This research compares three different artificial evolution approaches to the design of cooperative behavior in a group of simulated mobile robots. Two approaches specify robots that share a single genotype, where the second includes a learning mechanism, and the third uses different genotypes to specify each robot. The application domain is a multiple predator-single prey scenario in which a team of robots, termed: predators, collectively work to capture or slow a single robot, termed: the prey. Results indicate that the multiple genotype approach is superior in terms of deriving robust cooperative behaviour, given that this approach facilitates behavioural specialization in the predator team.