Enhancing Pedestrian Dead Reckoning Systems for Accurate Location Tracking in GNSS-Deprived and Challenging Environments Using Smartphone IMU Sensors
Date Issued
May 2025
Author(s)
Advisor
Abstract
Pedestrian dead reckoning (PDR) uses smartphone inertial sensors (accelerometer, gyroscope, magnetometer)
to estimate displacement by detecting steps and heading changes. However, IMU-based PDR
suffers from drift due to sensor biases and magnetic disturbances. This thesis proposes an enhanced PDR
framework using real SensorLog data for indoor environments. We implement robust step detection and
heading estimation, analyzing metrics such as step cadence, walking speed, path straightness, stop time,
and angular velocity to characterize pedestrian motion. Trajectories are visualized via a Python tool that
plots classic and color-coded paths. We fuse WiFi fingerprinting (RSSI) with the PDR estimates using a
Kalman filter to mitigate drift. This hybrid approach follows prior work, where quaternion-based EKF
reduced PDR error to 2.2 m over a 270 m walk, and advanced heading fusion achieved 80.
to estimate displacement by detecting steps and heading changes. However, IMU-based PDR
suffers from drift due to sensor biases and magnetic disturbances. This thesis proposes an enhanced PDR
framework using real SensorLog data for indoor environments. We implement robust step detection and
heading estimation, analyzing metrics such as step cadence, walking speed, path straightness, stop time,
and angular velocity to characterize pedestrian motion. Trajectories are visualized via a Python tool that
plots classic and color-coded paths. We fuse WiFi fingerprinting (RSSI) with the PDR estimates using a
Kalman filter to mitigate drift. This hybrid approach follows prior work, where quaternion-based EKF
reduced PDR error to 2.2 m over a 270 m walk, and advanced heading fusion achieved 80.
File(s)![Thumbnail Image]()
Name
AChristoforou-BSC-abstract.pdf
Size
536.06 KB
Format
Adobe PDF
Checksum (MD5)
a6c9d0c8826eba9f8052739a10a4b31f