Dr Michael Herrmann – University of Edinburgh
Self-Organized Criticality in Neurorobotics
Self-organised criticality is a principle for the generation of complexity in nature. It is characterised by power-law event distributions and has been used to describe phenomena in domains such as the dynamics of neural activity, natural evolution and biological motor control. We show that critical behaviour is brought about in a natural way in neural networks and can be achieved in autonomous robots by the optimisation of conflicting goals. The critical dynamics in these systems leads to self-organisation of behavioural options that later can be composed into meaningful behaviours e.g. by reinforcement learning. Furthermore, we discuss applications of the approach in prosthetics as well as methods for guidance of the process of behavioural self-organisation in order to bias the emerging behaviour towards promising regions in the behavioural space, to include background knowledge or to achieve more coherent representations.
School of Computing, Robert Gordon University, St Andrew Street, Aberdeen, Lecture Room A12, 14:00 – 15:00.