Spiking Neural Networks based Rate-Coded Logic Gates for Automotive Applications in BiCMOS

Abstract

Spiking Neural Networks (SNNs) represent the third generation of artificial neural networks. In this work, we evaluate the core element of SNN, the neuron circuit equivalent, in terms of temperature robustness for automotive applications. Thanks to the operating point stabilization, the proposed circuit-level neuron implementation achieves a broad frequency tuning range up to 42 MHz and operates over a wide temperature range from −40 °C to 125 °C. At the maximum spiking frequency of 42 MHz, the circuit consumes a DC power of only 300 nW. We use the proposed neuron circuit to realize two fundamental logic gates, AND and OR, by means of analog rate-encoded spiking neural networks. To the best of the authors’ knowledge, these are the first reported SNN-based logic gates measured over the automotive temperature range. We showcase the suitability of SNN circuit implementation for automotive applications. The circuits are realized in a 130 nm BiCMOS.