Optimal Conflict Resolution for Vehicles With Intersecting and Overlapping Paths

A collaborative centralized model predictive controller solving the problem of autonomous vehicles safely crossing an intersection is presented. The solution gives optimal speed trajectories for each vehicle while considering collision avoidance constraints between vehicles traveling on the same path before, after and/or within the intersection. This extends earlier results, where collision avoidance was only considered for vehicles with intersecting paths, with the possibility of vehicles on the same path and by this, the controller is not only one step closer to handling complex traffic intersections but can now be used for merging and splitting of roads, roundabouts and intersection networks. The proficiency of the extended controller is demonstrated by applying it to a four-way intersection. It is shown that the controller provides smooth, collision free trajectories in scenarios with and without vehicles traveling in the same lane. Further, it is evaluated how the solutions differ when using various cost functions and how the controller handles disturbances in the form of a sudden lane blockage. Lastly, it is discussed how the presented controller could also be extended to handle mixed-traffic scenarios and how soft constraints can be used to avoid infeasibility in the case of missing or noisy traffic data.