An agent-based collaborative evolutionary model for multimodal optimization

A novel approach to multimodal optimization called Roaming Agent-Based Collaborative Evolutionary Model (RACE) combining several evolutionary techniques with agent-based modeling is proposed. RACE model aims to detect multiple global and local optima by training a multi-agent system to employ various evolutionary techniques suitable for a specified multimodal optimization problem. Agents can exchange information during the search process enabling a cooperative search of optima between several populations evolving independently. Redundant search by multiple agents is avoided by having them communicate and negotiate about the space region searched. An agent can request and receive from another agent valuable information and genetic material for a better search of a certain region in the environment. Performance of the proposed agent-based collaborative evolutionary model is compared by means of numerical experiments with rival evolutionary techniques. Categories and Subject Descriptors...

• Type: Conference
• Year: 2008
• Where: GECCO
• Authors: Rodica Ioana Lung, Camelia Chira, Dumitru Dumitrescu
• Keywords: Collaborative Evolutionary Model, Evolutionary Techniques, GECCO 2008, Multimodal Optimization, Optimization