Comparison of Multi-Channel Ranging Algorithms for Narrowband LPWA Network Localization

Florian Wolf (1) , Mohamed Sana , Sébastien de Rivaz , Francois Dehmas (2) , Jean-Pierre Cances (1)

1 XLIM-SRI - Systèmes et Réseaux Intelligents
2 CEA-LETI - Commissariat à l'énergie atomique et aux énergies alternatives - Laboratoire d'Electronique et de Technologie de l'Information

Accurate radio signal based localization for Low Power Wide Area networks enables ubiquitous positioning for the Internet of Things. Narrowband communication and multipath propagation make precise localization challenging. Coherent multi-channel ranging increases bandwidth and provides improved temporal resolution through the aggregation of sequentially transmitted narrowband signals. This paper applies parametric estimators as well as a deep learning technique to multi-channel measurements obtained with 10 kHz signals. Ranging performances are compared via numerical simulations and real outdoor field trials, where parametric estimation and deep learning achieve 60 m and 45 m accuracy in 90% of the cases, respectively. Further work is required to study the impact of deep neural network training with a combination of synthetic and real data. Future research may also include the adaptation of multi-channel localization to differential network topologies.

Mots clés: deep learning; DNN; LPWA network localization; range estimation; frequency hopping; MUSIC

Domaines: Electronique; Traitement du signal et de l'image [eess.SP]

Florian Wolf, Mohamed Sana, Sébastien de Rivaz, Francois Dehmas, Jean-Pierre Cances. Comparison of Multi-Channel Ranging Algorithms for Narrowband LPWA Network Localization. 2019 The 5th International Symposium on Ubiquitous Networking (UNet), Nov 2019, Limoges, France. ⟨hal-02515127⟩