Formation Control of Multiple Mobile Robots Systems

This paper presents a vision-based framework for mobile robot detection and tracking using off-the-shelf cameras mounted on mobile robots. Target detection and pose estimation are performed from single frames using fiducial markers as key elements. The method consists in distributing an octagon shaped structure on the back part of each robot. These shapes are easy to extract from the images and posses unique ID codes to facilitate pose estimation. Pose estimation algorithms method about Projection ray attraction (PRA) is programmed and compared. Finally, a dual Unscented Kalman filter (DUKF) is implemented to smooth measured data and estimate unknown leader's velocities.