Multiagent Coordination Optimization Based Model Predictive Control Strategy With Application to Balanced Resource Allocation

Abstract

Model predictive control (MPC) is a heuristic control strategy to find a consequence of best controllers during each finite-horizon regarding to certain performance functions of a dynamic system. MPC involves two main operations: estimation and optimization. Due to high complexity of the performance functions, such as, nonlinear, non-convex, large-scale objective functions, the optimization algorithms for MPC must be capable of handling those problems with both computational efficiency and accuracy. Multiagent coordination optimization (MCO) is a recently developed heuristic algorithm by embedding multiagent coordination into swarm intelligence to accelerate the searching process for the optimal solution in the particle swarm optimization (PSO) algorithm. With only some elementary operations, the MCO algorithm can obtain the best solution extremely fast, which is especially necessary to solve the online optimization problems in MPC. Therefore, in this paper, we propose an MCO based MPC strategy to enhance the performance of the MPC controllers when addressing non-convex large-scale nonlinear problems. Moreover, as an application, the network resource balanced allocation problem is numerically illustrated by the MCO based MPC strategy.