Learning in Immune Network Algorithm for Multi-Robot Cooperation

Abstract

The multi-robot system frequently associated with the problem of robot coordination and cooperation as it requires real-time and distributed control. This paper describes biological immune system, immune response, and immune learning through somatic hypermutation. The relationship between immune system and multi-robot system is presented to show the connection between both systems. To improve the cooperative behavior in multi-robot systems, an immune network algorithm is proposed with the extension of learning ability. Jerne and Farmer models of immune network are referred as the foundation of our approach. The proposed algorithm is based on our previous conceptual model and designed particularly for multi-robots foraging task with five different action strategies. The learning concept in the antibody is applied to the robot action. Therefore, the robot swarm is expected to complete the task faster since robots adapt to the environment. For future work, the proposed algorithm will be implemented in a robot simulation environment called ARGoS.