2015 IEEE / WIC / ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT) (2015)
Dec. 6, 2015 to Dec. 9, 2015
The organizational design of a multi-agent system (MAS) is important for its efficiency, adaptability and robustness. However, finding suitable organizational structures for different MASs is a challenging problem. In this paper, we propose a Framework of Evolutionary Optimization for Agent Organizations (FEVOR) based on Genetic Programming for optimizing tree-structured MASs. FEVOR employs a flexible representation of organizations and may be applied to a wide range of organizational forms such as pure hierarchies, holarchies, and federations. Compared to existing work, FEVOR is capable of efficient quantitative search and less vulnerable to stalling at local optima due to its non-greedy nature. Extensive experiments for optimizing an information retrieval system have been conducted to demonstrate the advantages of FEVOR in generating suitable MAS organizations for adaptive environments.
Organizations, Peer-to-peer computing, Optimization, Sociology, Statistics, Algorithm design and analysis, Genetic programming
B. Li, H. Yu, Z. Shen, L. Cui and V. R. Lesser, "An Evolutionary Framework for Multi-agent Organizations," 2015 IEEE / WIC / ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT), Singapore, Singapore, 2015, pp. 35-38.