A PDR-based indoor positioning system using smartphone-embedded IMU sensors

This thesis proposes a novel pedestrian dead reckoning (PDR)-based method to improve the accuracy of the indoor positioning system by using a smartphone’s inertial measurement unit (IMU). We focus on the investigation of step detection, and step length estimation. In the step detection phase, a novel method is proposed to select the distinct signal for step detection corresponding to the recognized smartphone handheld modes. In the step length estimation phase, the optimized Polyak-Ruppert weight averaging ensemble method is employed to fuse the result of the step length estimation sub-models from multiple recognized smartphone handheld modes. The experiments are conducted in order to evaluate the performance of indoor positioning estimation by combining the result of the proposed method from step detection and step length estimation with the tilt-compensation approach in heading estimation phase. Thus, the result shows a lower positioning error compared with the conventional methods.

本论文提出了一种新颖的基于行人航位推算（Pedestrian Dead Reckoning, PDR）的方法，通过利用智能手机的惯性测量单元（Inertial Measurement Unit, IMU）提升室内定位系统的定位精度。本研究聚焦于步检测与步长估计的相关探究。在步检测阶段，本文提出了一种针对已识别的智能手机握持模式，选取适配步检测的差异化信号的新颖方法。在步长估计阶段，本文采用优化后的Polyak-Ruppert权重平均集成方法，对来自多种已识别智能手机握持模式的步长估计子模型结果进行融合。为评估所提方法的室内定位估计性能，本文将步检测与步长估计的结果，结合航向估计阶段的倾斜补偿方法进行融合后开展实验。实验结果表明，相较于传统方法，本方法的定位误差更低。