RFG-TVIU: robust factor graph for tightly coupled vision/IMU/UWB integration.

Area of Science:

• Robotics
• Sensor Fusion
• Navigation Systems

Background:

• High-precision navigation is crucial across various applications.
• Single sensors are insufficient for diverse navigation scenarios.
• Existing multi-sensor systems face challenges with sensor availability and data weighting.

Purpose of the Study:

• To develop a "plug and play" Vision/IMU/UWB multi-sensor tightly-coupled system.
• To enable real-time estimation of UWB base station coordinates without pre-calibration.
• To propose an adaptive robust factor graph model addressing dynamic sensor availability and observation weighting.

Main Methods:

• A factor graph-based tightly-coupled system integrating Vision, IMU, and Ultra-Wideband (UWB) sensors.
• Real-time estimation of UWB base station coordinates as system parameters.
• An adaptive robust factor graph model for dynamic sensor availability and weighted observation information.
• A novel adaptive estimation model for UWB ranging covariance without prior system information.

Main Results:

• The proposed system demonstrates superior performance compared to state-of-the-art methods.
• Significant Root Mean Square Error (RMSE) improvements were achieved: 62.83-64.26% in scene 1 and 70.32-82.15% in scene 2 (non-line-of-sight).
• The adaptive UWB covariance estimation effectively handles real-time changes.

Conclusions:

• The Vision/IMU/UWB tightly-coupled system offers robust and high-precision navigation.
• The adaptive factor graph approach enhances reliability in dynamic sensor environments.
• The system provides significant performance gains, particularly in challenging non-line-of-sight conditions.