Tethered robots get reliable power and communication through their cables, but those cables can snag on obstacles, move them unintentionally, or become damaged. The authors of this paper propose using cable tension itself as a sensing signal to judge whether a contacted obstacle is movable or fixed, which could help robots plan safer paths.
They ran experiments with a tethered mobile robot, varying robot velocity, contact angle, obstacle mass, and obstacle surface material. The key finding was that movable and non-movable obstacles produce different tension patterns: the initial tension rise matters, but the more telling signal is whether tension relaxes after peaking or stays high.
From those patterns, the team selected six features and trained a logistic-regression classifier on 117 labeled trials. It distinguished movable from non-movable obstacles with 95.5% balanced accuracy after 30 cm of post-contact travel. The authors position this as a step toward online, cable-aware autonomous navigation, though they note the approach currently relies on post-contact data rather than prediction before contact.