Optimization of a robotics gaze control system

Abstract

Gaze control is an important issue in the interaction between a robot and humans. In particular, the decision of who to pay attention to in a multi-party conversation is a way of improving a robot’s naturalness in human-robot interaction. A system based on a competitive neural network can decide who to look at with a smooth transition in the focus of attention when significant changes in stimuli are produced. One important aspect in this process is the configuration of the different parameters of such a neural network. The weights of the different stimuli have to be computed in order to achieve behavior similar to humans. This article explains how these weights can be obtained by solving an optimization problem. The experiments carried out and some results are also presented.