LEARNING AND STABILIZATION OF ALTRUISTIC BEHAVIORS IN MULTIAGENT SYSTEMS BY RECIPROCITY

Optimization of performance in collective systems often requires altru ism. The emergence and stabilization of altruistic behaviors are diffi cult to achieve because the agents incur a cost when behaving altruist ically. In this paper? we propose a biologically inspired strategy to learn stable altruistic behaviors in artificial multi-agent systems, n amely reciprocal altruism. This strategy in conjunction with learning capabilities make altruistic agents cooperate only between themselves, thus preventing their exploitation by selfish agents, if future benef its are greater than the current cost of altruistic acts. Our multi-ag ent system is made up of agents with a behavior-based architecture. Ag ents learn the most suitable cooperative strategy for different enviro nments by means of a reinforcement learning algorithm. Each agent rece ives a reinforcement signal that only measures its individual performa nce. Simulation results show how the multi-agent system learns stable altruistic behaviors, so achieving optimal(or near-to-optimal) perform ances in unknown and changing environments.