大规模网络化控制系统分布式随机模型预测控制方法研究

The large-scale networked control systems widely existing in practical applications are a class of uncertain systems and have the features of large scale, a great amount of constraints and objectives. The traditional control methods cannot satisfy the requirements of this kind of systems. Over the past years, distributed model predictive control (DMPC) has attracted great attention to industrial applications due to its good control performance, capability of handling constraints, and structural flexibility. Therefore, it is becoming an important tool to handle the large-scale networked control systems. However, the research of DMPC is still in an embryonic stage and lots of difficult but rather important problems still remain to be solved. This project aims to develop efficient distributed stochastic model predictive control (DSMPC) methods for large-scale networked stochastic systems with coupled probabilistic constraints. Firstly, the characteristics of both stochastic disturbances and model uncertainties will be analyzed, and the interactions among subsystems caused by coupling in the dynamics and constraints will be depicted. Based on this, the whole system model will be built. Secondly, the coupled constraints handling strategy will be explored in a distributed manner. Furthermore, some DSMPC schemes will be proposed and the performance of the controllers will be evaluated. Next, when communication is limited, a DSMPC will be designed to achieve the cooperation between subsystems. Finally, the theoretical results will be validated by multi-unmanned vehicle systems test platform. The implementation of this project will enrich and develop theoretical results of DSMPC and further provide theoretical and technical supports for engineering practices of large-scale networked control systems.

大规模网络化控制系统是现实中广泛存在的一类不确定系统，具有规模大、约束多、目标多等特点，传统的方法无法满足这类系统的控制要求。近年来，分布式模型预测控制以其良好的控制性能，有效处理约束的能力和较高的灵活性等优点得到关注，成为处理大规模网络化控制系统的重要工具。目前，该方法尚处于研究初期，仍存在诸多难点问题亟待解决。本项目旨在提出有效的分布式随机模型预测控制方法，重点解决具有概率约束的大规模网络化随机系统控制问题。首先，分析系统内部和外部不确定性及子系统间耦合关联，建立系统模型；其次，研究耦合概率约束的分布式处理策略，给出分布式随机模型预测控制器设计方法及理论分析结果；进一步，在通信受限情况下，设计具有协调性能的分布式随机模型预测控制方法；最后，通过多无人车实验平台验证理论结果。项目研究将丰富和发展分布式随机模型预测控制理论，有望为大规模网络化控制系统的工程实践奠定坚实的理论和技术基础。

本项目针对具有概率约束的大规模网络化不确定系统的控制问题开展了系统深入的研究，提出了一系列实用有效的分布式随机/鲁棒模型预测控制设计方法。该课题为目前控制领域最前沿的研究方向之一，综合了控制论、最优化理论、概率论和不变集理论等，研究难度大，现存成果少。首先，分析了系统内部和外部不确定性及子系统间耦合关联，建立了系统模型；其次，研究了耦合约束的分布式处理策略，给出了分布式鲁棒/随机模型预测控制器设计方法及理论分析结果；进一步，在通信受限情况下，为了减小在线计算量，设计了基于事件触发的分布式鲁棒/随机模型预测控制方法；最后，通过多无人车实验平台验证理论结果。本项目的研究丰富并完善了分布式鲁棒/随机模型预测控制理论，为大规模网络化控制系统的工程实践提供了坚实的理论和技术基础。..依托该项目，共发表录用国际SCI期刊论文12篇和EI国际会议论文2篇。6篇发表于控制领域顶级SCI期刊，6篇发表于控制领域重要SCI期刊，包括IEEE Transactions on Automatic Control、Automatica、International Journal of Robust and Nonlinear Control、Journal of the Franklin Institute和IET Control Theory & Applications等。同时，搭建了一个基于无线通信的多无人车系统控制实验平台，并对所提理论方法的有效性进行了实验验证。此外，项目负责人作为第五获奖人获2018年吴文俊人工智能自然科学奖二等奖。