The surge in the number of non-geostationary satellites has intensified interference issues. Interference detection in satellite communication with onboard processing is critical, as it relies heavily on accurately identifying interference events to ensure seamless and reliable data transmission. This paper delves into the critical problem of onboard interference detection and identification within satellite communication systems. A scenario comprising a satellite that receives radio frequency (RF) interference from a ground terminal not belonging to its network is considered. Conventional techniques, such as energy detection (ED), have shown difficulties in detecting weak interferences. In this context, artificial intelligence (AI)-based techniques are a promising solution to enhance performance when addressing RF interference. In particular, the case of brain-inspired spiking neural networks (SNNs) for energy-efficient implementations is studied. The primary contributions of this work encompass the creation of a labeled dataset for interference detection in noisy channel environments, the implementation of both SNN and convolutional neural network (CNN) models for interference detection and classification, and a thorough evaluation of their performance for onboard interference detection and identification in actual commercially available AI-acceleration chipset of Xilinx and Intel's research neuromorphic processor, Loihi 2. The results confirm that SNNs implemented on neuromorphic chips can offer orders of magnitude improvements in energy efficiency compared to CNNs implemented on specialist hardware while maintaining accuracy.
No takes yet. Share an insight, caveat, or question.
Eappen et al. (2025) studied this question.