Vision-Based Self-Assembly For Modular Multirotor Structures

Modular aerial robots can adapt their shape to suit a wide range of tasks, but developing efficient self-reconfiguration algorithms is still a challenge. Self-reconfiguration algorithms in the literature rely on high-accuracy global positioning systems which are not realistic for real-world applications. In this letter, we study self-reconfiguration algorithms using a combination of low-accuracy global positioning systems (e.g., GPS) and on-board relative positioning (e.g. visual sensing) for precise docking actions. We present three algorithms: 1) parallelized self-assembly sequencing that minimizes the number of serial "docking steps"; 2) parallelized self-assembly sequencing that minimizes total distance traveled by modules; and 3) parallelized self-reconfiguration that breaks an initial structure down as little as possible before assembling a new structure. The algorithms take into account the constraints of the local sensors and use heuristics to provide a computationally efficient solution for the combinatorial problem. Our evaluation in 2-D and 3-D simulations show that the algorithms scale with the number of modules and structure shape.