Predicting slip of objects that have been picked up
Robots using classical control have proven less effective when the manipulated objects are deformable due to their unpredictable behaviour. In this work at LIS, EPFL, we used a tactile sensor to continuously monitor forces when a deformable object (fabric) was grasped. We trained a model to distinguish between successful grasps, slippage and failure during a manipulation task. Slippage could be anticipated before failure occurred using data acquired over a 30 ms period with a greater than 95% accuracy.
This work was presented at the IEEE International Conference on Intelligent Robots and Systems (IROS) 2022. Access the paper on the publisher’s website.