Centralized MPC for autonomous intersection crossing

This paper develops a method for a safe and autonomous intersection crossing. A centralized system controls autonomous vehicles within a certain surrounding of the intersection and generates optimized trajectories for all vehicles in the area. A recently proposed design approach, 10, where this problem is expressed as a convex optimization problem using space sampling instead of time sampling, is formulated as a MPC problem solved by a QP algorithms so that it can be executed in real time. The MPC controller is then integrated in CarMaker using Matlab/Simulink so that the controller can be validated against the advanced vehicle models and sensor models available in CarMaker. Preliminary results of this validation are presented. Also, a method is designed to obtain time gaps between the vehicles to prevent the optimization problem to become infeasible when sensors give noisy measurements.