基于WiFi指纹的高精度室内定位融合算法

High-precision indoor localization fusion algorithm based on WiFi fingerprint

针对室内环境中指纹定位接收信号强度信息的高维时变特性以及动态目标定位误差的累积问题,提出了一种基于梯度提升决策树与粒子滤波相结合的融合算法。该算法首先利用梯度提升决策树算法建立的位置坐标与接收信号强度之间的非线性映射模型,对在线接收的信号强度数据进行特征分类判别,实现位置的初步估计;随着目标的运动,进一步结合粒子滤波方法,迭代地实现动态目标位置的精确预测;另外,将定位轨迹与实际轨迹进行对比,以验证该算法的稳定性。实验仿真结果表明:累积分布函数在80%的百分位处,提出算法的定位精度控制在1.19 m以内,明显优于基于支持向量机、随机森林等定位算法;同时较基于梯度提升决策树算法的定位精度提升了34.9%;所获得的定位轨迹与实际轨迹的趋势一致且趋于收敛。

Aiming at the high-dimensional time-varying characteristics of received signal strength information for fingerprint positioning and the accumulation of dynamic target positioning errors in the indoor environment,a fusion algorithm is proposed based on gradient boosting decision tree and particle filter.Firstly,the initial estimation of the position was achieved,using the nonlinear mapping model between the position coordinates and the received signal strength established by the gradient boosting decision tree algorithm to classify and discriminate the received signal strength data;then,with the movement of the target,the accurate prediction of the dynamic target position was iteratively realized in combination with the particle filtering method;in addition,a comparison of the positioning trajectory with the actual trajectory was made to verify the stability of the algorithm.The experimental results show that 80%of the proposed algorithms’positioning accuracy is controlled within 1.19 m,which is significantly better than the support vector machine and the random forest based positioning algorithm,and has increased by 34.9%in performance over the gradient boosting decision tree algorithms’positioning accuracy;the positioning trajectory also coincides with the trend of the actual trajectory and gradually converges.