(via USC)
Researchers from the Viterbi School of Engineering at USC have released a paper detailing their development of a robotic hand that is capable of surpassing the human’s ability in identifying objects based on textures. The specialized robotic hand is capable of reading textures due to the unique sensors built into each finger combined with an algorithm that is designed to mimic hand-movements made by humans when identifying objects.
The secret to object identification for the robotic hand lies in its specially designed fingertips which is made by Syntouch. The sensors, called Bio Tac Biomimetic Sensors, developed by the paper's co-author Jeremy Fishel, uses a rigid inner-core that is encased by an elastic skin filled with liquid. The skin of fingertip is outfitted with fingerprints which ‘sense’ vibrations when moved over a textured surface much like its human counterparts. The vibrations are detected by a hydrophone located in the fingers rigid core which helps in identifying objects. In conjunction with the hydrophone are pressure sensitive electrodes positioned throughout the fingers tip to determine the shape of an object based on force and pressure. The finger is even able to detect heat by using a thermistor located in the fingertip to identify thermal gradients.
The researchers tested the robotic hand using 117 objects ranging from hardware, fabric and stationary stores where it was successful at correctly identifying 95% of the items using 5 exploratory movements per object. The researchers state that the robotic hand could be used to develop improved prosthetics for hand amputees as well as implementing them into industrial and military applications. (Military - Terminator comes to mind).
Cabe
