Distributed model predictive control applied to urban traffic networks: Implementation, experimentation, and analysis

Distributed model predictive control (DMPC) operates large, dynamic systems with a network of distributed agents. The advantages over standard model predictive control (MPC) are the distribution of decision-making, facilitated reconfigurability and expansion, and robustness to failure. A chief difficulty to its deployment is the lack of a computational platform for implementing DMPC. To this end, this paper presents a prototype of computational platform that automatically implements DMPC from the mathematical specification of the problem and runs a distributed algorithm. The platform prototype is tested in the control of an urban traffic network.