Towards Visually-Guided Neuromorphic Robots



The Beobot Video Footage

Check out video segments filmed from our beobot prototype here. The AVI files are raw, unprocessed data and the MPEG files have attentional fixations superimposed. Clearly, the attentional system will not help much with basic low-level navigation: For example, it seldom fixates the track (because of its fairly uniform texture, such that nothing in the track itself is very salient). This is why we are developing a basic nagigation system based on a rapid extraction of the "gist" of the scene and its rough layout. The attentional system, however, demonstrates a remarkable ability at finding out obstacles such as pedestrians, and several landmarks which can be used for higher-level orienting and trajectory planning towards a given goal.

Copyright © 2005 by the University of Southern California, iLab and The Beobot Team. Last updated Thursday, 02-Sep-2010 10:05:32 PDT