A Collision-free Coordination Framework for Mixed-Vehicle Intersections

Proper coordination systems for autonomous vehicles (AVs) in complex road structures such as road intersections are of great importance. However, considering the fact that not only AVs but also human-driven vehicles (HVs) are involved at intersections, existing coordination strategies are limited in reliability and flexibility. In order to guarantee the vehicle safety in the mixed vehicle intersections, we propose a double-level coordination framework. In the centralized node, a high-level planner based on appropriate space-time resource allocation is used to provide feasible tunnels to AVs, to achieve high traffic throughput. Furthermore, on each AV, a low-level planner replans a collision-free practical trajectory to avoid HVs. Sensing, localization and control uncertainties are all considered in the framework to give a proper replanning interval in low-level planner, for improving the robustness of the system. Numerical results are provided and validate our analysis.