous vehicle to perform safe and comfortable overtaking of a slower moving leading vehicle. The contribution is an analysis of the comparisons between a convex relaxation with the standard mixed integer quadratic program. The main difference between the formulations is that the sampling is performed in the temporal domain in the standard formulation, but in the spatial domain for the convex relaxation. The case of varying lateral position and longitudinal velocity is studied for both algorithms and the solution quality and computational effort are discussed. The results are given in a case study where an ego vehicle is forced to accelerate when overtaking a leading vehicle, due to the presence of an oncoming vehicle. The results illustrate that the temporal and spatial formulation’s yield similar solutions but obtaining