Analog Spiking-Rate Neuromorphic System for Efficient RF Source Localization

Radiofrequency localization is pivotal for the internet of things ecosystem, requiring solutions that are both precise and energy-efficient. While traditional machine learning and artificial intelligence techniques provide high accuracy in RF localization, they often come at the cost of greater complexity and power usage. To address these challenges, this paper presents a novel RF analog spiking-rate neuromorphic system designed for high-precision, energy-efficient source localization. The system determines source positions within a 360-degree range on a two-dimensional plane, maintaining a high 1-degree angular resolution even in noisy conditions. Through training and testing on both simulated and experimental datasets developed for this study, the system achieves high accuracies, reaching 93.5% at a 20 dB signal-to-noise ratio. It maintains a minimal angle error of 0.1 degrees while consuming just about 403 nW of power.