Abstract
Tactile feedback is a key sensory channel that contributes to our ability to perform precise manipulations. Sensor skin provides robots with the sense of touch making them increasingly capable of, e.g., dexterous object manipulation. However, in applications like teleoperation, the complex sensory input of an infinite number of different textures must be projected to the human users skin in a meaningful manner. In addressing this issue, we deployed a deep gated recurrent unit-based autoencoder (GRU-AE) that captured the perceptual dimensions of tactile textures in latent space, thus, allowed the network to implicitly understand unseen textures. Furthermore, the expression of unknown textures in latent space allowed the definition of a control law to efficiently drive tactile displays, thus, allowed to convey tactile feedback in a psycho-physically meaningful manner. We experimentally verified our approach by evaluating the prediction performance of the GRU-AE on seen and unseen data that was gathered during active tactile exploration of objects of daily living. Then, we conducted a user study with a custom-made tactile display in which real tactile perceptions in response to active tactile object exploration were compared to the emulated tactile feedback using our proposed tactile feedback loop. The results suggest that our deep GRU-AE for tactile display control offers an efficient and intuitive method for efficient end-to-end tactile feedback during active tactile texture exploration for the implementation into real-time capable tactile feedback systems.
Funding Information
  • JSPS (19H01130, 19K14948)
  • Tateishi Science and Technology Foundation
  • Research Institute for Science and Engineering of Waseda University

This publication has 27 references indexed in Scilit: