基于时序逻辑的多机器人协作共融关键问题研究

With the contemporary advances in computer technology and artificial intelligence, robot with co-existing capability becomes an imminent popular topic in the community. This proposal attempts to tackle temporal logic based multi-robot cooperative planning and control problem from the perspective of multi－agent systems, special attention is paid on how to systematically establish a temporal logic based multi-robot cooperative planning and control methodology. The main topics of the proposed research include: (1) how to design and execute temporal logic to accommodate topological properties. (2) how to systematically perform logic modification when conflict of interest is detected in multi-robot scenarios, and how to find feasible solution with relaxed abstraction. (3) how to integrate extremum seeking based distributed optimization technique into temporal logic framework, to realize cooperation and co-existing between robot and environment and among robots as well. (4) how to carry out experiments using the proposed conjecture to fully investigate the strength and limit of the proposed method. The proposed effort is dedicated to solving some key problems in Multi-Robot cooperation, and once successful, it will help to sustain and subsequently improve the technical advantage of China's information technology, control industries as well as artificial intelligence communities.

随着计算机与人工智能的发展，共融机器人成了新的研究热点。本项目拟从多智能体网络控制角度研究基于时序逻辑的多机器人协作共融问题，力求系统地建立一套基于时序逻辑的多机器人协作体系。主要研究内容包括：（1）如何设计包含网络拓扑联通性信息的时序逻辑及其实现问题；（2）任务冲突时候逻辑的实时修正以及松弛可行解的求取问题；（3）如何将基于极值搜索的分布式优化算法与时序逻辑结合用于多机器人共融场景，实现机器人之间以及机器人与环境之间的交互共融；（4）如何将所设计的算法应用于解决多机器人协作共融场景，来检验算法的有效性与局限性。本项目的研究致力于解决当前机器人协作共融的几个关键问题，填补相关领域研究的空白，为提高我国信息现代化和人工智能领域的竞争实力做出贡献。

多机器人协作共融是当前我国智能制造和信息现代化领域发展的关键问题，如何在复杂环境下实现智能机器人和智能系统的集群协作也是自动化领域的研究热点，其在国防军工领域同样具有重要的应用价值和现实意义。本项目以智能无人系统集群共融为背景，研究了智能机器人集群感知、控制与优化的关键理论问题。特别值得一提的是，项目解决了基于时序逻辑的机器人集群路径规划、冲突解决等问题，并创新性地将其应用于解决航天器集群在轨任务规划问题；针对集群态势感知问题，项目提出了一种实时的连通性分布式快速感知算法，通过估计拉普拉斯矩阵特征值的上下限来实现对连通性的逼近，从根本上解决了集群态势感知的实时性和快速性问题；针对集群博弈问题，项目从理论上解决了线性系统分布式纳什博弈点的获取问题，通过对开环博弈终端条件的估计解决了传统博弈算法在集群共融场景中的应用；针对集群信息融合问题，项目通过引入分布均值跟踪算法设计了一种新的分布式卡尔曼滤波，从理论上克服了观测矩阵本地化的约束。另外，项目还研究了基于次梯度的分布式集群优化问题，通过改进一致性环节实现了分布式优化的有限时间收敛。通过这些关键问题的解决，为多机器人集群共融与协作奠定了坚实的理论基础，为我国相关领域的研究提供了有力的技术支撑。