MULTI-AGENT MODEL PREDICTIVE CONTROL FOR TRANSPORTATION NETWORKS: SERIAL VERSUS PARALLEL SCHEMES

摘要

We consider the control of large-scale transportation networks, like road traffic networks, power distribution networks, water distribution networks, etc. For control of these networks, we propose a multi-agent control scheme in which each agent employs Model Predictive Control. In order to obtain coordination and to improve decision making agents communicate with each other. We compare two Lagrangian-based communication and decision making schemes. One scheme is based on serial iterations between agents, while the other is based on parallel iterations. The schemes are explained theoretically and assessed experimentally by means of simulations on a particular type of transportation network, viz., a power distribution network. The serial scheme shows to have preferable properties compared to the parallel scheme in terms of solution speed and quality.