The problem of image based visual servoing for robots working in a cluttered dynamic environment is addressed in this paper. It is assumed that the environment is observed by depth sensors which allow to measure the distance between any moving obstacle and the robot. Also an eye-in-hand camera is used to extract image features. The main idea is to control suitable image moments and to relax a certain number of robot's degrees of freedom during the interaction phase. If an obstacle approaches the robot, the main visual servoing task is relaxed partially or completely, while the image features are kept in the camera field of view by controlling the image moments. Fuzzy rules are used to set the desired values of the image moments. Beside that, the relaxed redundancy of the robot is exploited to avoid collisions. After removing the risk of collision, the main visual servoing task is resumed. The effectiveness of the algorithm is shown by several case studies on a KUKA LWR 4 robot arm.
Visual servoing with safe interaction using image moments
2015-01-01
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
DDC: | 629 |
Visual servoing with moments of SIFT features
TIBKAT | 2006
|A bayesian deep neural network for safe visual servoing in human–robot interaction
BASE | 2021
|Visual Servoing Using Relevant 2-D Image Cues
British Library Conference Proceedings | 1993
|Weighted moments of SIFT-features for decoupled visual servoing in 6DOF
BASE | 2007
|