Instead of relying on time-divisional multiple access to reconstruct a two-dimensional map of pressure / temperature distribution, the researchers chose an event-based signal architecture, where sensors asynchronously transmit data upon value changes. This is similar to the skin’s biological mechanoreceptors, their source of inspiration, which can fire spike signals asynchronously and form very precise spatiotemporal patterns that reach the brain at a constant latency for sensory interpretation.
Under the title “A neuro-inspired artificial peripheral nervous system for scalable electronic skins“, they disclosed in the Science Robotics journal a neuromimetic architecture which they call Asynchronously Coded Electronic Skin (ACES), theoretically capable of simultaneously transmitting the thermotactile information of 10,000 sensors, at a constant readout latency of 1ms.
In the ACES platform, each sensor or “ACES receptor” combines a resistive sensor, a microcontroller, and several passive components to perform the necessary signal conditioning. The receptor can then capture and transmit stimuli information asynchronously as “events” using electrical pulses spaced in time. Because each ACES receptor has its own pulse signature (designed to be transmitted in 1ms), multiple sensors can transmit concurrently and without specific time synchronization on the same conductor trace which will propagate the combined pulse signatures to a decoders. At the receiving end of a common electrical path, the decoder leverages a neural network to identify all the transmitting receptors by correlating the received pulses against the known temporal arrangement of pulses for each receptor’s signature. Prior 2D mapping of the receptors’ spatial distribution allows the decoder to match each stimuli to a physical location on the e-skin.