Ultrastable neuroendocrine robot controller

Abstract

The work of a simulated neuroendocrine controller with ultrastable neurons and glands is sketched and tested in terms of stability and adaptability. The artificial neurons connect to each other and to motors, while hormones produced by behaviour-related glands regulate their output. The ultrastable nature of the cells allows them to maintain their homeostasis by random reconfiguration of their connections and parameters without reference to the global goal of the system. Interactions of these ultrastable components cause individual robot behaviours to emerge to certain extents. The presented results show that the controller as a whole is capable of not only configuring itself to perform random walk, obstacle avoidance, mineral collection and recharging, but also to stay robust or adapt to a number of environmental and body perturbations without a need for a body model.