An Improved Robust ESKF Fusion Positioning Method with a Novel UWB-VIO Initialization.

Area of Science:

• Robotics
• Sensor Fusion
• Localization

Background:

• Visual-inertial odometry (VIO) faces challenges in complex indoor environments due to illumination variations, sparse features, and inertial drift.
• Scale uncertainties and accumulated errors degrade VIO performance in mobile robot localization.
• Existing methods often require external calibration and coordinate transformations for Ultra-Wideband (UWB) integration.

Purpose of the Study:

• To propose a new UWB-VIO initialization method and an enhanced Robust error-state Kalman filter (Robust ESKF) fusion technique.
• To address scale drift and heading inconsistency during VIO initialization.
• To improve the robustness and stability of mobile robot localization in challenging indoor settings.

Main Methods:

• Developed a direction-consistent constrained initialization model for joint optimization of scale factor and yaw angle.
• Implemented an improved residual-weighted robust filtering mechanism to mitigate abnormal UWB ranging data and noise.
• Integrated UWB and VIO using a novel initialization and fusion approach, eliminating the need for separate calibration.

Main Results:

• The proposed method significantly reduces trajectory drift and attitude jumps in narrow-corridor environments.
• Achieved over 50% enhancement in average localization accuracy compared to conventional UWB-VIO algorithms.
• Demonstrated stable estimation and high precision even under severe multipath interference conditions.

Conclusions:

• The novel UWB-VIO initialization and Robust ESKF fusion method significantly improves mobile robot localization accuracy and robustness.
• The direction-consistent initialization effectively resolves scale and heading ambiguities.
• The adaptive filtering mechanism successfully suppresses UWB outliers, leading to superior performance in challenging environments.