Distributed Model Predictive Control Based on Dynamic Games

GUIDO SANCHEZ; LEONARDO GIOVANINI; MARINA MURILLO; ALEJANDRO LIMACHE

Advanced Model Predictive Control

Intech

Lugar: Croacia; Año: 2011; p. 65 - 90

In this paper, a framework for distributed model predictive control based on dynamic games is presented. The original optimisation is decomposed into several small coupled problems in a distributed structure, which is solved iteratively. The conditions to guarantee stability and persistent feasibility are derived, and a procedure for constructing an initial feasible solution is provided. Moreover, we show that the solution of the iterative process converges to the centralised MPC solution when the information of system?s cost function and constraints are available to all agents. The ef- fect of communication failures on the system performance and behaviour are also analysed. The gain in the overall computational load compared to the centralised problem is balanced by the increased communication requirements. This makes this approach more applicable to situations where the number of subsystems is large and local commu- nication is relatively fast and cheap. The control problem of a consensesus of autonomous robots in an environment with arbitrary obstructions is used to illustrate the effectiveness, practicality and limitations of the proposed control framework.