Visual control through narrow passages for an omnidirectional wheeled robot

Robotic systems are gradually replacing human intervention in dangerous facilities to improve human safety and prevent risky situations. In this domain, our work addresses the problem of autonomous crossing narrow passages in a semi-structured (i.e., partially-known) environment. In particular, we focus on the CERN’s Super Proton Synchrotron particle accelerator, where a mobile robot platform is equipped with a lightweight arm to perform measurements, inspection, and maintenance operations. The proposed approach leverages an image-based visual servoing strategy that exploits computer vision to detect and track known geometries defining narrow passage gates. The effectiveness of the proposed approach has been demonstrated in a realistic mock-up.