The article discusses a system that uses tactile information to estimate the rotation angle of a grasped object when slippage occurs during robotic manipulation tasks. The system uses a neural network to segment the contact region and an algorithm to estimate the rotated angle of that region. This method is applied to DIGIT tactile sensors and has been trained and tested with a publicly available dataset. The system showed a 95% and 90% result in Dice and IoU metrics in the worst-case scenario. The approach is capable of detecting slippage movement, providing a possible reaction to prevent the object from falling.

 

Publication date: 19 Jan 2024
Project Page: ?
Paper: https://arxiv.org/pdf/2401.09831