Collective transport of complex objects by simple robots: theory and experiments

Abstract

Ants show an incredible ability to collectively transport com-plex irregular-shaped objects with seemingly simple coor-dination. Achieving similarly effective collective transport with robots has potential applications in many settings, from agriculture to construction to disaster relief. In this pa-per we investigate a simple decentralized strategy for collec-tive transport in which each agent acts independently with-out explicit coordination. Using a physics-based model, we prove that this strategy is guaranteed to successfully trans-port a complex object to a target location, even though each agent only knows the target direction and does not know the object shape, weight, its own position, or the position and number of other agents. Using two robot hardware plat-forms, and a wide variety of complex objects, we validate the strategy through extensive experiments. Finally, we present a set of experiments to demonstrate the versatility of the simple strategy, including transport by 100 robots, trans-port of an actively moving object, adaptation to change in goal location, and dealing with partially observable goals.