A Bi-level Local and Global Model Predictive Control Architecture for Air Traffic Management

Abstract

A bi-Ievel architecture based on model predictive control is proposed for air traffic management. The aim is to reduce the computational complexity of online optimization in air traffic control, while producing accurate control decisions. We integrate a slow-rate centralized MPC controller including a global, long-term prediction window (top level of control), with several fast-rate decentralized MPC controllers including local, short-term prediction windows (bottom level of control). All aircraft are steered towards their destination in the shortest possible time, while forbidden areas are avoided. The performance of the proposed bi-Ievel control architecture is compared via a case study with (1) a slow-rate centralized MPC controller similar to the top control level, (2) a fast-rate decentralized MPC controller similar to the bottom control level, and (3) a fast- rate centralized MPC controller with a prediction horizon the same as the top control level and decision sampling times the same as the bottom control level: The results show that while the bi-Ievel MPC controller is significantly more efficient - w.r.t. the computation time - than the fast-rate centralized MPC controller, these two controllers are comparable w.r.t. accuracy of the decisions. Moreover, with a comparable computation time, the bi-Ievel MPC controller attains a superior control performance - considering the accuracy of decisions - compared to the slow-rate centralized and fast-rate decentralized MPC controllers individually. This shows that the integration of these two control approaches in the proposed bi-Ievel architecture can significantly boost their performance.