Win-fit: Efficient intersection management via dynamic vehicle batching and scheduling

Intersection management is one of the most challenging tasks in traffic control. In this paper, to significantly decrease the average delay for autonomous vehicles to cross an intersection without any collision, vehicles are dynamically grouped and a selected winner group is allowed to drive through the intersection as one batch unlike most existing approaches, which only schedule individual autonomous vehicles one by one. Further, certain vehicles in the non-winner groups are allowed to safely cross the intersection during idle time slots (if any) unused by the winner group to further decrease the delay. Our approach has the lower-bound time complexity, and decreases the average trip delay to cross a simulated intersection by 31%-95% compared to several state-of-the-art baselines.