Self-repair during continuous motion with modular robots

Through the use of multiple modules with the ability to reconfigure to form different morphologies, modular robots provide a potential method to develop more adaptable and resilient robots. Robots operating in challenging and hard-to-reach environments such as infrastructure inspection, post-disaster search-and-rescue under rubble and planetary surface exploration, could benefit from the capabilities modularity offers, especially the inherent fault tolerance which reconfigurability can provide. With self-reconfigurable modular robots self-repair, removing failed modules from a larger structure to replace them with operating modules, allows the functionality of the multi-robot organism as a whole to be recovered when modules are damaged.

Previous self-repair work has, for the duration of self-repair procedures, paused group tasks in which the multi-robot organism was engaged, this thesis investigates Self-repair during continuous motion, ``Dynamic Self-repair", as a way to allow repair and group tasks to proceed concurrently. In this thesis a new modular robotic platform, Omni-Pi-tent, with capabilities for Dynamic Self-repair is developed. This platform provides a unique combination of genderless docking, omnidirectional locomotion, 3D reconfiguration possibilities and onboard sensing and autonomy. The platform is used in a series of simulated experiments to compare the performance of newly developed dynamic strategies for self-repair and self-assembly to adaptations of previous work, and in hardware demonstrations to explore their practical feasibility. Novel data structures for defining modular robotic structures, and the algorithms to process them for self-repair, are explained.

It is concluded that self-repair during continuous motion can allow modular robots to complete tasks faster, and more effectively, than self-repair strategies which require collective tasks to be halted. The hardware and strategies developed in this thesis should provide valuable lessons for bringing modular robots closer to real-world applications.