This paper addresses the design of control policies for groups of up to 16 simple autonomous mobile robots (called s-bots) for the cooperative transport of heavy objects of different shapes and sizes. The s-bots are capable of establishing physical connections with each other and with the object (called prey). We want the s-bots to self-assemble into structures which pull or push the prey towards a target location. The s-bots are controlled by neural networks that are shaped by artificial evolution. The evolved controllers perform quite well, independently of the shape and size of the prey, and allow the group to transport the prey towards a moving target. Additionally, the controllers evolved for a relatively small group can be applied to larger groups, making possible the transport of heavier prey. Experiments are carried out using a physics simulator, which provides a realistic simulation of real robots, which are currently under construction.