Abstract
With the continuous popularization of Global Navigation Satellite System (GNSS) in various applications, the performance requirement for integrity is also increasing, especially in the field of safety-of-life. Although the existing Receiver Autonomous Integrity Monitoring (RAIM) algorithm has been embedded in the GNSS receiver as a standard method, it might still suffer from small fault detection and delay alarm problem for time series fault models. In an effort to solve this problem, a Deep Neural Network (DNN) for RAIM, named RAIM-NET, is investigated in this paper. The main idea of RAIM-NET is to propose a combination of feature vector extraction and DNN model to improve the performance of integrity monitoring, with a problem specifically designed for loss function, obtaining the model parameters. Inspired by the powerful advantages of Recurrent Neural Network (RNN) in time series data processing, a multilayer RNN is applied to build the DNN model structure and improve the detection rate for small faults and reduce the alarm delay for the time series fault event. Finally, real GNSS data experiments are designed to verify the performance of RAIM-NET in fault detection and time delay for integrity monitoring.
Highlights
Global Navigation Satellite System (GNSS) has been an essential part of various civil and military systems and has been constantly spreading to many new applications, such as self-driving car, internet of things, transportation, big data, etc. [1,2,3]
The main idea of Receiver Autonomous Integrity Monitoring (RAIM)-NET is to propose a combination of feature vector extraction and a Deep Neural Network (DNN) model to improve the performance of integrity monitoring, with a problem designed for loss function obtaining the model parameters
AlthAoluthgohugthhetheenedn-dt-ot-oe-ennddDDNNNss tthhaattuusesethtehreawradwatdaaatsathaes itnhpeuitnopf uthteonfetuhrealnneeutwraolrkneartewionrk are in fafsahshioionn, tthhee nnetwoorrkk aarrcchhitietcetcutruersesgegtemt omreoraenadnmdomreocroemcpolmexpldeuxedtuo ethteo ltahcek loafckbaocfkgbraocukngdround knowknloedwgleedfgoerfomromreorcehcahllaellnegnigninggtatasskkss [[2299]]
Summary
Global Navigation Satellite System (GNSS) has been an essential part of various civil and military systems and has been constantly spreading to many new applications, such as self-driving car, internet of things, transportation, big data, etc. [1,2,3]. The performance of the snapshot algorithms is not stable due to the complex and changeable environment of the receiver in many new applications To solve this problem, other sensors or datasets are adopted as aids to provide additional measurements for RAIM. To solve the non-linear and non-Gaussian problem for integrity monitoring, Peng proposed a temporal RAIM algorithm based on particle filter [23]. The Back-Propagation (BP) neural network was used to adjust the particles to improve the performance in fault detection under the conditions of non-Gaussian measurement noise [26]. The main idea of RAIM-NET is to propose a combination of feature vector extraction and a DNN model to improve the performance of integrity monitoring, with a problem designed for loss function obtaining the model parameters.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have