Sensor-Coupled Fractal Gene Regulatory Networks for Locomotion Control of a Modular Snake Robot

In this paper we study fractal gene regulatory network (FGRN) controllers based on sensory information. The FGRN controllers are evolved to control a snake robot consisting of seven simulated ATRON modules. Each module contains three tilt sensors which represent the direction of gravity in the coordination system of the module. The modules are controlled locally and there is no explicit communication between them. So, they can synchronize implicitly using their sensors, and coordination of their behavior takes place through the environment. In one of our experiments, all the three tilt sensors are available for the FGRNs and a simple controller is evolved. The controller is a linear mapping of one input sensor to the output. It is only based on one sensor input and ignores the other sensors as well as the regulatory part of the network. In another experiment, the controller's input uses one of the other sensors that carries less information. In this case, the evolved controller blends sensory information with the regulatory network capabilities to come up with a proper distributed controller.