Leveraging Fine-Grained Self-correlation in Detecting Collided LoRa Transmissions.

Recently LoRa has become one of the most attractive Low Power Wide Area Network (LPWAN) technologies and is widely applied in many distinct scenarios, such as health monitoring and smart factories. However, affected by the signal collision of uplink transmissions, a base station fails to decode concurrent transmissions. To solve this challenge and improve the performance of the base station, it is necessary to detect the collided transmission accurately. Existing researches focus on extracting the corresponding payload from collided signals based on the information of detected preamble. In this paper, we present FSD, a novel approach that achieves an effective preamble detection from collided LoRa packets, which exploits the inherent fine-grained similarity of LoRa. We implement and evaluate the design on commodity LoRa devices and USRP B210 base stations. The experiment results show that the accuracy and precision are improved by up to 20% than the continue peaks detection method and time-domain cross-correlation method.