Communication and Computer Vision Based Control

Computer vision is a promising technology for the control of autonomous robots. Through computer vision, a robot can visually map its surrounding area, recognize specific objects or structures and even determine its position relative to a given landmark. It is preferable to process this video data on a remote client computer rather than taxing the on board server-side processor, freeing the robots computation power to process other sensor information such as sonar readings. However, when large amounts of information must be transmitted to a remote client, even with today's advanced communication technology, there are inherent bandwidth limitations. We consider a network with a transfer rate of 11mbps (mega-bits per second), which is roughly enough to transmit only one 640x480 256 color image per second. Ultimately, we intend to investigate the limitations of video transfer over a network in the context of receiving video data from an autonomous mobile robot. As a demonstration, we will integrate sonar and video information into a simple control algorithm for a Pioneer II mobile robot. Other applications for this technology include: terrain mapping, medical surgery, and even robots that are sent to explore other planets.