Fast AHRS Filter for Accelerometer, Magnetometer, and Gyroscope Combination with Separated Sensor Corrections

Area of Science:

• Robotics
• Sensor Fusion
• Navigation Systems

Background:

• Attitude and heading reference systems (AHRS) are crucial for navigation and control.
• Existing AHRS filters often face challenges with accuracy and computational efficiency, especially when using low-cost sensors.
• Microelectromechanical systems (MEMS) sensors offer a cost-effective solution but require sophisticated filtering techniques.

Purpose of the Study:

• To propose a new predictor-corrector filter for AHRS.
• To develop two variants of the filter: one for high accuracy using SO(3) operations and another for fast evaluation using linearization.
• To compare the proposed filters against state-of-the-art methods using MEMS sensor data.

Main Methods:

• A predictor-corrector filter structure utilizing gyroscope data for prediction and independent correction steps for accelerometer and magnetometer data.
• Implementation of two quaternion-based filter variants: one employing special orthogonal group SO(3) operations and another using linearization.
• Validation using data from low-cost MEMS sensors with ground truth provided by a VICON optical system.

Main Results:

• Both proposed filter variants demonstrated superior accuracy compared to conventional methods.
• The filters achieved this improved accuracy at a lower computational cost.
• The SO(3) variant provided high accuracy for nonlinear operations, while the linearization variant offered fast evaluation.

Conclusions:

• The novel predictor-correctors filters offer a significant advancement for AHRS applications.
• These filters provide a better trade-off between accuracy and computational cost, particularly for MEMS-based systems.
• Public availability of data and source codes facilitates further research and development in sensor fusion for AHRS.