SoRa: Waveform Features Based Soft Range Information Estimation Enabling Robust Tracking in Cluttered Environments

Soft range information (SRI) is promising for high-accuracy localization in harsh environments, because it gives all distance possibilities in combination with channel characteristics. In this paper, we propose a probabilistic tracking method based on SRI of waveform features, namely SoRa. Specifically, in order to estimate SRI efficiently, we weigh the correlation, physical meaning, and computational complexity to choose the best feature set among physical and raw features. Then the SRI is introduced into the probabilistic observation model of the particle filter. According to experimental validation, the proposed SoRa achieves the localization mean errors of less than 34 cm and the 90 -th percentile errors of less than 60 cm. Moreover, it is robust to scene changes, and the generalizability outperforms the conventional machine learning-based localization models.