Aggregation for modular robots in the pivoting cube model

Abstract

In this thesis, we present algorithms for self-aggregation and self-reconfiguration of modular robots in the pivoting cube model. First, we provide generic algorithms for aggregation of robots following integrator dynamics in arbitrary dimensional configuration spaces. We describe solutions to the problem under different assumptions on the capabilities of the robots, and the configuration space in which they travel. We also detail control strategies in cases where the robots are restricted to move on lower dimensional subspaces of the configuration space (such as being restricted to move on a 2D lattice). Second, we consider the problem of finding a distributed strategy for the aggregation of multiple modular robots into one connected structure. Our algorithm is designed for the pivoting cube model, a generalized model of motion for modular robots that has been effectively realized in hardware in the 3D M-Blocks. We use the intensity from a stimulus source as a input to a decentralized control algorithm that uses gradient information to drive the robots together. We give provable guarantees on convergence, and discuss experiments carried out in simulation and with a hardware platform of six 3D M-Blocks modules.