复杂网络图模型的信息评估：协同信息共享的新方法

7 浏览量更新于2024-08-27 收藏 758KB PDF 举报

"AFastGraphic-basedInformationValuationAlgorithmforCooperativeInformationSharing HaixiaoHu,YangXu*,YulinZhang,MingLiu SchoolofComputerScienceandEngineering,UniversityofElectronicScienceandTechnologyofChina Chengdu,Sichuan,611731,P.R.China Email:xuyang@uestc.edu.cn 摘要—在动态且部分可观察的环境中的多代理团队合作决策中，信息共享至关重要。除了建立全面的信息覆盖外，如果团队成员能够自主评估潜在接收者并确定通信位置，协调效率将显著提升。尽管在本体图匹配和自然语言处理领域已经对信息评估方法进行了深入研究，但这些模型在处理大规模分布式代理的动态协调时无法实现快速推理。本文提出了一种基于复杂网络图模型的快速信息评估方法，该方法有助于识别信息的重要性。如同人类模糊信息评估一样，关键在于重要信息在融合到复杂信息图中时，其价值会显著变化。因此，我们在图模型中计算新信息的语义值，并构建局部图评估算法来估算信息图的演变，而非执行耗时的全局图搜索。这使得我们可以将本地评估算法转化为座席信息库的有效查询，从而让他们能迅速作出决策。虽然这种决策可能不完全精确，但正如人际交流中那样，对于多主体团队来说，信息共享效能足以传播有价值的信息。我们通过多主体合作的案例研究验证了所提信息评估方法的可行性。关键词：信息共享，协同决策，复杂网络图，信息评估，局部图算法在本文中，作者首先阐述了当前信息评估方法的局限性，即在处理大量分布式代理的动态协作时速度较慢。接着，他们介绍了一种新的基于复杂网络图模型的方法，以解决这个问题。复杂网络图模型能够有效地模拟信息在多代理系统中的交互和传播过程。通过计算新信息的语义值，这个模型能够量化信息的重要程度。然后，作者提出了局部图评估算法。这个算法的关键在于，它避免了全图搜索的高计算成本，转而采用局部评估策略。这种方法能够更快地反映出信息图的变化，而且可以将这些评估转化为易于执行的数据库查询，从而帮助团队成员迅速做出决策。为了证明这个算法的实用性，作者进行了多主体合作的案例研究。案例分析结果表明，提出的快速信息评估方法能够在保持高效性的同时，有效地支持团队内的信息共享和决策过程。这项研究为多代理系统的协同信息共享提供了一个高效且适应性强的评估工具，这对于动态环境中的决策制定具有重要意义。未来的研究可能会进一步优化这个算法，以提高其精度，并扩展到更复杂的多代理系统中。" "快速的基于图形的协同信息共享信息评估算法是一种针对多代理团队在动态环境中的协作决策问题的解决方案。该算法利用复杂网络图模型，结合局部图评估算法，提高了信息评估的速度，同时保持了信息共享的效能。通过案例研究，证明了该方法在实际应用中的可行性。"

A Fast Graphic-based Information Valuation

Algorithm for Cooperative Information Sharing

Haixiao Hu, Yang Xu*, Yulin Zhang, Ming Liu

School of Computer Science and Engineering, University of Electronic Science and Technology of China

Chengdu, Sichuan, 611731, P.R.China

Email: xuyang@uestc.edu.cn

Abstract—Information sharing is critical to multi-agent team

for cooperative decision making in dynamic and partially ob-

servable environments. Other than building a full information

coverage, if agents in a team can be self-directed to valuate a

potential receiver and where to communicate, the coordination

efﬁciency will be greatly enhanced. Although intensive studies

of information valuation approaches have been developed in

ontology graph matching and natural language processing, these

models fail to perform fast reasoning for large-scale decentralized

agents dynamic coordination. In this paper, we propose a fast

information valuation approach based on a complex network

graph model, which helps to indicate the information importance.

Similar to vague information valuation by human, the key is that

important information always signiﬁcantly changes their complex

information graph with its incorporation. Therefore, we calculate

the semantic based value of this new information in a graph

model and build a local graph evaluation algorithm to estimate

information graph evolution, instead of performing expensive

complete graph search. The advantage is that the local valuation

algorithm can be easily transformed into efﬁcient queries in

agents’ information base so that they can make fast decisions.

Although the decision may not be precise, similar to human

communication, the information sharing performance is good

enough to disseminate valuable information in the multi-agent

team. We demonstrate the feasibility of the proposed information

valuation approach in a multi-agent cooperation case study.

I. INTRODUCTION

Multi-agent coordination is popular in various application

domains including disaster response, anti-terrorism and explo-

ration [8]. In those systems, agents are required to complete

tasks cooperatively to achieve the common goal. However,

when the number of agents scale up, they are highly distributed

and have to operate in a dynamic and partially observable

environment. In this case, information sharing is critical for

distributed agents to make rational decisions [11]. Although

intensive studies have been carried out, state of the art fails to

build practical algorithms for large-scale multi-agent systems,

where agents are required to evaluate how a potential receiver

could beneﬁt from a given information whilst considering

the communication cost. With decision-theoretic approaches,

decentralized agents inherit the constraint of partial observ-

ability and their valuation falls within DEC-POMDP model,

which is intrinsically NEXP-COMPLETE. Therefore, precise

computation of information utilities is either infeasible or

computational costly in a large system [2].

Heuristic approaches for the information sharing problem

always start from the assumption that communication cost can

be rationally estimated [4]. Given a good knowledge of the

potential receiver, e.g., in [10], the receivers’ knowledge can

be built solely from their incoming messages, the agent should

be able to score the importance of the new information if it is

incorporated into the receiver’s knowledge. There have been

many approaches. For example, most straightforwardly, the

information importance can be crudely counted by checking

whether it is already known by the receiver explicitly or

implicitly [7]. Some other researches calculate agent’s belief,

and approximate the information importance with two criteria:

improvement to reach a synchronized belief [9] or reduction

of the uncertainty [3]. However, those algorithms are of high

computation cost to perform intensive information base scan

throughout to build statistical or probability inference model,

and have poor performance in a large multi-agent system. In

the semantic web community, the information base is always

abstracted into a semantic graph [5]. Based on this graph,

either statistic or graph searching algorithms are put forward

[12]. In such models, agents in a team have to maintain a

complete view of knowledge graphs for each potential receiver

so that information valuation can be performed. However,

it is of high memory cost as well as long decision making

times, due to its computational complexity. Therefore, a fast

information valuation method that matches the needs of large-

scale multi-agent information sharing decision is still very

necessary.

II. PROBLEM STATEMENT AND MODELING

A. Information Sharing Problem

In large teams, decentralized agents have to make decisions

about how to share information based on their limited obser-

vation of the team. Due to their heterogeneities, not all agents

in the system are interested in a given piece of information

[13]. In this case, other than building full information coverage

around the team, information sharing aims at ensuring that

agents get the information they are interested in at a timely

manner. For example, in a city rescue domain, an agent

driving the ﬁre track cares more about the city trafﬁc condition

than the news of ﬁnding a victim. Therefore, to effectively

share information, agents have to estimate their teammates’

information needs from their own knowledge.

More generally, let agent a be the candidate receiver of

information I, and IB = {I

, I

, ..., I

, ...} denotes its infor-

mation base, which contains a set of information known or

2017 IEEE International Conference on Systems, Man, and Cybernetics (SMC)

Banff Center, Banff, Canada, October 5-8, 2017

下载后可阅读完整内容，剩余5页未读，立即下载

weixin_38707153

粉丝: 7
资源: 949

复杂网络图模型的信息评估：协同信息共享的新方法

COA郊狼算法_COA郊狼算法_coa算法_

【协同任务】蚁群算法多组群无人机协同任务路径规划【含Matlab源码 1578期】.zip

【白冠鸡优化算法】基于白冠鸡优化算法求解单目标优化问题（COOT）含Matlab源码.zip

基于定向A*算法的多无人机航迹规划分步策略

地质导向钻井信息可视化与协同决策虚拟系统

基于GPU平台的全搜索运动估计算法优化研究与实现之软件工程分析.docx

行业分类-电子-关于基于GIS技术的电网检修工作强度图形化展示方法的说明分析.rar

DV-Hop定位算法

航迹规划算法Astar-KO (1),航迹规划算法的优势,matlab

matlab实现合作协同进化算法优化二维函数源码

最新资源