Research on Fusion Navigation Framework and Algorithm Based on Fully-connected Neural Network
Chunsheng Xu,Yunqing Liu,Zhanchen Zhu,Shuning Zhang,Ershen Wang,Jingyi Yi,Yongkang Wang
DOI: https://doi.org/10.1088/1361-6501/ad5198
IF: 2.398
2024-05-31
Measurement Science and Technology
Abstract:Fusion navigation and positioning have evolved into crucial technologies and methodologies within navigation systems. Currently, the majority of combined positioning frameworks employ Kalman filtering algorithms for data fusion. To explore more efficient and high-precision fusion architectures and algorithms, we introduce a fusion navigation framework based on a fully-connected neural network (FCNN). Initially, we conducted an analysis of existing fusion positioning technology, data fusion algorithms, and the application of artificial intelligence algorithms in navigation positioning. Drawing insights from the federated Kalman filter (FKF) architecture, FCNN, and attention mechanism, we propose a fusion navigation framework centred on FCNN. Finally, fixed-point and trajectory determination experiments were carried out in both open and semi-shielded environments. The results demonstrate that, compared to the traditional FKF architecture, the FCNN fusion navigation framework, coupled with the attention mechanism fusion algorithm, effectively accommodates data, mitigates errors, and achieves superior positioning accuracy.
engineering, multidisciplinary,instruments & instrumentation