A Novel Algorithm to Detect and Prevent Slip during Dynamic Object Manipulation using Tactile Feedback

Abstract

Robotic object manipulation has increased exponentially over the last couple of decades across the field of manufacturing, medical robotics, assistive care and warehousing. Detection and prevention of slip of objects plays a vital role in secure object grasping and manipulation. Through the sensory feedback provided by their skin, humans possess the remarkable ability to readily perceive slip. In order to attain a level of skill comparable to humans, robots must be equipped with artificial tactile sensing integrated into their system. In our work, we use an optical tactile sensor, GelSight sensor for real time tactile feedback. We decompose the problem of slip detection as a classification problem by using the input from the GelSight sensor as features. Once slip is detected appropriate corrective actuation is taken to prevent slip. Through this approach we are able to achieve an accuracy of 99% for slip detection for an assortment of objects of different sizes, shapes, textures and rigidity. We then shift gears, and study how slip can be prevented once it is detected. We develop a simplistic and then a more robust algorithm incorporating tactile feedback to detect and prevent slip during dynamic manipulation tasks at different speeds of manipulation. Finally we discuss the drawbacks of the current approach and how these can be addressed in the future work.