Distributed Model Predictive Control for MLD systems: Application to freeway ramp metering

Abstract

This paper deals with mixed logical dynamical systems controlled via model predictive schemes. For this class of systems a centralized MPC approach is firstly introduced, in which the finite horizon optimal control problem is a mixed-integer quadratic programming problem aiming at minimizing the deviations of the system variables from their equilibrium points. Since the application in real time of these MPC schemes is sometimes limited because of the high computational load necessary to solve the finite horizon optimal control problem, a distributed MPC scheme is proposed, characterized by two different and alternative algorithms. An important application is then introduced to assess the proposed approaches, i.e. ramp metering freeway traffic control. Referring to this applicative case, the two distributed control algorithms are compared, via simulation, with the centralized MPC scheme and with a completely decentralized one.