DOGE: An Extrinsic Orientation and Gyroscope Bias Estimation for Visual-Inertial Odometry Initialization (2412.08135v1)

Abstract: Most existing visual-inertial odometry (VIO) initialization methods rely on accurate pre-calibrated extrinsic parameters. However, during long-term use, irreversible structural deformation caused by temperature changes, mechanical squeezing, etc. will cause changes in extrinsic parameters, especially in the rotational part. Existing initialization methods that simultaneously estimate extrinsic parameters suffer from poor robustness, low precision, and long initialization latency due to the need for sufficient translational motion. To address these problems, we propose a novel VIO initialization method, which jointly considers extrinsic orientation and gyroscope bias within the normal epipolar constraints, achieving higher precision and better robustness without delayed rotational calibration. First, a rotation-only constraint is designed for extrinsic orientation and gyroscope bias estimation, which tightly couples gyroscope measurements and visual observations and can be solved in pure-rotation cases. Second, we propose a weighting strategy together with a failure detection strategy to enhance the precision and robustness of the estimator. Finally, we leverage Maximum A Posteriori to refine the results before enough translation parallax comes. Extensive experiments have demonstrated that our method outperforms the state-of-the-art methods in both accuracy and robustness while maintaining competitive efficiency.

• The paper introduces a rotation-only constraint that accurately estimates extrinsic orientation and gyroscope bias without relying on translational parallax.
• The paper employs an advanced weighting mechanism with failure detection to integrate visual and inertial uncertainties for enhanced calibration precision.
• The paper leverages a Maximum A Posteriori framework to outperform state-of-the-art methods in robustness and accuracy on the EuRoC dataset.

Extrinsic Orientation and Gyroscope Bias Estimation for Improved Visual-Inertial Odometry Initialization

This paper introduces a novel methodology named DOGE for the initialization of Visual-Inertial Odometry (VIO), specifically addressing the estimation of extrinsic orientation and gyroscope bias which can degrade system performance over time as extrinsic parameters shift due to factors like temperature changes or mechanical stress. Traditional VIO initialization methods often rely on pre-calibrated extrinsic parameters; however, these can become inaccurate during extended use, particularly the rotational components. This misalignment necessitates a more robust approach for navigation in real-world applications, such as AR/VR systems and unmanned aerial vehicles, where precision is paramount.

Innovations of the Proposed Method

1. Rotation-Only Constraint: The DOGE method employs a unique rotation-only constraint for estimating extrinsic orientation and gyroscope bias without necessitating translational parallax. This constraint tightly integrates gyroscopic data with visual observations and is effective even in scenarios of pure rotational motion, which are commonplace in realistic navigation tasks.
2. Advanced Weighting and Failure Detection Strategy: The paper introduces a sophisticated weighting mechanism that incorporates both visual and inertial uncertainties into the constraints. This approach enhances the precision and robustness of the gyroscope bias and orientation estimation. An accompanying failure detection strategy helps ascertain the success of the initialization process, allowing for a recalibration as necessary.
3. Maximum A Posteriori Estimation: Before the onset of significant translational motion, DOGE leverages a MAP estimation framework to refine the estimates of extrinsic orientation and gyroscope bias, ensuring that early estimates are as accurate as possible.

Experimental Evaluation

Extensive tests conducted on the EuRoC dataset demonstrate that the proposed method outperforms existing state-of-the-art approaches in terms of both accuracy and robustness, maintaining competitive efficiency. The DOGE method consistently provides reliable extrinsic parameter estimation, even under significant rotational deviations—surpassing classical and recent techniques like VINS-Mono and other initialization algorithms.

Here are several highlighted results from the experiments:

• The DOGE method consistently outperformed others, maintaining lower errors in both gyroscope bias and extrinsic orientation across various datasets with different levels of rotational deviation.
• The implemented failure detection mechanism effectively minimized cases of incorrect initialization, showing a lower incidence of undetected errors compared to comparable methods.

Implications and Future Prospects

The implications of this work are substantial for the development of autonomous navigation systems, particularly in environments where rotational changes are frequent or where precise calibration cannot be ensured continuously. The proposed approach could significantly enhance the reliability of VIO systems through robust initialization, thereby improving the overall stability and performance of real-time navigation tasks.

Furthermore, this method offers a promising foundation for future research into more adaptive VIO initialization strategies, potentially integrating dynamic recalibration processes that consider ongoing environmental influences. Additional exploration could focus on how this methodology could be adapted for systems with varying intrinsic IMU parameters, especially in computationally constrained environments like mobile AR devices.

In summary, this paper presents a detailed and well-constructed framework for improving VIO initialization through innovative use of gyroscope bias and extrinsic orientation estimations, providing insights that hold potential significance for future advancements in VIO technology.