Error-Tolerant Cyclic Sequences for Vision-Based Absolute Encoders

A method for obtaining error-tolerant cyclic sequences similar to de Bruijn sequences is presented. These sequences have a number of potential applications, including use as absolute rotary encoders. This investigation is motivated by the desire to use a vision-based system to obtain the angular position of the wheels of mobile robots as they rotate about their axes. One benefit of this approach is that the actual wheel orientation is observed (as opposed to non-collocated measurements of wheel angles via encoders on the motor shaft). As a result, ambiguities from backlash are eliminated. Another benefit of this system is the ability to apply it quickly to existing systems. Several methods are developed for increasing the robustness of these encoders. An imaging simulator is used to compare the accuracy of a variety of encoding schemes subjected to several levels of image noise.