Prediction-Based Dynamic Load Balancing Using Agent Migration for Multi-agent System

Multi-agent system in the ubiquitous computing environment can provide customized services to the users by effectively utilizing the distributed resources. The existing dynamic load balancing approach invokes the migration of agents even for temporal imbalance of the loads. In this paper we propose a prediction-based dynamic load balancing scheme which can effectively avoid unnecessary agent migration. An experiment reveals that the proposed approach significantly reduces the service response time compared with the existing scheme. It enables the agent system to quickly adapt to the environment change through effective load balancing.