Wavelet denoising of inertial sensor input data for GNSS-INS integration: simulation approach

Abstract

Integrated Global Navigation Satellite System (GNSS) and Inertial Navigation System (INS) architectures are widely employed to achieve reliable and continuous navigation under conditions of signal blockage, interference, and degraded satellite visibility. Although GNSS provides long-term positioning accuracy, its performance deteriorates in challenging environments, whereas INS offers continuous navigation as a self-contained system but suffers from accumulated inertial sensor errors and drift over time. These limitations significantly affect positioning reliability, particularly during GNSS-denied operation. This paper presents a simulation-based investigation of wavelet-driven denoising techniques for preprocessing inertial sensor acceleration measurements in GNSS–INS integrated navigation systems. The study systematically evaluates and compares multiple wavelet-domain denoising strategies including decimated wavelet transform (DWT), un-decimated wavelet transform (UDWT), and least-square-method (LSM)-assisted adaptive thresholding integrated with UDWT for suppressing inertial sensor noise while preserving signal characteristics. Unlike conventional implementations that employ a single wavelet configuration, the proposed framework establishes a comparative methodology to analyze the trade-off between noise attenuation, reconstruction fidelity, computational complexity, and implementation constraints under practical navigation conditions. A realistic simulation environment incorporating inertial sensor error sources, GNSS observation uncertainty, and representative motion conditions was developed to evaluate denoising performance. The results demonstrate that DWT provides reduced memory usage and lower computational burden but may introduce reconstruction distortion due to coefficient down-sampling. In contrast, UDWT preserves temporal continuity and improves signal reconstruction quality at increased computational cost. The LSM-assisted UDWT approach further enhances denoising performance by reducing reconstruction artifacts and improving signal smoothness. The study demonstrates that wavelet-based preprocessing of inertial measurements can improve INS robustness and support more reliable GNSS–INS integrated navigation under real-time operational scenarios.

The authors' abstract, as published at the source. Scientific Reports, 2026 · DOI ↗