A Self-Organizing Map Based Approach to Adaptive System Formation

Multi-agent systems are considered to be potential solutions to complex tasks. Cellular self-organizing (CSO) multi-agent systems have been proposed that take a field-based approach to regulate agent behaviors. One difficulty in designing CSO systems is to generate rules to map given tasks to agent behaviors. This paper proposes an approach for adaptive system formation based on a field analysis and self-organizing map (SOM) algorithm. The tasks are captured as multiple task fields. The relationship among the agents is translated into a social field. Each agent has multiple function modes corresponding to the task fields. SOM and a function mode selection algorithm are devised to match the social field of the system with the task fields. Computer simulations have demonstrated the effectiveness of this approach and its potential in designing CSO systems for solving system formation tasks.