基于多模信息融合的时空协同优化及自适应频率驯服算法研究

Research on Spatiotemporal Cooperative Optimization and Adaptive Frequency Disciplining Algorithm Based on Multi-Modal Information Fusion

摘要: 针对复杂环境下时频信号处理的高精度需求，提出了一种基于多模信息融合的时空协同优化及自适应频率驯服算法。通过整合北斗、GPS、惯性导航系统等多源传感器数据，结合独立分量分析(ICA)进行噪声分离与特征提取，构建了多模信息融合系统架构。在此基础上，设计了时空协同优化模型，利用图神经网络和固定区间平滑算法(FIS)提升时空信号处理精度。同时，进一步提出多相位锁相环(MP-PLL)和加权自适应卡尔曼滤波算法，有效抑制晶振频率漂移与跳变野值。实验结果表明，该算法极大地提升了定位精度、时间准确度及频率稳定性，显著超越国家现行标准，为高精度导航和授时系统提供了创新解决方案。

Abstract: To address the high-precision requirements of time-frequency signal processing in complex environments, this paper proposes a spatiotemporal cooperative optimization and adaptive frequency disciplining algorithm based on multi-modal information fusion. By integrating multi-source sensor data from BeiDou, GPS, and inertial navigation systems, and employing Independent Component Analysis (ICA) for noise separation and feature extraction, a multi-modal information fusion system architecture is constructed. On this basis, a spatiotemporal cooperative optimization model is designed, utilizing graph neural networks and Fixed Interval Smoothing (FIS) to enhance spatiotemporal signal processing accuracy. Furthermore, a Multi-Phase Phase-Locked Loop (MP-PLL) and a weighted adaptive Kalman filter algorithm are proposed to effectively suppress oscillator frequency drift and outliers. Experimental results demonstrate that the algorithm significantly improves positioning accuracy, time synchronization precision, and frequency stability, markedly surpassing current national standards, thereby providing an innovative solution for high-precision navigation and timing systems.