Design of a Decentralized Strategy for Layered Self-Assembly of 3D Structures Using Robotic Blocks

In self-assembly tasks, local interactions between robotic structure parts induce a collective behaviour that guides the robots to assume a desired shape. In this work, we propose a self-assembly strategy for building three-dimensional structures using robotic blocks. The assembly is executed layer-by-layer, and each layer grows from a single position called seed position. The robots follow a set of pre-programmed behaviours to perform the assembly task. Moreover, each robot is only capable of local sensing and empty positions in the structure are not known a priori. Robots carry a blueprint of the structure, containing relative coordinates for where blocks should be placed. A graph is extrapolated from the blueprint and used to define the structural properties necessary to analyze correctness and efficiency of the proposed technique. Simulated results show that the distance that robots travelled during assembly is likely to be the shortest path possible. Also, graph-based metrics are applied to evaluate the selection of different seed positions, drawing a relationship between structure blueprint and the distance travelled by the robots. Finally, physical experiments demonstrate the applicability of the proposed algorithms in realistic scenarios.