Continuous Fusion of Motion Data Using an Axis-Angle Rotation Representation with Uniform B-Spline.

Area of Science:

• Robotics and Automated Driving
• Sensor Fusion
• Motion Estimation

Background:

• Existing motion data fusion methods (filter-based, pose-graph-based) have limitations, including careful parameter tuning, unidirectional fusion, and discrete results.
• Filter-based methods require meticulous parameter settings and typically process data chronologically.
• Pose-graph methods allow bidirectional fusion but necessitate pre-integration for inertial measurement unit (IMU) data, yielding discrete outputs.

Purpose of the Study:

• To develop a unified, continuous approach for fusing motion data from diverse sources like IMUs and localization systems.
• To overcome the limitations of discrete fusion results inherent in current filter-based and pose-graph-based methods.
• To enhance the robustness, accuracy, and efficiency of motion data fusion in robotics and autonomous driving applications.

Main Methods:

• A novel continuous fusion approach utilizing uniform B-splines as the optimization backbone.
• Integration of motion measurements from inertial measurement units (IMUs).
• Incorporation of pose data from external localization systems.
• Application of axis-angle representation for rotation data.

Main Results:

• The proposed B-spline-based approach successfully fuses motion data continuously, accurately, and robustly.
• Evaluations on real-world data validate the effectiveness of the continuous fusion concept.
• Achieved superior performance compared to existing discrete fusion techniques.

Conclusions:

• The uniform B-spline-based continuous fusion method provides a unified solution for integrating motion data from IMUs and localization systems.
• This approach delivers accurate, robust, and continuous motion fusion results, advancing the field of robotics and automated driving.
• The findings support the viability and benefits of continuous fusion methodologies over discrete ones.