A Lightweight α-μ Fading Environment-Based Localization Toward Edge Implementation

This letter presents a lightweight localization method utilizing Received Signal Strength (RSS) of Wireless Fidelity (Wi-Fi) signals considering a generic $\alpha -\mu $ fading environment. Despite the usefulness of RSS-based methods, inaccuracies arise from signal randomness caused by shadowing and small-scale fading effects. Furthermore, fingerprinting methods are computationally exhaustive due to the multiple RSS sample collection. We consider a lightweight range-based approach, modeling small-scale fading as $\alpha -\mu $ distribution, and present the formulation of a Maximum Likelihood Estimation (MLE) estimator for deriving location estimates using a single RSS sample. Moreover, to address the issue of exploding gradients resulting from divergent terms, a clipped gradient ascent variant is proposed. The proposed estimator outperforms other estimators based on established Rayleigh, and Nakagami-m fading distributions in comprehensive tests conducted in indoor and outdoor environments. A major advantage of the proposed method is its lower computational complexity over existing methods, and hence it can be easily deployed on edge devices.