Spiking Neuron with Sensing Coil Based on a Volatile Memristor

Area of Science:

• Neuromorphic Engineering
• Spiking Neural Networks
• Edge Intelligence Hardware

Background:

• The integration of sensing and processing is crucial for energy-efficient edge intelligence.
• Existing spiking neuron models like the leaky integrate-and-fire (LIF) neuron lack inherent sensory capabilities.
• Developing hardware that directly translates sensory input into neural signals is a key research challenge.

Purpose of the Study:

• To present a novel hardware implementation of a sensory neuron.
• To embed electromagnetic sensing directly into neuronal dynamics.
• To demonstrate a metal-sensitive proximity sensor with spiking output.

Main Methods:

• A novel sensory neuron was designed by coupling a volatile memristor with an LC tank circuit, creating a memristor-resistor-inductor-capacitor (MRLC) neuron.
• The MRLC neuron's design embeds electromagnetic sensing directly into its dynamics.
• The functionality was validated through both circuit simulations and physical experiments.

Main Results:

• The MRLC neuron successfully functions as a metal-sensitive proximity sensor, generating spiking outputs.
• The circuit exhibits diverse dynamical behaviors, including regular spiking, bursting (2-5 spikes/burst), and quasi-chaotic activity.
• The neuron demonstrates sensing memory through hysteresis-like multistability, surpassing basic LIF neuron capabilities.

Conclusions:

• The proposed MRLC neuron represents a significant advancement in energy-efficient spiking edge intelligence by integrating sensing and processing.
• This hardware implementation enables direct transduction of proximity information into spike trains, offering richer dynamics than traditional models.
• The MRLC neuron's capabilities pave the way for more sophisticated and energy-efficient neuromorphic systems at the edge.