We consider a system of cooperating robots transporting an object in which each individual robot is small and resource constrained and the number of robots is large precluding architectures that support a centralized coordination and control system. In nature, a swarm of ants can transport large payloads that are beyond the capability of individual organisms by using simple strategies that only require coordination with their immediate neighbors. We adopt a similar paradigm in which robots communicate with their immediate neighbors although the wireless communication links that are typical in small robots can be noisy and therefore not robust. We derive decentralized coordination algorithms that allow robots to converge to a load distribution system that is optimal in the sense of minimizing interaction forces while achieving the desired resultant wrench to allow manipulation and transport. We demonstrate the key ideas using simulations for 100 robots.