Neuromorphic Sensors with Visible Light Communications

Abstract

Spiking neural networks (SNN) can control single-joint robotic arms' precise rotation and force when shape memory alloy (SMA) actuators are used. SNN receives feedback from neuromorphic sensors for controlling anthropomorphic fingers, which typically respond to the flexion angle and the force applied to the fingertips. The robotic fingers and hands are in relative motion with the robot's body, which typically includes the main unit for the limb's motion control. An elegant method to implement the connection between the main control unit and the neuromorphic sensors on limbs is to use visible light communication technology. This work presents a system that uses the recently introduced optical axons to connect the neuromorphic force sensors placed on fingers and the main electronic SNN that controls the finger's motion. The SNN behaviour is evaluated comparatively with and without optical axons when the hand moves relatively to the neural control unit (NCU), which regulates the finger's force. The expected results show that despite small oscillations of the finger force during steady-state, the robotic hand can hold on to an object while moving in the vicinity of NCU.