A Visual Attention Framework for Search Behavior by a Humanoid Robot

This paper presents a framework for a visual search behavior of a 3D object in a 3D environment performed by HRP-2 humanoid robot. Object search is a sensor planning problem in which appropriate sensor configuration must be selected in order to allow a proper recognition. The sensor planning is formulated as an optimization problem in which the goal is to maximize the target detection probability while minimizing the energy/distance and time to achieve the task. The knowledge of the target position is encoded in a probability distribution map which will be updated after each sensing operation. Specificities of the humanoid robot as well as the characteristics of the recognition system are taken into account to restrict the sensor configuration space to meet heavy time constraints on the behavior in order to have a reactive and near human like reaction time