TL;DR
This paper introduces a new family of lightweight, low-power INS/GNSS systems designed for challenging flight applications, including aerobatic maneuvers and aircraft motion surveying. The core innovation lies in a sophisticated 42+ state Extended Kalman Filter (EKF) data fusion algorithm that accounts for sensor aging, a critical factor for MEMS inertial sensors, and demonstrates superior performance compared to a competitor's system in extreme conditions.
What Problem Does This Paper Address?
Accurate navigation and motion tracking are crucial for a wide range of applications, from unmanned aerial vehicles (UAVs) to high-performance aircraft. Inertial Navigation Systems (INS) combined with Global Navigation Satellite Systems (GNSS) are a common solution, but achieving high accuracy, especially under extreme dynamic conditions like aerobatic flight, presents significant challenges. These challenges include:
- Sensor Drift and Aging: MEMS inertial sensors, while lightweight and low-power, are prone to drift and aging effects that degrade accuracy over time. Existing systems often don't deeply consider these factors.
- Extreme Dynamics: Aerobatic maneuvers involve very high angular rates (up to 400°/s) and high g-forces (up to 10g), which can overwhelm standard INS/GNSS systems, leading to significant trajectory errors.
- GNSS Outages: During aerobatic maneuvers, the aircraft's orientation can lead to prolonged GNSS signal outages, requiring the INS to maintain accuracy independently for extended periods.
- System Integration and Usability: Developing a robust, plug-and-play system that integrates various inertial sensor technologies (from low-cost MEMS to high-performance FOG/RLG) and GNSS receivers, while remaining user-friendly, is a complex engineering task.
The paper aims to address these issues by presenting a novel INS/GNSS system architecture and a powerful data fusion algorithm that can handle extreme dynamics, mitigate sensor aging effects, and maintain accuracy during GNSS outages.
Methods
The paper details a new family of iNAT INS/GNSS systems from iMAR Navigation. The key methodological contributions include:
- Modular Hardware Architecture: The iNAT systems are built on a common hardware platform that allows for the integration of various inertial sensor technologies (MEMS, FOG, RLG, HRG) and GNSS receivers. This modularity allows for different performance classes while maintaining a consistent software interface.
- 42+ State Extended Kalman Filter (EKF): The core of the system's performance lies in a highly sophisticated EKF data fusion algorithm. This filter is designed to:
- Estimate and Correct Aging Parameters: It explicitly estimates and corrects parameters affected by sensor aging and hysteresis, a crucial aspect for MEMS sensors that is often overlooked.
- Handle Extreme Dynamics: The algorithm is designed to process very high dynamic motions, including advanced coning compensation.
- Robust Outlier Detection: A powerful outlier detection mechanism is employed to minimize the impact of distorted aiding information.
- Tightly-Coupled Data Fusion: This approach allows for accurate navigation even with limited GNSS satellite visibility or during outages.
- In-Motion Alignment: The system features a novel "In-Motion Alignment" routine that allows for alignment even while the aircraft is moving, eliminating the need for lengthy ground alignment procedures and enabling immediate takeoff.
- High-Precision Time Synchronization: The system ensures time stamping of all sensor data with better than 1 µs accuracy, which is critical for fusing data with varying latencies, especially between inertial sensors (up to 1 kHz) and GNSS data (up to 500 ms latency).
Datasets
The paper primarily uses flight data from real-world applications for evaluation:
- Red Bull Air Race 2016: Data was collected from flights during the Red Bull Air Race, specifically from Matthias Dolderer's aerobatic aircraft. This dataset represents extreme dynamic conditions, including high g-forces, high angular rates, and significant GNSS outages.
- Airborne Gravimetry Test Flight: Data from a 3.3-hour flight in the USA was used to demonstrate the system's performance in gravimetric surveying applications.
The paper does not introduce new benchmark datasets but rather uses existing challenging real-world scenarios to validate the system's performance.
Key Insights and Ideas
- Sensor Aging is Critical for MEMS: The paper strongly emphasizes that for MEMS inertial sensors, accounting for aging and hysteresis effects is not just an optimization but a necessity for achieving high and consistent accuracy, especially in demanding applications.
- 42+ States are Necessary for Extreme Dynamics: The choice of a 42+ state EKF is presented as a key enabler for handling the complex dynamics and sensor characteristics encountered in aerobatic flight, going beyond simpler 16-state filters often used in UAV applications.
- In-Motion Alignment is a Game-Changer for Aviation: The "In-Motion Alignment" feature significantly reduces operational overhead by allowing immediate takeoff without lengthy ground procedures, a critical advantage for time-sensitive aviation operations.
- Tightly-Coupled Fusion Excels in GNSS-Challenged Environments: The paper demonstrates that a tightly-coupled INS/GNSS fusion approach is vital for maintaining navigation accuracy during prolonged GNSS outages, a common occurrence in aerobatic flight.
- High-Performance Inertial Sensors Unlock New Applications: The paper highlights that by combining advanced data fusion with high-performance inertial sensors (like RLG), applications such as airborne gravimetry can achieve remarkable accuracy without specialized equipment.
Key Results
- Superior Trajectory Accuracy in Aerobatics: In Red Bull Air Race tests, the iNAT-M200/SLN system demonstrated significantly better trajectory accuracy compared to a competitor's INS/GNSS system, with deviations of up to 50 meters or more for the competitor during GNSS outages and high-dynamic maneuvers.
- High-G Performance: The system successfully handled accelerations peaking at over 11g and prolonged GNSS outages of up to 15 seconds or more during loops and barrel rolls without discontinuities in its navigation solution.
- Accurate Navigation Under Extreme Conditions: Estimated deviations for the iNAT-M200 system were reported as:
- Position: better than 1 m RMS (NED)
- Velocity: better than 0.1 m/s RMS (East/North/Down)
- Attitude: Roll/Pitch better than 0.05 deg RMS, Heading better than 0.1 deg RMS.
- Airborne Gravimetry Performance: For gravimetric surveying, the iNAT-RQH system achieved navigation performance significantly better than 0.2 nm/h and static alignment errors of less than 0.03 deg in heading. Attitude errors were as low as <0.0006 deg in roll/pitch and <0.003 deg in heading, with velocity errors <8 mm/s.
- Low Power and Lightweight: The iNAT-M200/SLN is noted for its low power consumption (8W) and light weight (0.8 kg), making it suitable for small unmanned vehicles.
Limitations and Future Work
- Focus on Specific Applications: While the paper presents results for aerobatic flight and gravimetry, the full breadth of applications for the iNAT system family is not exhaustively explored.
- "Plug & Play" vs. Aiding Sources: The authors mention that while their system can incorporate magnetic heading and true airspeed as aiding sources, they believe that for surveying applications, a pure INS/GNSS approach is preferable to avoid complexity in setup, especially for UAVs where magnetic interference can be an issue. This suggests a trade-off between ultimate flexibility and ease of use.
- Future Work (Implied): The paper's success in demonstrating high performance in extreme conditions and for specialized applications like gravimetry suggests that future work could focus on further miniaturization, integration with other sensor modalities, and broader adoption in various unmanned and manned aviation platforms. The emphasis on sensor aging also points towards continued research in robust sensor modeling and calibration techniques.
Summary updated