RadarPDR: Radar-Assisted Indoor Pedestrian Dead Reckoning

Area of Science:

• Robotics and Navigation
• Sensor Fusion
• Signal Processing

Background:

• Pedestrian dead reckoning (PDR) is crucial for indoor navigation but suffers from accuracy issues due to smartphone sensor errors.
• Existing PDR methods relying solely on inertial sensors accumulate significant tracking errors.

Purpose of the Study:

• To propose a novel radar-assisted PDR (RadarPDR) scheme to improve indoor pedestrian tracking accuracy.
• To address the limitations of current inertial sensor-based PDR systems.

Main Methods:

• Integration of a frequency-modulation continuous-wave (FMCW) radar with smartphone inertial sensors.
• Development of a segmented wall distance calibration model to mitigate radar ranging noise.
• Fusion of radar-derived wall distance with inertial sensor data (acceleration, azimuth).
• Implementation of a hierarchical particle filter (PF) combined with an extended Kalman filter for position and trajectory adjustment.

Main Results:

• RadarPDR demonstrates efficient and stable performance in practical indoor scenarios.
• The proposed scheme significantly outperforms traditional inertial sensor-based PDR methods.
• Improved accuracy in walking direction, step detection, and step length estimation.

Conclusions:

• RadarPDR offers a robust solution for accurate indoor pedestrian tracking and navigation.
• The fusion of FMCW radar with inertial sensors effectively compensates for individual sensor limitations.
• This approach provides a foundation for more reliable indoor navigation services.