Coordinating Adaptive Behavior for Swarm Robotics Based on Topology and Weight Evolving Artificial Neural Networks

Abstract

Swarm robotics is the research field of multi-robot systems which consist of many homogeneous autonomous robots without any type of global controllers. Generally, since a task given to this system can not be achieved by a single autonomous robot, emergent cooperative behavior is expected in a robot swarm by a certain mechanism through the interactions among the robots or with an environment. In this paper, an evolutionary robotics approach, i.e., the method that the robot controllers are designed by evolutionary algorithms with artificial neural networks, is applied. Among many approaches to evolving artificial neural networks, two approaches, NEAT and MBEANN, are adopted for computer simulations. Although a conventional neural network evolves only synaptic weights with the condition of a fixed topology, NEAT and MBEANN evolve not only their synaptic weights but also their topologies. As a benchmark of swarm robotics, cooperative package pushing problems by ten autonomous robots are conducted to examine their performance. The emerged behavioral characteristics are discussed.