基于滚动时域估计与预测控制的多智能体分布式协调控制研究

Aiming for the distributed cooperative control of multi-agents, the design and combination of the algorithms of distributed moving horizon estimation fusion and distributed model predictive control are studied, to construct the holistic distributed cooperative control framework and to propose the specific distributed control strategy. The distributed moving horizon estimation fusion is studied aiming at the time-varying communication topology of multi-agents. And the algorithm is proposed by adopting the graph theory, matrix analysis, quadratic boundedness technology, hybrid multiple models method and consensus theory, which can guarantee the convergence, optimality and global consensus of all agents' estimations. With utilizing the distributed estimations and fusions, the individual control optimization problem with recursive feasibility gurantee is built for each agent, by designing the compatibility, collision avoidance and terminal constraints, and by tightening the designed constraints. Then the distributed model predictive control algorithm is proposed with stability guarantee. According to the double layered structure, the distributed moving horizon estimation fusion algorithm and the distributed model predictive control algorithm are combined to construct the holistic distributed cooperative control framework, based on which the specific distributed control strategy is presented. Due to providing the theoretical and technological improvements on both the distributed estimation fusion and distributed control, and providing the solusion for the practical application of multi-agents, the achievements of this project possess with the importances in both the theoretical and practical fields.

面向多智能体的分布式协调控制问题，基于滚动时域估计与预测控制方法，重点研究分布式估计融合与控制算法的设计与结合，构建综合估计融合与控制的分布式协调控制框架，并基于该框架给出具体控制方案。在分布式估计融合中，针对多智能体通讯拓扑时变的情况，采用图论、矩阵分析、二次有界技术、混合多模型方法与一致性理论等手段，提出有收敛性、最优性与全局一致性保证的分布式滚动时域估计融合算法。在分布式控制中，基于分布式估计融合信息，通过相容约束、避碰约束与终端约束的设计以及对约束收缩处理，构造有递归可行性保证的子优化问题，并提出有稳定性保证的分布式预测控制算法。基于“双层结构”的先进理念，将分布式滚动时域估计融合与分布式预测控制算法结合，构建测控一体的分布式协调控制框架并给出具体控制方案。研究成果可进一步完善分布式估计融合与分布式控制的理论与技术、解决多智能体系统的实际应用问题，具有重要的理论价值与实际意义。

本项目针对多智能体系统的协调控制问题，基于滚动时域估计与预测控制方法，研究以下四个方面内容：1. 考虑多智能体控制中需要基于多传感器网络进行协同定位的情况，基于多维标度与多种滤波估计方法，设计了具有鲁棒性与更高精度的分布式协同定位算法；2. 考虑被跟踪目标在运动模式与目标类型方面的不同、传感器量测数据不完全与信息交互网络时变等情况，基于滚动时域估计方法，设计了具有一致性与收敛性的分布式估计融合算法；3. 考虑时变通讯拓扑下异构多智能体系统的跟踪与编队、带有有界扰动的非完整动力学智能体系统的跟踪等问题，基于以预测控制为主的多种方法，设计了具有递归可行性与稳定性的分布式控制算法；4. 考虑模型参数时变、带有有界扰动与饱和控制输入的控制对象，基于滚动优化的方式将估计与控制问题综合在优化问题中进行研究，设计了具有估计收敛性与控制稳定性的鲁棒输出反馈控制方案并给出了具体的控制算法。以上研究成果解决了多智能体协调控制中的一些关键理论与技术问题，进一步推进了多智能体系统的实际应用，具有一定的理论与实用价值。.本项目共计发表学术论文18篇，其中SCI检索4篇、EI检索10篇、核心期刊论文4篇，申请国防专利2项，培养博士生1名、硕士生2名。