Distributed Control And Optimization In Complex Environment

With the development of cloud computing and big data,distributed control and optimization has become one of the most important development directions of the modern system and control science with its unique advantages in multi-agent systems.The basic problem is how to design distributed controls or optimization algorithms based on the information interaction between neighborhoods so that each agent in the multi-agent system can achieve the goal of consensus or optimal consensus.The problem of distributed control is the basic control problem in multi-agent systems,which means that all agents achieve the same goal,such as state,position,etc.The distributed optimization problem is that the multi-agent system needs to optimize a global or local cost function while achieving the goal of consensus.It is worth noting that multi-agent systems will be disturbed by various factors in complex environment,such as noise with uncertain variance and singular perturbation caused by large-time scales.In this case,the distributed control and optimization problems still need further theoretical analysis and exploration.The distributed control and optimization of multi-agent systems in complex environment is studied in this paper.The distributed control problem of continuous stochastic multi-agent systems under G-expectation frame,the distributed control problem of stochastic multi-agent systems under singular perturbation,and the distributed control problem of Markovian switching with two time scales are studied respectively.Furthermore,the distributed convex optimization problem with constraints in continuous stochastic multi-agent systems and the nonconvex and nonsmooth distributed optimization problem for discrete multi-agent systems are explored.Major academic contributions are included,for the stochastic multi-agent and singularly perturbed multi-agent systems,distributed control and distributed optimization are studied.The problem of distributed control under G-expectation framework is explored for the first time.Lp weak average consensus,Lp strong consensus and in the senses of Choquet-capacity and quasi surely are analyzed based on the distributed control protocol.For the singularly perturbed continuous stochastic systems with time scale differences,the distributed control algorithm is proposed.With the help of graph theory,singular perturbation method and Kalman filter,the consensus is analyzed and the consensus of convergence rate is characterized.Finally,the distributed optimization problem of multi-agent system is studied based on the subgradient projection algorithm and the distributed proximal bundle method for solving the continuous-time stochastic distributed optimization with constrained sets and the nonconvex nonsmooth distributed optimization.Furthermore,the performance of the algorithm is analyzed.Meanwhile,the effectiveness of the designed distributed control protocols and the feasibility of the algorithms are verified by simulation.