A Novel Multidimensional Hybrid Position Compensation Method for INS/GPS Integrated Navigation Systems During GPS Outages

To mitigate the accuracy precipitous descent problem of the integrated navigation system that combines inertial navigation system (INS) and global positioning system (GPS) under complicated environments, especially in GPS outages, a novel multidimensional hybrid prediction method is proposed. First, a variational mode decomposition–improved wavelet denoising (VMD-IWD) data preprocessing algorithm is developed to reduce the internal noise of the INS, which offers distinct feature data for system integration. Second, a temporal dynamic attention neural network (TDANN) is proposed to predict GPS position increments to compensate for the missing navigation information of the system during GPS outages. The TDANN model combines spatial neighborhood aggregation (SNA), squeeze-excitation (SE), and time dynamic attention (TDA) modules to establish the nonlinear relationship between the inertial measurement unit (IMU) and GPS. When GPS is available, angular rate, specific force, and position increments are fed into the neural network for training simultaneously. When GPS outages occur, TDANN generates pseudo-GPS signals to correct the positioning accuracy of integrated navigation. Finally, simulation experiments and real road data tests are implemented to evaluate the performance of the proposed methods. The comparison results indicate that the proposed methodology can provide more stable and reliable navigation solution than the other existing machine learning algorithms in the event of GPS outages.