网络化多智能体系统预测控制设计与分析

This proposed project establishes a dynamical model for networked multi-agent systems with internal uncertainties, external disturbances and communication constraints (for example, network delay, data dropouts, bandwidth limitation, varying network structure etc). The focus of the project is on predictive control and consensus analysis of networked multi-agent systems with cost function coupling, coupling constraints, and communication constraints. It proposes a networked distributed predictive control method based on compatible constraints and error penalty, a networked distributed predictive control method with coupling constraints, a networked distributed predictive control method based on dynamical coordination rules and a networked double module distributed predictive control method without communication. Combing the networked predictive control method and robust control method, the effects of communication constraints, model uncertainties, and external disturbances to the control performance are overcome. The stability and consensus of networked multi-agent predictive control are analyzed using the ultimate consistent bounded method and convergence estimation method. This project will establish a foundation of theory, experiments and applications for the research of predictive control of networked multi-agent systems.

针对网络化多智能体系统所处的外部环境和具有的内部特性，建立包含通信受限(如：网络延时、数据丢失、带宽限制，网络结构可变等)、内部不确定性和外部干扰的网络化多智能体系统动态模型。将重点讨论代价函数耦合、耦合约束条件和通信失效下的网络化多智能体系统的预测控制及其一致性问题。提出基于相容约束和偏差惩罚的网络化分布式预测控制方法，带耦合约束条件的网络化分布式预测控制方法，基于动态协调规则的网络化分布式预测控制方法，带通信失效的网络化双模分布式预测控制方法。结合网络化预测控制方法和鲁棒控制方法，克服系统中的通信受限、模型不确定性和外部干扰对控制性能的影响。采用终极一致有界、收敛界估计等方法进行网络化多智能体系统预测控制的稳定性和一致性分析。最终，为实现网络化多智能体系统的预测控制研究奠定一个理论、实验和应用的基础。

随着网络化控制技术的迅速发展，网络化多智能体控制在人类日常生活和科学工程领域中有着广泛的应用潜力，如：物联网，智能交通、智能电网，无人机编队，卫星组网等。由于存在通信受限(如：网络延时、数据丢失、带宽限制，网络结构可变等)、内部不确定性和外部干扰，这给网络化多智能体控制系统的设计、分析与实现带来了巨大的挑战。该项目考虑了网络化多智能体系统的综合建模；提出了多种网络化多智能体预测控制方法，如：网络化同构多智能体预测控制方法，网络化异构多智能体预测控制方法，网络化广义多智能体预测控制方法，网络化多智能体预测跟踪方法，网络化分组多智能体控制方法等；分析了网络化多智能体预测控制系统的安全性、鲁棒性和抗干扰性；获得了闭环网络化多智能体预测控制系统的稳定性和一致性判据；研发了网络化多智能体控制系统实时仿真实验平台；进行了大量的网络化多智能体预测控制系统的实时仿真和实验。为实现网络化多智能体系统的预测控制研究奠定了一个理论、实验和应用的基础。