Attitude and position control using real-time color tracking

Abstract

A variety of positioning methods and sensors have been developed over the years. Most methods rely on active sensors (such as sonars or lasers) which have range and power restrictions, or rely on computationally complex (and often slow) methods such as recovering positions from stereo or optical flow. This paper describes a system that can determine a robot’s position and orientation, in all six degrees of freedom, relative to a simple passive target. The system can determine its position relative to the target from a single image frame, and process the image and calculate the position faster than a camera’s frame rate (60Hz). It uses a standard, uncalibrated color video camera as its sensor. The system runs on an inexpensive microprocessor (a Motorola 68332) leaving many cycles left over. By careful design of the target and the vision processing hardware and software, we are able to selectively perceive only the relevant parts of the target, greatly simplifying and speeding all the necessary computations. They system is being developed for autonomous spacecraft station keeping and docking.