Solving All-to-All Communication with CA Agents More Effectively with Flags

摘要

We have investigated the all-toall communication problem for a multi-agent system modeled in cellular automata. The agents' task is to solve the problem by communicating their initially mutually exclusive information to all the other agents. In order to evolve the best behavior of agents with a uniform rule we used a set of 20 initial configurations, 10 with border, 10 with cyclic wrap-around. The behavior was evolved by a genetic algorithm for agents with (1) simple moving abilities, (2) for agents with more sophisticated moving abilities and (3) for agents with indirect communication capabilities (reading and writing flags into the environmental cells). The results show that the more sophisticated agents are not only more effective but also more efficient regarding the effort that has to be made finding a feasible behavior with the genetic algorithm.