Observability analysis of a matrix Kalman filter-based navigation system using visual/inertial/magnetic sensors

Area of Science:

• Robotics and Autonomous Systems
• Sensor Fusion
• Navigation Systems

Background:

• Integrated navigation systems commonly use sensor fusion to improve accuracy and robustness.
• Nonlinearities in sensor models present challenges for traditional filtering techniques.
• Visual, inertial, and magnetic sensors offer complementary data for navigation.

Purpose of the Study:

• To implement a matrix Kalman filter (MKF) for an integrated navigation system utilizing visual, inertial, and magnetic sensors.
• To analyze and establish the observability conditions for the nonlinear system within the MKF framework.
• To validate the derived observability conditions through experimental testing.

Main Methods:

• Implementation of a matrix Kalman filter (MKF) by transforming the nonlinear process model into a pseudo-linear one.
• Application of the observability rank criterion, utilizing Lie derivatives, to determine system observability.
• Experimental validation of the theoretical observability conditions.

Main Results:

• The matrix Kalman filter (MKF) was successfully implemented for the integrated navigation system.
• The study identified specific observability conditions: excitation of at least one rotational degree of freedom and observation of at least two independent horizontal lines and one vertical line.
• Experimental results confirmed the validity of these identified observability conditions.

Conclusions:

• The developed matrix Kalman filter (MKF) provides a viable approach for integrated navigation systems.
• The established observability conditions are crucial for ensuring reliable performance of visual/inertial/magnetic sensor-based navigation.
• Experimental validation underscores the practical applicability of the theoretical findings.