Sensor-Driven Algorithm for Self-Reconfiguration of Modular Robots

The problem addressed is the deterministic 2D reconfiguration of a connected mass of identical hexagonal robots from an initial well-defined and restricted set of arbitrary shape configurations composed of connected modules (I), to overlapping, connected straight chain goal configurations (G). We present algorithms to accomplish system-wide shape change using only a constant amount of information at each processor, relying on the sensory information gathered via light and contact sensors to complete the reconfiguration. In our algorithms, a module uses only sensory information obtained from sensors on its six adjacent cells. No preprocessing is needed in this algorithm if' I adheres to certain admissibility conditions. The addition of a preprocessing algorithm increases the number of admissible configurations possible for I while adding information to be stored at each robot because each module has to be aware of the position of modules on the system periphery.