噪声环境下异构多智能体系统的一致性研究

Agents in the complex environments may have different dynamic characteristics and are inevitably subjected to noises. The heterogeneity and noise significantly affect the convergence performance and bring substantial difficulties to the design of the corresponding control law. This project will investigate the consensus control of the heterogeneous multi-agent systems in noisy environments, including the cases with delays and switching topologies. Firstly, the error compensations are introduced to transform the consensus problems of the heterogeneous multi-agent systems into the stability problems of stochastic systems. Then, stochastic Lyapunov function/functional, martingale theory and large deviation technology are applied to establish the stochastic stability criteria of such stochastic systems. Finally, the stochastic stability criteria will be applied to design the consensus protocol and find the corresponding consensus conditions. The explicit dependencies between heterogeneous multi-agent consentability and the system parameters are revealed, where the system parameters include the noises, delays, communication graph, and the original heterogeneous multi-agent system dynamics. The results of this project will enrich the consensus control theory of heterogeneous multi-agent systems, and provide the theoretical foundation for the study of practical distributed control systems.

复杂环境中的智能体往往具有不同的动态特性和结构，也不可避免地受到噪声的影响。多智能体系统的异构特性和噪声显著影响了系统的性能，也给一致性控制协议设计和性能分析带来了困难。本项目将在噪声环境下，考虑时滞、切换拓扑等实际情况，研究异构多智能体系统的一致性问题。首先引入合适的误差补偿，建立反映异构多智能体系统一致性控制的随机系统模型，然后利用随机 Lyapunov 函数／泛函、鞅论、大偏差等方法构造随机系统稳定性的条件，并由此设计噪声环境下异构多智能体系统的一致性控制协议，揭示其一致性与噪声强度、时滞、拓扑结构及系统动态特性的显式依赖关系。本项目的研究成果将丰富异构多智能体系统一致性控制理论，为涉及分布式控制领域的研究提供理论依据。

人工智能和分布式计算等领域的交叉融合形成了新的研究领域—多智能体协同控制，其中异构多智能体系统的应用更为广泛。而且在复杂的环境中，系统的动态变化不可避免地受到各种不确定性因素(噪声、时滞、拓扑切换等)的影响。因此，研究不确定性环境下异构多智能体系统的控制及应用尤为重要。本项目主要研究了噪声环境下异构多智能体系统一致性控制问题。首先提出噪声环境下一致性控制协议的设计方法，然后分别分析了噪声环境下，时滞和拓扑结构切换对系统一致性的影响，得到了噪声环境下异构多智能体系统均方一致和几乎必然一致的充分条件，这些条件揭示了拓扑结构、控制增益、时延参数和噪声强度之间的关系，为异构多智能体在复杂环境下的一致性研究提供有效的理论方法。本项目的研究成果丰富了随机时滞系统稳定性理论和噪声环境下多智能体系统的一致性控制理论，同时也将研究成果应用于无人机编队控制和航天器编队控制。