The advantage of reconfiguration is central to modular robotic systems. With this benefit, however, comes a complex and interesting challenge: how does a modular robot recognize which shapes are useful or familiar? The ability for a modular robot to determine which configurations are needed for various tasks is a fundamental requirement for increased autonomy.

For example, if a modular robot forms into a snake-like configuration, it should recognize its current state and select the correct corresponding slithering motions. In this work, we propose a general approach to solving the configuration recognition problem for modular robots. Our work is distinguished from previous work in that we describe our system from the perspective of a modular robot recognizing its own shape (rather than a program giving instructions on how to construct a robot given a set of parts), we introduce a new, succinct matrix representation of modular robots, we implement our algorithm on our CKBot system, and we show the generality of the approach and how it extends beyond simple configuration-dependent gaits.

Publication