Scholarlink

<p>This paper proposes a novel distributed learning control framework for multi-agent systems characterized by uncertain coupled nonlinearities in communication-constrained environments. The framework aims to address the core challenges of achieving robust performance and state synchronization amidst complex state-dependent couplings and external disturbances, without relying on continuous communication. The core of the proposed framework lies in constructing a unified interval modeling and control optimization architecture. Specifically, Radial Basis Function (RBF) neural networks are utilized to real-time identify unknown boundaries of nonlinear coupling effects, which are then characterized as time-varying intervals. Based on this interval characterization, an interval-aware Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm is designed. This algorithm drives agents to learn robust control policies capable of actively suppressing the worst-case impacts of the uncertain intervals. Furthermore, the framework incorporates a hybrid event-triggered mechanism. This mechanism optimizes the interplay between a state-adaptive threshold and a consensus-based triggering condition, significantly reducing communication requirements while strictly guaranteeing system stability and synchronization accuracy. Theoretical analysis rigorously proves the Input-to-State Stability (ISS) of the closed-loop system and the absence of Zeno phenomenon. Simulation results on a position-pressure-temperature coupled nonlinear multi-agent system demonstrate that the proposed framework maintains precise state synchronization while significantly reducing communication overhead.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Reinforcement learning-based interval event-triggered distributed synergetic control of uncertain coupled nonlinear system

  • 期刊论文

  • 发表日期：

    2026年2月11日

查看全文
  • Xiang Li,
  • Yaqiu Liu,
  • Yunlei Lv,
  • Lina Liu

摘要

This paper proposes a novel distributed learning control framework for multi-agent systems characterized by uncertain coupled nonlinearities in communication-constrained environments. The framework aims to address the core challenges of achieving robust performance and state synchronization amidst complex state-dependent couplings and external disturbances, without relying on continuous communication. The core of the proposed framework lies in constructing a unified interval modeling and control optimization architecture. Specifically, Radial Basis Function (RBF) neural networks are utilized to real-time identify unknown boundaries of nonlinear coupling effects, which are then characterized as time-varying intervals. Based on this interval characterization, an interval-aware Multi-Agent Deep Deterministic Policy Gradient (MADDPG) algorithm is designed. This algorithm drives agents to learn robust control policies capable of actively suppressing the worst-case impacts of the uncertain intervals. Furthermore, the framework incorporates a hybrid event-triggered mechanism. This mechanism optimizes the interplay between a state-adaptive threshold and a consensus-based triggering condition, significantly reducing communication requirements while strictly guaranteeing system stability and synchronization accuracy. Theoretical analysis rigorously proves the Input-to-State Stability (ISS) of the closed-loop system and the absence of Zeno phenomenon. Simulation results on a position-pressure-temperature coupled nonlinear multi-agent system demonstrate that the proposed framework maintains precise state synchronization while significantly reducing communication overhead.