TL;DR
This paper introduces a novel integrated navigation algorithm for marine vessels operating in challenging middle-to-high latitude regions. By unifying Earth models in both the SINS mechanization and filter, and employing an Adaptive Robust Kalman Filter (ARKF) with a hybrid-correction scheme, the algorithm significantly improves navigation accuracy and robustness, especially in the presence of uncertain DVL measurement errors.
What Problem Does This Paper Address?
Accurate navigation is crucial for marine operations, particularly in polar and high-latitude regions. Traditional navigation systems, like Strapdown Inertial Navigation Systems (SINS), suffer from accumulated errors and periodic oscillations. While integrating with Doppler Velocity Logs (DVLs) can mitigate these issues, existing methods face challenges:
- Meridian Convergence: Standard geographic frames become problematic at high latitudes due to meridian convergence. Grid frames are proposed as a solution, but switching between frames can introduce errors.
- Inconsistent Earth Models: Many integrated systems use different Earth models for SINS mechanization (e.g., reference ellipsoid) and the filter (e.g., sphere), leading to inherent errors and reduced accuracy.
- Measurement Noise and Outliers: DVLs, while valuable, can provide noisy or erroneous data, especially in challenging marine environments. Traditional Kalman Filters (KF) are sensitive to these uncertainties and can lead to filter divergence or inaccurate results.
- Correction Scheme Limitations: Existing output-correction schemes don't address SINS internal errors, while feedback-correction can be overly sensitive to DVL noise.
This paper aims to address these issues by proposing a unified, robust, and accurate integrated navigation solution for marine applications in middle-to-high latitudes.
Methods
The proposed algorithm integrates a grid SINS with a DVL using a novel approach:
- Unified Earth Model: The SINS mechanization and the integrated filter model both utilize the World Geodetic System 1984 (WGS-84) reference ellipsoid Earth model. This unification eliminates errors arising from inconsistent models.
- Grid Frame: A grid frame is employed to avoid meridian convergence issues inherent in traditional geographic frames at high latitudes.
- Adaptive Robust Kalman Filter (ARKF): To handle uncertain and potentially noisy DVL measurements, an ARKF is utilized. This filter uses an adaptive factor to detect and reduce the impact of measurement errors, making it more robust than a standard KF.
- Hybrid-Correction Scheme: A novel hybrid-correction scheme is introduced, combining:
- Output-Correction Channel: Corrects navigation output errors for immediate accuracy.
- Feedback-Correction Channel: Corrects SINS internal errors for long-term accuracy.
- Two-Order Switching Criterion: A sophisticated criterion that intelligently switches between output and feedback correction based on time and the ARKF's adaptive factor, ensuring optimal performance and preventing filter errors from propagating.
Datasets
The paper uses simulated data generated by a grid SINS trajectory generator and semi-physical experimental data.
- Simulated Data: Used to compare different filter models, correction schemes, and filter algorithms. DVL output errors were simulated to include zero-mean Gaussian white noises, constant errors (ocean current), and large random noises.
- Semi-Physical Experimental Data:
- IMU Data: Obtained from a high-precision three-axis turntable experiment to extract gyroscope and accelerometer errors.
- DVL Data: Extracted from a ship-mounted experiment in the Dalian sea area to capture real-world DVL error characteristics, including noise, large amplitude errors, and constant velocity errors when bottom lock is lost.
Key Insights and Ideas
- Unified Earth Models are Crucial: The paper strongly demonstrates that using a consistent Earth model (reference ellipsoid) for both SINS mechanization and the filter significantly improves navigation accuracy, especially in middle latitudes where the difference between geographic and geocentric latitudes is more pronounced.
- ARKF for Robustness: The ARKF's ability to adapt to uncertain DVL measurements is key to improving the overall system's robustness, particularly when dealing with non-Gaussian noise or outliers that would challenge a standard KF.
- Hybrid Correction for Comprehensive Error Management: The hybrid-correction scheme effectively balances the need to correct immediate navigation outputs (output-correction) with the necessity of addressing internal SINS errors (feedback-correction), all while intelligently managing the influence of DVL noise.
- Grid Frame Advantage: The grid frame proves effective in avoiding meridian convergence issues, making it a suitable choice for middle-to-high latitude marine navigation without the need for frame switching.
Key Results
- Improved Accuracy: The proposed grid SINS/DVL integrated navigation algorithm with a unified reference ellipsoid Earth model and ARKF-based hybrid-correction scheme achieved better navigation accuracy compared to traditional methods.
- In middle latitude regions, accuracies for latitude, longitude, and height improved by up to 19.21% (RMS) and 19.09% (maximum error) when using the reference ellipsoid model compared to a sphere model.
- In high latitude regions, similar improvements were observed, with accuracies improving by up to 13.20% (RMS) and 13.21% (maximum error).
- Enhanced Robustness: The ARKF-based hybrid-correction system demonstrated superior robustness against DVL measurement errors, especially large random noises and constant errors, compared to KF-based systems.
- Long-Term Performance: The hybrid-correction scheme, by addressing both output and internal SINS errors, maintained better navigation accuracy over extended periods compared to output-correction alone.
Limitations and Future Work
- Limitations:
- The paper focuses on the first two DVL failure scenarios (noisy output and constant velocity errors when bottom lock is lost), not explicitly discussing the case where no velocity information is provided by the DVL.
- While the ARKF is robust, the paper acknowledges that the feedback-correction channel of the hybrid system can still be influenced by DVL output errors, especially large ones.
- Future Work: The authors suggest that in the future, more kinds of external measurements will be employed to further improve the integrated navigation accuracy and reliability.
Summary updated