BiGS: BioTac Grasp Stability Dataset

Autonomous grasping of unknown objects is a fundamental requirement for robots performing manipulation tasks in real world environments. Even though there has been a lot of progress in the area of grasping, it is still considered an open challenge and even the state-of-the-art grasping methods may result in failures [1]. A reliable prediction of grasp stability helps to avoid such failures and provides an option to re-grasp the object safely. Since the majority of grasping failures happen at the contact points, which are occluded for vision systems, tactile feedback plays a major role for predicting grasp stability. The human-inspired biomimmetic tactile sensor (BioTac) [2] is equipped with a 19-electrode array and a hydroacoustic sensor surrounded by silicon skin inflated with incompressible and conductive liquid. This design provides rich tactile feedback similar to the slowly-adapting and fastadapting afferents present in the human skin [3]. Latest developments in classification algorithms [4] allow us to explore the potential of large amounts of data from these sensors. Our goal is to provide a publicly accessible grasp-stability dataset collected using the BioTacs and, thus, enable further development of algorithms capable of reliable grasp stability prediction.