Table 1. Experimental set-up.

Experiment
Number of robots	100
Simulation length	1,000,000 time-steps
Number of repeats	64
Number of pucks	150 green, 150 red, immediately replaced in a random location when picked up
Arena	See Fig. 1
Controller
Controller	Feed forward neural net with activation function
Input nodes	8 obstacle sensors, 16 puck detectors, 2 bias and 2 recurrent nodes
Output nodes	2 (left and right motor values)
Evolution
Representation	Real valued vectors
Chromosome length	54
Initial weight distribution	Randomly drawn from distribution
Mutation	Gaussian perturbation
Parent selection	Rank-based
Robot lifetime	2000 time-steps
Egg-phase duration	200 time-steps
Communication range	ca. 9 body lengths
Crossover	none

The standard settings for experiments reported in this paper. Some experiments vary one or more of these values as indicated in the experiment description.