Sequence learning in mobile robots using avalanche neural networks

Abstract

This paper describes the implementation of a neural network for sequence learning that is based on a neurocomputational theory of learning. The network is implemented on a physical mobile robot in order to learn to reproduce sequences of motor actions. At the onset of a conditioned stimulus the robot is presented with a sequence of visual stimuli that produce reactive motor actions of different duration. Initial results show that after learning the robot can approximate the motor sequence with no visual stimulation. © Springer-Verlag Berlin Heidelberg 2001.