Energy-Optimal Coordination of Autonomous Vehicles at Intersections

The problem of coordinating automated vehicles at intersections is naturally posed within the optimal control framework, using objectives such as minimization of energy consumption. In this paper we extend previous work to include relevant nonlinearities in the vehicle models and propose a cost function that directly captures both energy consumption and travel time. The problem is a so-called Economic MPC (EMPC) problem, which entails both numerical and theoretical challenges. To address these issues, we propose to use a previously presented procedure to tune a MPC with a quadratic objective to approximate the EMPC. We evaluate the performance of both linear and nonlinear approximating MPC controllers in simulation. In particular, we demonstrate that a standard linear MPC can be tuned to so that the losses with respect to the EMPC is below 1%.