Implementation and Validation of Bioplausible Visual Servoing Control.

摘要

In this report, we compare the performance of the Lucas-Kanade algorithm with feature tracking with the bioplausible optical flow algorithm to achieve pose stabilization in the context of one-dimensional (1-D) attitude stabilization. These results have been benchmarked against an ideal controller in both simulation and robotic experimentation. Within both environments, the accuracy, stability, and settling time were evaluated as a function of contrast and system noise. Both algorithms achieved reasonable performance when compared to the ideal controller, but in most cases the Lucas-Kanade algorithm outperformed the bioplausible algorithm. Within the evaluated serial implementation, the Lucas-Kanade algorithm was also faster. It is anticipated that both performance and processing speed will improve for the bioplausible algorithm when it is implemented in a parallel instantiation.