A comparison of approaches to the evolution of homogeneous multi-robot teams

The application of artificial evolution to the design of homogeneous multi-robot teams encounters the basic yet important issue of how such teams are to be generated. The standard approach used with homogeneous systems is to evaluate teams comprising identical copies of a single evolutionary individual. However an alternative would be to use a separate evolutionary individual to specify each member of a team. Intuitively the former seems a far better choice, however, so little consideration has been given to the latter approach that there is insufficient empirical evidence on which to discount it. The paper reports on a comparison of the two approaches over multiple runs in the context of a non-trivial co-operative task carried out by simulated mobile robots controlled by arbitrarily recurrent neural networks. Surprisingly, it was found that the latter approach performed significantly better than the former. Analysis and further experimentation were undertaken yielding a possible explanation of this result