Abstract
Zero velocity update is a common and efficient approach to bound the accumulated error growth for foot-mounted inertial navigation system. Thus a robust zero velocity detector (ZVD) for all kinds of locomotion is needed for high accuracy pedestrian navigation systems. In this paper, we investigate two machine learning-based ZVDs: Histogram-based Gradient Boosting (HGB) and Random Forest (RF), aiming at adapting to different motion types while reducing the computation costs compared to the deep learning-based detectors. A complete data pre-processing procedure, including a feature engineering study and data augmentation techniques, is proposed. A motion classifier based on HGB is used to distinguish “single support” and “double float” motions. This concept is different from the traditional locomotion classification (walking, running, stair climbing) since it merges similar motions into the same class. The proposed ZVDs are evaluated with inertial data collected by two subjects over a 1.8 km indoor/outdoor path with different motions and speeds. The results show that without huge training dataset, these two machine learning-based ZVDs achieve better performances (55 cm positioning accuracy) and lower computational costs than the two deep learning-based Long Short-Term Memory methods (1.21 m positioning accuracy).
Highlights
For more than 15 years, research in pedestrian navigation has focused on solutions based on inertial signals to provide accurate and continuous indoor/outdoor positioning solutions
Zero-velocity detectors (ZVDs) find the instants that the inertial measurement unit (IMU) sensors are static corresponding to the stance phase of a walking gait
The second step predicts the ZUPT instants using artificial intelligence (AI)-based techniques. This process is applied to all three algorithms: Random Forest (RF), Histogrambased Gradient Boosting (HGB) and Long short-term memory network (LSTM)
Summary
For more than 15 years, research in pedestrian navigation has focused on solutions based on inertial signals to provide accurate and continuous indoor/outdoor positioning solutions. The use of walking phases derived from biomechanical knowledge [3], [4] was quickly adopted to replace the lack of GNSS (Global Navigation Satellite System) measurements in buildings. They are used to periodically bound the propagation of the inertial errors in the positioning algorithms [5]. It is found that these features enable robust detection of ZUPT even with rapidly changing movements This is less true for LSTM trained on globally small databases, since LSTM does not allow random segmentation of the learning database at the sample level to prevent the risk of breaking temporal correlations
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have