基于ADASYN和随机森林的UWB非视距识别方法

NLOS Identification Method for UWB Based on ADASYN and Random F orest

室内UWB(超宽带)定位系统中视距、非视距信号不平衡是常见现象,过分依赖平衡数据集进行NLOS(非视距传播)识别面临着数据采集工作量增加和识别性能降低问题。为解决这一挑战,提出ADASYN和随机森林结合的非视距识别方法。应用ADASYN算法将减少视距、非视距信号数据量之间的差距,可实现样本数量平衡;结合随机搜索与五折交叉验证,实现随机森林模型超参数的自动优化以提升模型的整体效率。将所提出的方法与其他常见的机器学习方法进行比较,实验结果表明,所提出方法在具有最优识别性能的同时有效地缩减了调参时间。

The imbalance beween line-of-sight(LOS)and non-line-of-sight(NLOS)signals is a common issue in indoor UWB positioning systems.Relying too heavily on balanced datasets for NLOS identification leads to increased data collection workload and reduced identification performance.To address this challenge,we propose a NLOS identification method that combines the ADASYN and random forest algorithms.The ADASYN algorithm is used to reduce the disparity between LOS and NLOS signal data,achieving sample balance.Random search and five-fold cross-validation are combined to automatically optimize the hyperparameters of the random forest model,enhancing the overall efficiency of the model.The proposed method is compared with other common machine learning methods,and experimental results show that the proposed method not only achieves optimal identification performance but also significantly reduces parameter tuning time.