The application of recurrent neural networks to improve the accuracy of navigation systems of mobile objects
Аuthors
*,Saint Petersburg State University of Aerospace Instrumentation, 67, Bolshaya Morskaya str., Saint Petersburg, 190000, Russia
*e-mail: ammar.assad225@gmail.com
Abstract
The relevance of this study stems from the critical need to enhance the accuracy of navigation systems for mobile objects, particularly in scenarios where the Global Positioning System (GPS) experiences disruptions. While integrated systems combining GPS and Inertial Navigation Systems (INS) offer high-precision navigation, their performance degrades significantly during GPS outages. To address this challenge, this article introduces a novel EKF-RNN method, which integrates a Extended Kalman Filter (EKF) with a Recurrent Neural Network (RNN) to improve navigation accuracy under such adverse conditions.
The novelty of the research lies in the innovative use of RNNs to process navigation data during GPS failures. The RNN leverages historical GPS measurements alongside data from INS sensors, such as gyroscopes and accelerometers, to generate corrective outputs. These outputs are then used to refine the measurements within the EKF framework, thereby enhancing the estimation of navigation parameters.
To validate the effectiveness of the proposed method, experiments were conducted using two real-world datasets. A comprehensive comparison was made between the traditional EKF and the EKF-RNN methods across key performance metrics, including orientation errors (attitude), horizontal positioning errors, and velocity (V_(ned) ). The results demonstrate that the EKF-RNN method outperforms the conventional EKF across all evaluated parameters. Specifically, the EKF-RNN achieves higher accuracy in orientation estimation, significantly reduces horizontal positioning errors, and improves velocity estimation.
The findings showed the effectiveness of the EKF-RNN method in enhancing the robustness and accuracy of navigation systems during GPS disruptions. This makes it a promising solution for deployment in navigation systems for mobile objects, particularly in applications where reliability and precision are paramount.
Keywords:
Global Positioning System, Inertial Navigation System, Extended Kalman Filter, Recurrent Neural NetworkDownload