Error compensation method of GNSS/INS integrated navigation system based on PSO-LSTM

The integrated navigation system based on the Global Navigation Satellite System and Inertial Navigation System (GNSS/INS), can provide continuous and dependable positioning information for airborne motion. The GNSS signal outages are easily caused by electromagnetic interference environments such as high-rise buildings, and canyons high-voltage towers. A long period of missing GNSS positioning information will lead to a rapid decline in the navigation precision of the GNSS/INS system. In order to solve this problem, the optimum value of Long Short-Term Memory (LSTM) is obtained through Particle Swarm Optimization (PSO). An LSTM network model based on the PSO algorithm (PSO-LSTM) is designed. It is used to assist the integrated navigation of GNSS/INS based on Kalman Filter (KF). When GNSS signals are available, the PSO-LSTM model is trained with carrier dynamic information and navigation information. When GNSS signals are not available, the PSO-LSTM model is used to obtain pseudo-GNSS signals for Kalman Filter measurement updates. In order to verify the validity of the algorithm, unmanned aerial vehicle (UAV) flight tests of GNSS signal outages are performed. The results of the positioning are compared to the conventional LSTM model. When the GNSS signal experiences outages of 30 s, the maximum positioning error of the LSTM model is 4.142 m. The PSO-LSTM model is 2.883 m, which decreases by 30.4 % relative to LSTM. When the GNSS signal experiences outages of 60 s, the maximum positioning error of the LSTM model is 5.992 m. The PSO-LSTM model is 2.898 m, which decreases by 51.2 % relative to LSTM. When the GNSS signal experiences multiple outages, the maximum positioning error of the LSTM model is 11.362 m. The PSO-LSTM model is 6.042 m, which decreases by 46.8 % relative to LSTM.