Distributed Model Predictive Control for Intersection Automation Using a Parallelized Optimization Approach

Road intersections are usually a bottleneck in big cities and might lead to severe congestion during rush hour traffic. With highly automated vehicles leveraging vehicle-to-vehicle communication, traffic lights and signs might become dispensable one day. For this particular scenario, we propose a distributed model predictive control approach which enables multiple vehicles to pass the intersection simultaneously in a safe and efficient way. The resulting non-convex distributed control problem is solved in parallel without any nested iterations which is achieved by applying constraint prioritization. Constraint prioritization introduces priorities on vehicles such that a vehicle with lower priority has to give right of way in case of a potential conflict without prescribing an intersection passing order. To cope with non-convex safety constraints, we apply a semidefinite programming relaxation with randomization to determine appropriate feasible solutions. Simulation results finally prove the efficacy of our approach.