World simulation as a tool for performance evaluation of algorithms in robotics

In this work, we present a simulation tool for representing virtual robotic worlds for performance evaluation of algorithms used in robotic tasks. We take advantage of the controlled scenario for obtaining ground truth data for the algorithms under test. This reference data is not available when the real robot is performing its tasks. Our system provides virtual perceptual data to the algorithms under test and responds eventually to commands issued by the algorithms. In this paper, we show the application of the world simulator to a visual tracking task.