tactile-sensing

Distributed sensor arrays capable of detecting multiple spatially distributed stimuli are considered an important element in the …

The paper proposes a method using motion and tactile features to automatically segment atomic actions from teleoperated demonstrations for complex robotic tasks, provides a dataset of pick-and-place teleoperation with a dexterous hand, and shows the proposed features generalize between episodes and similar objects while tactile sensing improves activity recognition within demonstrations.

The paper presents a pipeline to predict safe robotic grasps of unknown objects using depth and tactile sensing. Three methods are compared to find optimal grasp points during tactile exploration - grid search, standard Bayesian Optimization, and Unscented Bayesian Optimization. Results show Unscented Bayesian Optimization provides higher confidence in discovering safe grasps with fewer exploratory observations. It converges faster to robust grasp points away from edges compared to other methods.

The paper presents an approach using depth sensing and tactile feedback to predict safe robotic grasps of completely unknown objects through haptic exploration and comparing grid search, standard Bayesian Optimization, and unscented Bayesian Optimization, with results showing unscented Bayesian Optimization provides higher confidence grasps with fewer exploratory observations.