2016/nts16

Neuromorphic Tactile Sensing

Members: James Wright, Michael Pfeiffer, Wang Wei Lee

Organizers:: James Wright (MARCS Institute, Western Sydney University) Jonathan Tapson (MARCS Institute, Western Sydney University) Wang Wei Lee (Singapore Institute for Neurotechnology) 'Nitish V. Thakor' (Singapore Institute for Neurotechnology)


Focus and goals of this topic area

This workshop aims to provide tutorials on microneurography, an introduction to the neurophysiology of the peripheral nervous system, including the somatosensory system and mechanotransduction, and an examination of the state of the art in peripheral nervous system interfaces for neuroprosthetic control. It also provides an opportunity for attendees to engage with a multi-disciplinary team of experts who are interested in the neuromorphic approach to tactile processing.

Our tutorials will also provide an introduction to biological tactile sensing, followed by artificial tactile sensors and models for spike conversion. Techniques for spike‐based machine learning will also be discussed. The attendees will work on projects of varying levels of difficulty, depending on their prior knowledge.

Newcomers to the field will get to work on fabrication of tactile sensor arrays, and implementing biofidelic models for spike conversion in embedded systems. Advanced attendees will work on implementing learning algorithms for recognition of shape, texture and slippage. Other advanced projects involve fusion of multiple sensor arrays, and adaptive algorithms to suit changing sensor behaviors. Attendees may also pursue real‐time implementations of these algorithms on neuromorphic hardware such as the SpiNNaker or TrueNorth. Attendees can also look forward to a mini‐competition to implement the most reliable event‐based slip detection algorithm.

The MARCS Institute will provide a microneurography recording system, including electrodes, amplifiers, stimulus isolators and trained operators. We will also provide BioADC systems for the acquisition of surface EMG signals. We will provide a data set of single and multiunit recordings from the afferent and efferent peripheral nervous system. We will provide as well some simple robotic effectors. We also plan to extend the opportunity to commercial exoskeleton manufacturers to attend with example development units.

Attendees will be able to implement surface EMG based control of simple robotic end effectors, as well as using recordings of spike trains from the efferent peripheral nervous system for offline control. Advanced attendees will be able to work on decoding live spike trains acquired from the somatosensory system of volunteers with real time algorithms, on neuromorphic hardware as part of a closed loop neuroprosthetic. Depending on availability attendees may also be able to incorporate these algorithms into commercially available exoskeletons.

We will provide tactile signal acquisition systems, which include sensor arrays and electronics, as well as materials to fabricate more of them. Skeleton code for signal acquisition will be provided, as well as existing codes to train spike based learning algorithms. We will also provide a pair of prosthetic hands and existing software for integration with the tactile sensors.



Invited Participants of the Topic Area

Name Institution Expertise Time Website
'Vaughan Macefield' Western Sydney University Neurophysiology 26 June - 01 July  http://www.westernsydney.edu.au/marcs/our_team/researchers/professor_vaughan_macefield
'Sliman Bensmaia' University of Chicago Neurophysiology 11 July - 16 July  http://bensmaialab.org/lab-members/
'Masaki Sekino' University of Tokyo Bio-Harmonized Electronics 13 July - 16 July  http://www.bee.t.u-tokyo.ac.jp/index.html
Calogero Maria Oddo Scuola Superiore Sant’Anna Biorobotics 26 June - 30 June  http://sssa.bioroboticsinstitute.it/user/102
Mathew Schiefer Case Western Reserve University Neural Engineering  http://fescenter.org/about-fes-center/who-we-are/investigators/schiefer-matthew-phd/


Projects

This is a list of potential projects that students can work on during the workshop:

  • Fabrication of tactile sensor arrays
  • Implementing biofidelic models for spike conversion in embedded systems
  • Implementing learning algorithms for recognition of shape, texture and slippage
  • Real-time implementations of these algorithms on neuromorphic hardware such as the SpiNNaker or TrueNorth
  • Surface EMG control of robotic effectors
  • Decoding of spike trains recorded from volunteers, as well as recordings brought by the organisers

Recommended Reading and Resources

  • Papers and tutorials that should be studied before Telluride.


Available Hardware and Equipment

  • We will bring microneurographic recording equipment suitable for recording spikes from the human PNS
  • A Tactile signal acquisition systems, and components to fabricate additional systems
  • A WidowX robotic gripper with custom-fitted tactile sensor arrays
  • Multi-channel surface EMG acquisition system

Results

To be added during the workshop


Lectures, Tutorials, and Slides

  • To be added during the workshop

Attachments