Date: May 2011
A robot neural controller where obstacle-avoidance and random-walk neurons observe level of their irritation and randomly reconfigure their connections to motors when the irritation reaches a specified threshold is presented and tested.
Individual neurons have self-recurrent connections and receive function-specific input. The neural controller shows either robustness or adaptivity to noise and sensory and motor function perturbations by utilising ultrastability and subsumption principles.
Keywords:
Obstacle avoidance, random walk, artificial neural network, ultrastability, homeostasis
{Please enable JavaScript in order to post comments}