In this paper, we consider the consensus problem of dynamical multiple agents that communicate via a directed moving neighborhood random network. Each agent performs random walk on a weighted directed network. Agents interact with each other through random unidirectional information flow when they coincide in the underlying network at a given instant. For such a framework, we present sufficient conditions for almost sure asymptotic consensus. Some existed consensus schemes are shown to be reduced versions of the current model.
GCS: Graph-based Coordination Strategy for Multi-Agent Reinforcement Learning
许多真实世界的场景涉及一组代理，他们必须协调他们的策略以实现共同的目标。以往的研究主要集中在分散控制上，以最大化共同的报酬，而很少考虑控制策略之间的协调，这在动态复杂环境中是至关重要的。在这项工作中，我们建议将联合团队策略分解为图形生成器和基于图形的协调策略，以实现代理之间的协调行为。图形生成器采用编码器-解码器框架，输出有向无环图（DAG）来捕获底层动态决策结构。我们还在图生成器中应用了DAGness-constrained和DAG depth-constrained优化来平衡效率和性能。基于图的协调策略利用生成的决策结构。同时训练图生成器和协调策略以最大化贴现收益。对协同高斯挤压、协同导航和Google Research Football的实验结果表明了该方法的优越性。
Communicating Plans, Not Percepts: Scalable Multi-Agent Coordination with Embodied World Models
Department of Computer Science University of Wisconsin-Madison（计算机科学系 明尼苏达大学）；Department of Computer Science National University of Singapore（计算机科学系 新加坡国立大学）
Emergent Coordination in Multi-Agent Language Models
D’Amore-McKim School of Business（达莫-麦克金商学院）；Khoury College of Computer Sciences（科里学院计算机科学学院）；Network Science Institute（网络科学研究所）；Northeastern University（东北大学）
