Research On Integrated Navigation Data Fusion Methods During GPS Signal Outages

In modern military war, the demand of the tracking accuracy and reliability for high dynamic aircraft, missiles and ships is higher and higher. Single system has been unable to meet the requirements, in order to solve this problem, there are complementary advantages by two or more navigation systems together. A combination of Global positioning system (GPS) and inertial navigation system(INS) are used in integrated navigation system to realize vehicle navigation and positioning. In the process of combination, we need to correct a certain frequency of inertial navigation by GPS to limit the accumulation of the error over time. But it’s unable when the GPS signal is disturbed or obscured. Therefore in this paper main studies high dynamic vehicle GPS/INS integrated navigation system signal processing methods during GPS outages. On the aspect of how to increase the accuracy and reliability, this paper main works as follows:1. The plane as the carrier of high dynamic, we simulated its multiple flight status to make trajectory by Matlab, as the foundation of the INS accelerometer and gyroscope module. Through calculating the INS locations of the integrated system and comparing with the real trajectory, we have verified the INS accumulated error over time. Therefore GPS data to correcting it frequently, show the necessity of the study.2. Research the theory and characteristic of artificial neural network, analyse the problems of standard Back Propagation(BP) neural network in convergence speed and generalization ability. In the application of signal processing model, we put forward a kind of BP neural network INS error prediction model based on Levenberg-Marquardt(L-M) bayesian regularization method. When GPS signal outages use the model to predict INS error for fixing its position data. Compared with BP neural network prediction model, we have verified its advantages.3. In generally, we will take random number or decide it from the experience for initial parameters of the BP neural network model, to destabilize the network model and to make network into a local minimum. In this paper proposes an hybrid modified particle swarm optimization method and BP neural network algorithm, which is used to establish the INS error model. We optimize the initial parameters of the neural network by particle swarm algorithm, and prediction model is produced by training the BP neural network. Hybrid algorithm is verified by matlab simulation, it can effectively improve the convergence speed and generalization ability of the model.