Source-linked AI summary
Agent-based computing from multi-agent systems to agent-based Models: a visual survey
Muaz A. Niazi, Amir Hussain
TL;DR
The broad application base creates confusion about the exact semantics of terms in the literature. The paper presents a visual and scientometric survey connecting agent-based computing across diverse areas, including non-computing domains, while noting limits on campus-level attribution and detailed sub-domain analysis.
Problem
The broad application base leads to confusion regarding the exact semantics of various terms in the literature.
Method
The paper presents a detailed visual and scientometric survey spanning multi-agent systems and artificial intelligence.
Results
The analysis links diverse areas through citing documents and reveals agent-based computing across non-computing domains ranging from life sciences to ecological sciences.
Takeaways & Limitations
Agent-based computing spans interconnected research areas beyond computing, including life sciences and ecological sciences.
Takeaways & Limitations
The survey does not provide detailed analyses of each sub-domain individually.
Abstract
from arXiv · showhide
Agent-Based Computing is a diverse research domain concerned with the building of intelligent software based on the concept of "agents". In this paper, we use Scientometric analysis to analyze all sub-domains of agent-based computing. Our data consists of 1,064 journal articles indexed in the ISI web of knowledge published during a twenty year period: 1990-2010. These were retrieved using a topic search with various keywords commonly used in sub-domains of agent-based computing. In our proposed approach, we have employed a combination of two applications for analysis, namely Network Workbench and CiteSpace - wherein Network Workbench allowed for the analysis of complex network aspects of the domain, detailed visualization-based analysis of the bibliographic data was performed using CiteSpace. Our results include the identification of the largest cluster based on keywords, the timeline of publication of index terms, the core journals and key subject categories. We also identify the core authors, top countries of origin of the manuscripts along with core research institutes. Finally, our results have interestingly revealed the strong presence of agent-based computing in a number of non-computing related scientific domains including Life Sciences, Ecological Sciences and Social Sciences.
Introduction
Agent-based computing spans agent design, multi-agent systems, and simulation across diverse scientific domains, creating substantial terminological and conceptual overlap. The paper addresses this breadth through citation analysis and visualization of shared concepts, journals, categories, authors, institutions, and manuscript origins.
- Agent-based computing encompasses both agent design for validating future physical-agent operation and simulation for understanding emergent phenomena.
- Its communities may share little beyond the notion of an agent, contributing to confusion about terminology across the literature.
- Agent-based modeling overlaps with multiagent systems and synonymous labels such as individual-based modeling within agent-based computing.
- Applications extend from social and biological sciences to environmental, business, wireless-sensor, and ad-hoc network modeling.
- The study uses citation analysis and visualization to survey the domain's diversity, identify shared concepts, and map core journals, subject categories, authors, institutes, and countries.
Background
Scientometric visualization combines quantitative relationship mapping with visual displays of scientific domains and their evolution. The paper applies these ideas to agent-based computing through co-citation, clustering, and network-based analysis.
- Scientometrics quantitatively studies scientific communication using citation, social-network, and related techniques to map relationships among knowledge-based entities.
- Domain visualization represents large amounts of research-front data at a high-level view of a scientific map.
- Co-citation analysis measures document proximity and distinguishes journal, author, and document co-citation perspectives.
- Author co-citation visualization uses frequent joint citation as an indicator of closeness between authors and their bodies of literature.
- The survey analyzes agent-based computing with co-citation networks, complex-network methods, Network Workbench, and CiteSpace.
- Its objectives include identifying paper clusters, topic bursts and timelines, core journals, subject categories, and productive authors, institutes, and countries.
Methodology
The study constructs a Web of Science corpus using topic searches covering major agent-based-computing subdomains from 1990 to 2010. It combines Network Workbench for global complex-network analysis with CiteSpace for citation visualization and cross-validation.
- The data were retrieved from Thomson Reuters Web of Science and restricted to journal articles.
- The topic search covered agent-based, multi-agent-based, and individual-based models; agent-oriented software engineering; and AI agents and multi-agent systems.
- Searches covered SCI-EXPANDED, SSCI, A&HCI, and CPCI-S, with analysis restricted to 1990–2010.
- Bibliographic records included authors, titles, abstracts, and references, with cited references producing 32,576 nodes.The search-keyword details and selection rationale were documented in the appendix.
- The two-tool design enabled both complementary use of tool strengths and cross-validation by comparing their outputs.
- Network Workbench supported global complex-network analysis, while CiteSpace provided citation-network visualization and was used as the key visualization tool.
Results
The analysis traces agent-based computing’s growth, citation structure, clusters, and institutional spread, revealing links between computing and non-computing research areas. Visualization highlights established publication venues, cross-disciplinary clusters, and connections between multiagent systems and agent-based modeling.
- 148 articles were published in 2009, following gradual growth from the early 1990s.
- 1,630 citations were recorded in 2009, rising from a very small number in earlier years.
- Analysis using NWB: 32,576 nodes formed the extracted citation network, which included cited references as well as original records.
- Analysis using NWB: The citation graph contained 78 weakly connected components, with the largest component comprising 31,104 nodes.
- CiteSpace: Key journals and authors spanned computer science, ecology, biology, and related fields, while the United States led manuscript origins followed by England, China, Germany, France, and Canada.
- CiteSpace: Index-term clustering linked agent-based modeling, multiagent systems, ecology, social sciences, economics, and biosciences through citing documents.
Conclusions and Future Work
The paper presents a visual and scientometric survey of agent-based computing across journal articles and diverse sub-domains. Its analysis finds substantial representation beyond computer science, while identifying further sub-domain analysis as future work.
- Conclusions and Future Work: The analysis identifies results from the entire agent-based computing domain rather than from a single sub-domain.The paper describes its contribution as a detailed visual and scientometric survey of the domain.
- Conclusions and Future Work: The survey covers journal articles across agent-based modeling and simulation, agent-based software engineering, multi-agent systems, and artificial intelligence.The analysis uses data from recognized Web of Science databases and aims to cover these diverse areas within agent-based computing.
- Conclusions and Future Work: Despite roots in computer science, agent-based computing has relatively high citation presence across non-computing domains.The paper specifically identifies life sciences, ecological sciences, and social sciences among these domains.
- Conclusions and Future Work: The survey reports that agent-based computing appears in life sciences, ecological sciences, and social sciences.These domains are presented as part of the field’s broader spread beyond computing.
- Conclusions and Future Work: Future work will analyze individual sub-domains in greater detail and extend visualization-based analysis to areas such as Cybernetics and Consumer Electronics.Planned work includes multi-agent systems, agent-oriented software engineering, and agent-based modeling.
Appendix A: Details of Search Keywords
The search strategy was designed to retrieve papers focused on agent-based modeling or multiagent systems while reducing ambiguity from shared terminology. The query searched titles and returned 1,064 records on 8 September 2010.
- Appendix A: Details of Search Keywords: The search used keywords shared across sub-domains and potentially overlapping with chemical or biological terminology.The authors note that terms such as agent could refer to biological or chemical agents.
- Appendix A: Details of Search Keywords: The authors limited retrieval to papers focused on agent-based modeling or multiagent systems.This restriction addressed ambiguity in the shared keyword vocabulary.
- Appendix A: Details of Search Keywords: The search was performed on article titles using an ISI Web of Knowledge query.The displayed query begins with title terms for agent-based, individual-based, multi-agent, multiagent, and ABM variants.
- Appendix A: Details of Search Keywords: 1,064 records were retrieved on 8 September 2010.The date and record count are reported together as the retrieval metadata.