Adaptive Neural Unscented Kalman Filter

2025-03-07

AI Summary

To address the degradation or divergence of filtering performance in autonomous underwater vehicle (AUV) navigation caused by time-varying uncertainty in process noise covariance during tight coupling of inertial sensors and Doppler velocity logs (DVL), this paper proposes an Adaptive Neural Unscented Kalman Filter (AN-UKF). The core contribution is the design of ProcessNet—a lightweight, end-to-end regression network that enables real-time, data-driven adaptive estimation of the process noise covariance matrix and seamlessly integrates it into the unscented Kalman filter framework. Leveraging unscented transformation and recursive Bayesian estimation, the method demonstrates robustness and efficacy on real AUV navigation data. Compared with conventional UKF, adaptive UKF, and EKF, AN-UKF reduces positioning error by 32% while maintaining full numerical stability—no divergence observed throughout the entire trajectory.

Technology Category

• Intelligent Robots: State Estimation
• Machine Learning: Calibration & Uncertainty Quantification
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

The unscented Kalman filter is an algorithm capable of handling nonlinear scenarios. Uncertainty in process noise covariance may decrease the filter estimation performance or even lead to its divergence. Therefore, it is important to adjust the process noise covariance matrix in real time. In this paper, we developed an adaptive neural unscented Kalman filter to cope with time-varying uncertainties during platform operation. To this end, we devised ProcessNet, a simple yet efficient end-to-end regression network to adaptively estimate the process noise covariance matrix. We focused on the nonlinear inertial sensor and Doppler velocity log fusion problem in the case of autonomous underwater vehicle navigation. Using a real-world recorded dataset from an autonomous underwater vehicle, we demonstrated our filter performance and showed its advantages over other adaptive and non-adaptive nonlinear filters.

Problem

• Handles nonlinear scenarios in Kalman filtering
• Adjusts process noise covariance in real time
• Improves autonomous underwater vehicle navigation accuracy

Innovation

• Adaptive Neural Unscented Kalman Filter
• ProcessNet for noise covariance estimation
• Nonlinear sensor fusion for underwater navigation