.png){kind=icon}
.png){kind=icon}
Return
Factor Graph Optimization Localization Method Based on GNSS Performance Evaluation and Prediction in Complex Urban Environment
.png){kind=icon}
PRE
AI
DOI：10.1109/JSEN.2025.3542058.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
This article proposes an online global navigation satellite system (GNSS) positioning performance evaluation and position prediction method to handle the degradation of positioning accuracy due to the complex urban denial environment. A dynamic trust (DT) function is constructed by combining multiparameter metrics to dynamically filter inavailable information and optimize information utilization. An improved indirect position prediction model based on bi-directional long short-term memory (BiLSTM) and strapdown inertial navigation system toward the heading error divergence model (SINS-HEDM) is constructed to enhance the accuracy of the navigation system. In order to reduce the interference of human driving behavior on the direction information in the position, the position is decomposed into distance and direction. BiLSTM is employed to predict vehicle movement distances between adjacent moments, and SINS-HEDM is designed to compensate for heading errors in SINS. A robust factor graph optimized (FGO) fusion method is presented for achieving reliable vehicle positioning in urban GNSS-denied environments. A comparative experiment is adopted to demonstrate the superiority of the proposed method.
Keywords:
Global navigation satellite system
Predictive models
Vehicle dynamics
Satellites
Bidirectional long short term memory
Accuracy
Receivers
Position measurement
Performance evaluation
Inertial navigation
Bi-directional long short-term memory (BiLSTM)
dynamic trust (DT) function
factor graph
heading error dispersion model
vehicle position prediction