Accurate Stride-Length Estimation Based on LT-StrideNet for Pedestrian Dead Reckoning Using a Shank-Mounted Sensor

Area of Science:

• Robotics and Artificial Intelligence
• Sensor Fusion and Navigation

Background:

• Pedestrian dead reckoning (PDR) is a crucial self-contained positioning technology.
• Accurate pedestrian stride length estimation is vital for PDR system performance.
• Existing methods struggle with varying pedestrian walking speeds, increasing PDR errors.

Purpose of the Study:

• To propose a novel deep learning model, LT-StrideNet, for enhanced pedestrian stride length estimation.
• To develop a shank-mounted PDR framework utilizing the LT-StrideNet model.
• To improve the adaptability and accuracy of PDR systems across different walking speeds.

Main Methods:

• Developed LT-StrideNet, a hybrid deep learning model combining Long Short-Term Memory (LSTM) and Transformer architectures.
• Implemented a shank-mounted PDR framework incorporating LT-StrideNet for stride length estimation.
• Utilized peak detection with a dynamic threshold for pedestrian stride detection.
• Employed an Extended Kalman Filter (EKF) for sensor fusion (gyroscope, accelerometer, magnetometer).

Main Results:

• The proposed LT-StrideNet model effectively estimates pedestrian stride length.
• The stride length estimation method demonstrates adaptability to changes in pedestrian walking speed.
• The shank-mounted PDR framework exhibits excellent positioning performance.

Conclusions:

• LT-StrideNet offers a significant advancement in pedestrian stride length estimation.
• The developed PDR framework provides robust and accurate positioning capabilities.
• The approach effectively mitigates errors caused by variations in walking speed.