Research On Vehicle INS/GNSS Integrated Navigation Filtering Algorithm

With the development of intelligent transportation systems and smart cities,more Location-based Services(LBS)will be provided to traffic users to enhance traffic safety and experience,the application of unmanned technology has become a tendency,and will also highly rely on precise location information.We all know that the Global Navigation Satellite System(GNSS)can provides accurate location information with time independent,but in urban environments,satellites signals can be greatly interfered by the environment,because satellites signals can be easily obstructed and reflected by buildings,and if receivers only capture signals with low Signal-to-Noise ratio or with multipath,it can cause great positioning mistake.By the way,the Strap-down Inertial Navigation System(SINS)can provide precise relative position information in a short time,but limited by error accumulation,it cannot work for a long time independently.Fortunately,the integrated navigation system provides an effective solution for solving problems of single navigation system with poor positioning accuracy and stability,by designing SINS/GNSS integrated navigation systems,using the satellite positioning information to periodically correct the inertial system and suppressing its error,to ensure long-term stability of the inertial system,which can effectively solve the problem of accurate location information acquisition in urban environments.Sensors information fusion and filtering algorithm is the key technology for integrated navigation.In dealing with state estimation and noise filtering,those filters based on the minimum mean square error criterion(MMSE)are proved optimal in dealing with the Gaussian noise.However,the method based on MMSE can only captures first-order and second-order statistics of the noise.If noises are non-Gaussian,especially heavy-tailed and outlier noises,the filter performance under MMSE will degraded.The non-Gaussian noise in vehicle integrated navigation systems appeared as measurement noise,in most previous studies,the measurement noise of integrated navigation systems was treated as Gaussian white noise,of course some studies based on non-Gaussian assumptions,but massive calculations are required.In order to eliminate non-Gaussian noise in integrated navigation system,and consider the nonlinear characteristics of actual system,the maximum correntropy criterion unscented Kalman filter(MCC-UKF)is applied,and adaptive strategy of MCC-UKF was researched.The correntropy is defined as a statistical metric ofthe similarity between two random variables in information theoretic,which is different from the MMSE,it measures not only the second-order information but also higher-order information between a pair of scalar random variables.The Unscented Kalman filter under Maximum Correntropy,its equation is solved by the weighted least squares method,and the relationship between the correntropy and the optimal estimation is established,that helps to iterate optimal estimation value which under correntropy criterion in the filtering model.The algorithm selects Gaussian kernel function to effectively capture high-order statistical information of noise.When the system experiences large deviations,especially impulsive noise,the negative exponential term in the Gaussian kernel and the threshold value of the kernel size weakens the outlier relative to the correntropy.At last,the simulation and analysis of the integrated navigation system is performed.By comparing with the traditional UKF,the effectiveness of the MCC-UKF to solve non-Gaussian noise is verified.