Autonomous Multi-material Construction with a Heterogeneous Robot Team.

We present a construction model that allows robots with different construction capabilities, using materials of different physical properties and sizes, to modify unstructured environments in a distributed system. Building steps are computed reactively so that they can respond to changes in the environment and imperfect assembly. The reactive approach allows robots to coordinate and add material to the same structure. Each robotic agent uses an abstract model of the environment to compute a set of legal construction steps based on its current knowledge of the world, and we show that in this setting more knowledge results in more legal moves. We exploit this capability by letting the system use a variety of materials and choose the most appropriate material given its current knowledge of the state of the structure. We demonstrate the approach by running the system on a variety of terrains and with mixed materials, including both deformable and rigid components.