Evaluation of attentional control in active systems using a 3D simulation framework

Abstract

In active vision systems, attentional control is used to determine the relevant parts of a scene and to direct perception towards these parts. To test and evaluate active vision systems, we have implemented a 3D simulation framework capable of simulating a broad scope of environments from simple block worlds to complex photorealistic scenes. The simulator allows full control of all aspects of the simulation, acting and moving inside virtual environments. In this paper, we demonstrate its use for evaluating our attentional control system. The attention model is based on a novel two-stage selection mechanism and especially focuses on the dynamic and three-dimensional aspects of its environment.