PUB - Publikationen an der Universität Bielefeld

It is still a challenging problem to self-reconfigure modular robots to different morphologies to adapt to dynamic environments. To tackle this problem, a new computational framework inspired from biological morphogenesis is suggested in this chapter. First, we introduce two reconfigurable modular robots, Cross-Cube and Cross-Ball, which are developed for various complex pattern reconfigurations using the pro-posed morphogenetic approach. Then, a hierarchical morphogenetic self-reconfiguration model is presented for both Cross-Cube and Cross-Ball. In this hierarchical model, the layer 1 controller is responsible for the adaptive pattern generation based on the current environmental constraints and the task requirements in hands. The layer 2 controller automatically generates target configurations to guides the modules to converge into the target pattern. Both the layer 1 and layer 2 controllers are generic and can in principle be applied to other reconfigurable modular robots. The controller in layer 3 is hardware dependent that mainly deals with the physical constraints of module movements. This hierarchical morphogenetic model is applied to each module of the reconfigurable modular robot in a distributed manner, where each module makes its configuration movements only based on its local sensory information and shares information with its local neighboring modules. Extensive simulation results have demonstrated the feasibility and efficiency of the proposed module design as well as the corresponding hierarchical morphogenetic model for both Cross-Cube and Cross-Ball modular robots to construct various configurations.