低功耗广域（LPWA）无线定位具有长时效、覆盖范围大和穿透性强的优势，成为近年特定物联网关注的重点。然而，LPWA信号观测包含时变多径噪声、通信频次低和基站部署稀疏严重影响定位精度。本项目研究内容：首先，研究复杂噪声分布参数化表示和定位信号稀疏表示方法，构建噪声鲁棒与信号稀疏的NR-SS模型，为LPWA无线定位抑制噪声影响提供理论基础；其次，提出无线指纹地图的鲁棒增量更新算法和多观测时空在线定位算法，保障地图更新的质量与效率，提升鲁棒定位精度；最后，选择矿井作业和地下管廊救援多人员定位场景，验证所提模型的正确性与算法的有效性。通过本项目的研究，形成自主抑制复杂噪声的鲁棒稀疏学习方法，为多径环境下LPWA稳健定位提供理论支撑，为遮挡严重场景中的常态化定位监测和紧急救援提供新思路。

Low power wide area (LPWA) wireless localization has the advantages of long life, wide coverage and strong penetration, which has become the focus of specific Internet of Things in recent years. However, the localization accuracy is seriously affected due to time-varying multipath noise in the observations, low communication frequency and sparse deployment of base stations. The research contents of this project are as follows. Firstly, by studying the parametric representation of complex noise and sparse representation of localization signal, we present a noise-robust and signal-sparse (NR-SS) model for localization noise suppression. Secondly, a robust incremental algorithm for wireless fingerprint map updating and a multi-observation spatio-temporal algorithm for online localization are proposed to improve the localization efficiency and accuracy. Finally, the effectiveness of the proposed model and algorithms is verified in the multi-person localization scenarios of mine operation and underground pipeline rescue. This project will develop a robust sparse learning approach for LPWA localization with multipath interference, which provides a new idea for normal localization monitoring and emergency rescue in serious occlusion scenes.