Abstract

Road traffic networks are large-scale systems that demand distributed control strategies. Distributed model predictive control (DMPC) arises as a feasible alternative for traffic control. Distributed strategies decompose the whole traffic network into different subnetworks with local optimal controllers that make decisions on actions to be taken by the actuators responsible for traffic control (traffic lights, routing signals, variable speed limits, among others). However, subnetworks are interacting elements of the whole traffic network. Hence, local control decisions made for one sub-network affect and are influenced by the decisions taken for the other subnetworks. Under these circumstances, the DMPC traffic problem can be treated as a game where the rules are provided by the physical system, the players are the local optimal controllers, their strategies are the control sequences, and the payoffs are the local performance indices (such as the total time spent by the users in the network). This configuration allows the achievement of a computational burden reduction, with a compromise between local and global performance. Since DMPC local controllers are able to communicate with each other, the control of the traffic network corresponds to a cooperative game. In this chapter, game-theory-based DMPC is developed and tested for control of urban and motorway networks.