Online Calibration of Inertial Sensors Based on Error Backpropagation

Area of Science:

• Engineering
• Geomatics
• Computer Science

Background:

• Global Navigation Satellite Systems (GNSS) are primary for outdoor vehicle localization but fail in urban canyons and tunnels.
• Inertial Navigation Systems (INS) provide dead reckoning but suffer from accumulating integration errors.
• Low-cost Inertial Measurement Units (IMUs) exhibit thermal drift, necessitating frequent calibration.

Purpose of the Study:

• To develop a novel method for online (continuous) calibration of inertial sensors using GNSS data.
• To improve vehicle localization accuracy in environments with intermittent or poor GNSS signal quality.

Main Methods:

• Data fusion was performed using an extended Kalman filter (EKF).
• Inertial sensors were calibrated via error backpropagation.
• The system calibrates INS sensors during good GNSS reception and predicts position using INS during GNSS outages.

Main Results:

• The proposed online calibration method significantly improved localization precision.
• Performance was superior compared to offline calibrated inertial localization using the same sensors.
• The algorithm effectively handles GNSS signal degradation and failure scenarios.

Conclusions:

• Online calibration of inertial sensors using GNSS data is a viable and effective approach.
• This method enhances the robustness and accuracy of vehicle localization systems.
• The approach offers a practical solution for navigation in challenging environments like tunnels and urban areas.