Abstract:

Aiming at the problem that the traditional fault-tolerant method of tightly coupled global navigation satellite system/inertial navigation system (GNSS/INS) is insufficient to adapt to the environment and to solve the fault, an adaptive fault-tolerant method based on long-term and short-term memory neural network is proposed. In this method, GNSS pseudo-range and pseudo-range rate prediction models were established based on long and short-term memory neural network. When a fault occurs, the dimension of the fault observation is located by the component detection method, and the relative differential precision of positioning is introduced to analyze the impact of the fault observation on the positioning accuracy of the system and to realize the dynamic selection of the isolation and reconfiguration strategy. Utilizing the actual measurement data, simulation experiments are conducted by setting up multiple environments from three perspectives: number of visible stars, geometric configuration, and fault duration. The simulation results show that the proposed method has better adaptability to complex environments, and can effectively reduce the localization error of the system during faults existing period and improve the fault detection performance of the system.

Key words: tightly coupled global navigation satellite system/inertial navigation system(GNSS/INS), fault-tolerant, long-short term memory (LSTM) neural network, positioning accuracy, fault detection