A Bilevel Virtual Platoon Based Coordination Framework for CAVs at Unsignalized Intersection

Emerging connected and automated vehicles (CAVs) are believed to play a critical role in the next-generation transportation systems. However, complicated intersection conditions and several safety problems challenge the optimal cooperative driving of CAVs. This paper proposes a bilevel coordination framework for CAVs at unsignalized intersections to improve efficiency and robustness while guaranteeing safety. The upper level focuses on a conflict-free and real-time scheduling problem for optimal passing order, and the results are transmitted to each CAV in the lower level for distributed control. To guarantee safety, we first project the conflicting vehicles on virtual lanes based on the conflict zones inside the intersection. Furthermore, the scheduling problem is transformed into a packing problem and formulated as a mixed-integer linear programming (MILP) for higher traffic efficiency. We propose three practical principles to obtain the real-time solutions by removing several redundant constraints in the optimization problem. Eventually, considering the inevitable bounded disturbances (e.g., unmodeled dynamics and equipment measurement errors), a distributed robust model predictive controller is designed to form and keep the desired virtual platoons under uncertainty. Numerical simulations validate that the proposed method saves 4.2%-6.6% total evacuation time under heavy traffic volume.