This website requires JavaScript.

Toward multi-target self-organizing pursuit in a partially observable Markov game

Lijun SunYu-Cheng ChangChao LyuYe ShiYuhui ShiChin-Teng Lin
Jun 2022
摘要
The multiple-target self-organizing pursuit (SOP) problem has wideapplications and has been considered a challenging self-organization game fordistributed systems, in which intelligent agents cooperatively pursue multipledynamic targets with partial observations. This work proposes a framework fordecentralized multi-agent systems to improve intelligent agents' search andpursuit capabilities. We model a self-organizing system as a partiallyobservable Markov game (POMG) with the features of decentralization, partialobservation, and noncommunication. The proposed distributed algorithm: fuzzyself-organizing cooperative coevolution (FSC2) is then leveraged to resolve thethree challenges in multi-target SOP: distributed self-organizing search (SOS),distributed task allocation, and distributed single-target pursuit. FSC2includes a coordinated multi-agent deep reinforcement learning method thatenables homogeneous agents to learn natural SOS patterns. Additionally, wepropose a fuzzy-based distributed task allocation method, which locallydecomposes multi-target SOP into several single-target pursuit problems. Thecooperative coevolution principle is employed to coordinate distributedpursuers for each single-target pursuit problem. Therefore, the uncertaintiesof inherent partial observation and distributed decision-making in the POMG canbe alleviated. The experimental results demonstrate that distributednoncommunicating multi-agent coordination with partial observations in allthree subtasks are effective, and 2048 FSC2 agents can perform efficientmulti-target SOP with an almost 100% capture rate.
展开全部
图表提取

暂无人提供速读十问回答

论文十问由沈向洋博士提出,鼓励大家带着这十个问题去阅读论文,用有用的信息构建认知模型。写出自己的十问回答,还有机会在当前页面展示哦。

Q1论文试图解决什么问题?
Q2这是否是一个新的问题?
Q3这篇文章要验证一个什么科学假设?
0
被引用
笔记
问答