Reproduce “Robust Localization of Key Fob Using Channel Impulse Response of Ultra Wide Band Sensors for Keyless Entry Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Abhiram Kolli, Filippo Casamassima, Horst Possegger, Horst Bischof
Researchers at Graz University of Technology and NXP Semiconductors developed a robust multi-head self-supervised neural network for key fob localization in keyless entry systems using Ultra-Wideband (UWB) Channel Impulse Response (CIR) features. This architecture demonstrated improved adversarial robustness against common attacks, achieving up to a 67% higher accuracy under FGSM attacks compared to standard neural networks, while maintaining a compact model size of 82.0 kB.
Key insights
- Carefully selected pre-computed features from UWB CIR data are highly effective for key fob localization, outperforming models that use raw CIRs.
- The proposed RBF neuron dynamics with a multi-head self-supervised denoising architecture can inherently enhance adversarial robustness without explicit adversarial training.
- Training a denoising head with a custom loss that includes adversarial examples helps the network learn more robust and discriminative features, improving resistance to various attacks.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.