基于位置信息指纹的蓝牙/WiFi混合定位方法

Hybrid positioning method using bluetooth and WiFi based on location information fingerprint

关键词:  蓝牙  WiFi  混合指纹  位置信息指纹  蓝牙/WiFi混合定位

Key words:  Bluetooth  WiFi  Hybrid fingerprint  Location information fingerprint  Bluetooth/WiFi hybrid positioning

摘要:

针对蓝牙/无线保真（WiFi）混合定位精度不理想、稳定性差等问题，提出了基于位置信息指纹的蓝牙/WiFi混合定位方法，该方法由离线阶段与在线阶段组成。在离线阶段，首先将采集的蓝牙/WiFi信号强度分为2组，第1组用于构建蓝牙、WiFi和蓝牙/WiFi混合指纹库；第2组作为训练指纹，分别与蓝牙、WiFi及蓝牙/WiFi混合指纹库匹配定位，以获得蓝牙、WiFi及蓝牙/WiFi混合指纹估计位置。随后，基于指纹估计位置和参考点构建位置信息指纹库。在在线阶段，先进行蓝牙、WiFi和蓝牙/WiFi混合指纹定位，然后结合蓝牙、WiFi和混合指纹估计位置生成在线位置信息指纹，最后，利用K近邻（KNN）算法实现与位置信息指纹库的匹配定位。实验结果表明，提出方法在2个公开数据集上的定位效果明显优于加权K近邻（WKNN）、高斯过程回归（GPR）和支持向量机（SVM）方法。在数据集一中，提出方法的均方根误差（RMSE）比WKNN、GPR和SVM最少减小了41.21%、48.33%和67.56%；在数据集二中，提出方法的平均绝对误差（MAE）为 0.914 m，远优于WKNN、GPR和SVM。

Abstract:

To address the issues of poor accuracy and stability in Bluetooth/WiFi hybrid positioning, a hybrid positioning method using Bluetooth and WiFi based on location information fingerprint is proposed, which includes an offline phase and an online phase. In the offline phase, Bluetooth and WiFi signal strength data are collected and segmented into two groups. The first group is used to construct Bluetooth, WiFi, and Bluetooth/WiFi hybrid fingerprint databases, while the second group is used to train fingerprints, which are subsequently matched with the Bluetooth, WiFi, and hybrid fingerprint databases to estimate positions for Bluetooth, WiFi, and Bluetooth/WiFi. A location information fingerprint database is then constructed based on these estimated positions and reference points. In the online phase, Bluetooth, WiFi, and Bluetooth/WiFi hybrid fingerprint positioning is performed. The estimated positions of Bluetooth, WiFi, and hybrid fingerprints are combined to generate online location information fingerprints, which are then matched with the location information fingerprint database using the K-nearest neighbors (KNN) algorithm. Experimental results show that the proposed method significantly outperforms weighted K-nearest neighbors (WKNN), Gaussian process regression (GPR), and support vector machine (SVM) methods in terms of positioning accuracy on two public datasets. In Dataset 1, the root mean square error (RMSE) of the proposed method decreased by at least 41.21%, 48.33%, and 67.56% compared to WKNN, GPR, and SVM, respectively. In Dataset 2, the mean absolute error (MAE) of the proposed method was 0.914 meters, significantly outperforming WKNN, GPR, and SVM.