Distributed Model Predictive Control Using Cooperative Contract Options

To achieve a common goal in an optimal way systems are becoming increasingly interconnected. While operating as autonomous as possible, the subsystems should exploit the capabilities and desires of neighboring systems in a cooperative way to avoid conflicts and optimize overall performance. Increasing interconnections, together with autonomous behavior, however, challenge the system design and control. Distributed model predictive control strategies allow to break the complexity of interconnected systems, as each subsystem employs its own controller, taking information obtained or provided by the neighboring systems into account. We propose a distributed model predictive control strategy, in which the subsystems provide multiple options of possible future behaviors to their neighbors. The options are provided in form of contracts - outer bounds of the variables of the coupling variables. The neighboring systems take these options in their predictions into account and choose the most suitable one. By this they exploit the capabilities and influences of the neighbors in a cooperative way. We outline conditions which guarantee repeated feasibility and illustrate the approach considering autonomously driving vehicles entering a highway. (C) 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.