Large-scale robotic self-assembly using alignment-agnostic docking

Swissler, Petras Juozas

doi:https://doi.org/10.21985/n2-wr8d-2e60

Work

Large-scale robotic self-assembly using alignment-agnostic docking

Public

Download PDF

Download All Files (.zip)

This thesis presents a new self-assembling robotic system that diverges from traditional approaches by enabling robots to build structures not constrained to a lattice and formed based on environmental conditions, rather than on a priori blueprints. Robot self-assembly is a class of behavior in swarm robotics in which many robots join their bodies to form large structures. The field of swarm robotics, which studies the large-scale cooperation of robot groups presents many benefits relative to traditional robotics. In general, robot swarms are extremely well-suited to parallelizable tasks, are highly adaptable with regards to the types of tasks they take on, are robust to failures in any individual member, and are highly scalable. Unfortunately, traditional lattice-based and blueprint-based approach to robot self-assembly are often at odds with these benefits of swarm robotic systems, limiting their use to highly controlled environments. Inspired by the techniques social insects use for self-assembly, I developed a novel hardware-algorithm system able to form non-latticed structures based solely on environmental interactions. Specifically, this system includes: the Continuous Docks, an attachment mechanism enabling robot to form robust and reliable connections without requiring specific alignment; the FireAnt series of robots, which use these docks to attach to like robots and which independently climb over arbitrary arrangements of peer robots; and the ReactiveBuild algorithm, which enables robots to build a variety of structures tailored to their environment using only a simple set of rules. I validated the hardware portions of this work through a series of experiments and demonstrations and validated algorithmic portions through extensive simulation. The research towards this thesis presents a profoundly different way of approaching the robotic self-assembly problem and addresses many longstanding challenges in this field. By enabling robots to adapt the structure to arbitrary environments and to attach regardless of contact location, the work presented here makes good on many of the promised benefits of the swarm robotics approach. The work also provides a path for these self-assembling robotic systems to move from the tightly controlled lab environments in which they almost exclusively operate to the messy and uncontrolled environments of the outdoor world. It is only by this move out of the lab that we can realize a future of ubiquitous use of robotic self-assembled structures in real-world applications.

Creator