Decoupled cooperative trajectory optimization for connected highly automated vehicles at urban intersections

The increasing market penetration of connected vehicles supports the development of highly automated vehicles for various traffic situations. Especially intersections form a bottleneck for the traffic flow and thus offer a high potential not only to increase the efficiency, but also to ensure safety. This paper presents a decoupled and decentralized approach using graph-based methods to optimize longitudinal trajectories for multiple vehicles at urban intersections. The approach enables the vehicles to cooperate, while avoiding collisions, considering dynamic influences like traffic lights, and minimizing a cost function. Furthermore, several heuristics are introduced, reducing the computational effort to solve these complex tasks. Simulations of an intersection scenario using the Monte Carlo method show a reduction of summarized costs, which represent travel time, efficiency and driving comfort, by ~28% compared to a driver model and by ~2.6% compared to a non-cooperative system.