Study On Support Vector Regression Aided GNSS/INS Integrated Navigation Model

GNSS/INS integrated navigation combines the advantage of GNSS navigation and INS navigation,and has been widely used in military and civil fields,Based on the simulation of GNSS/INS integrated navigation system,the integrated navigation model based on support vector regression is studied with the consideration of system fault or abnormal situation.The main research contents are:(1)Considering the poor ability of GNSS/INS integrated navigation in the region where the GNSS signal is not accessible,a method based on support vector regression and observation difference to eliminate colored noise is proposed.The method utilized grid search and cross-validation to find the reasonable parameter of SVR,setting up nonlinear mapping model between the increment of position,velocity and the increment of acceleration,attitude,when the GNSS signal is outage,the model can predict the increment of position and velocity of user,which is used to get the position and velocity of user.Because of the accumulation of noise,the predicted position and velocity errors can be treated as colored noise,while designed into state matrix,it can be eliminated by differenced method,achieving stable and reliable navigation solution under the condition of short interruption of GNSS signal.(2)Considering the shortage of the normal robust adaptive kalman filtering algorithm in identifying the source of system abnormal on the condition of lacking redundant observations,some low precision navigation solutions are generated,a robust adaptive kalman filtering algorithm based on support vector regression is put forward.The algorithm can separate the anomaly of observation and dynamic model using the predicted velocity and position aided anomaly detection under the condition of the whole system abnormality,which can improve the overall accuracy of navigation solution.The simulation results show that support vector regression aided GNSS/INS Integrated Navigation behaves higher stability and reliability.