An improved pedestrian dead reckoning algorithm based on smartphone built-in MEMS sensors

Abstract

With the advantages of low cost, small size and lightweight, the smartphone has played a more and more important role in pedestrian indoor positioning. In particular, the application of Pedestrian Dead Reckoning (PDR) in the smartphone built-in MEMS sensors makes the application of smartphone more extensive. However, the zero-bias instability of the smartphone built-in MEMS sensors leads to the rapid accumulation of pedestrian trajectory calculation error. To solve this problem, we used the bias drift model and Kalman filter (KF) to denoise the original MEMS data. The dual-feature step detection model of peak domain and time domain was established to provide accurate step information for step length estimation and heading correction. Based on the Weinberg model, the three-steps constraint step length estimation (TCSLE) model was proposed to estimate step length accurately. Then, based on the improved heuristic drift elimination (iHDE), the adaptive drift elimination (ADE) model was proposed to identify different walking states. The correction models under different walking states were established to correct the heading angle accurately. Finally, the pedestrian trajectory was reconstructed using accurate step length and heading information. To verify the performance of the PDR algorithm based on the above model, three experimenters with different heights and genders were recruited, and three mobile phones with different sensor performance were selected. The experimenters moved smoothly and steadily with hand-held mobile phone, and 18 sets of experiments were carried out along two paths. The experiment results shown that the step length deviation was less than 1.4871 %, the horizontal positioning error was less than 1.6070 m, and the relative positioning error was less than 1.1816 %D. The proposed PDR algorithm has strong adaptability and robustness, and meets the needs of pedestrian indoor positioning.