An agent-based clustering algorithm using potential fields

Abstract

In this paper, a novel clustering algorithm using an agent-based architecture along with a force-based clustering algorithm is proposed. To this end, a set of simple mobile agents that have limited processing power is used. These agents communicate in a pairwise manner to exchange their position information. As opposed to the bio-inspired clustering algorithms that need a set of local rules to specify the agent movements, in this paper the agent motions are driven from attractive and repulsive potential fields that are created by the data points and the other agents respectively. Each agent moves according to the resulted force from applying the potential fields and announces its next position to the other agents. The movement of agents is continued until all of them reach equilibrium points.978-1-4244-1968-5/08/$25.00 ©2008 IEEE