Listen to this content
GNSS receivers combined with inertial navigation systems (INS) have been widely applied to various mobile platforms.
However, in Arctic regions, GNSS positioning accuracy is severely degraded from low satellite elevation angles, frequent ionospheric disturbances, and insufficient visible satellites.
Moreover, the limited validation of existing onboard navigation systems further exacerbates the challenges of Arctic navigation.
To address these issues, a new research paper describes a hybrid neural network model based on temporal convolutional networks (TCN) and long short-term memory (LSTM) networks. The hybrid solution has been tested in the Artic with successful results.
The paper, “Robust GNSS/INS Integrated Navigation in Arctic GNSS-Challenged Environments Based on TCN-LSTM and MDAREKF,” is authored by Wei Liu, Tengfei Qi, Yuan Hu, Kaiwei Zhu, Tsung-Hsuan Hsieh and Shengzheng Wang of Shanghai Maritime University (DOI 10.1088/1361-6501/ae5279).
The proposal combines the pseudo-measurement information of GNSS predicted by the model with INS for integrated navigation to compensate for the interruption of GNSS and correct the error of INS.
Considering the potential bias in predicted pseudomeasurements, an adaptive robust extended Kalman filter (AREKF) algorithm based on Mahalanobis distance is further developed to dynamically adjust the innovation covariance matrix, thereby enhancing filter robustness.
Field experiments conducted on an Arctic survey vessel demonstrate that the proposed TCN-LSTM combined with AREKF significantly improves both the robustness and accuracy of integrated navigation under GNSS-constrained environments. In particular, during GNSS outages of 50 seconds, 140 seconds and 400 seconds, the proposed method reduces the horizontal root mean square error (RMSE) by 47%, 38% and 76% respectively.
