Research On Relative Navigation Method Of UAV Formation Based On Robust Adaptive Filtering

The future war will be a confrontation between the system and the system,and the unmanned aerial vehicle(UAV)formation battle as a new type of air combat model will gain widely attention from all the military powers.The UAV formation flight,is some UAVs fly according to a certain shape and keeps it throughout the whole flying process.The high precision navigation technology is the key to achieving the formation flight.In order to ensure the communication and coordination control accuracy among all the UAVs,and improve the integration combat efficiency,the relative navigation of the UAV formation based on the vision sensor is thoroughly studied in this paper.The main reserach contents are as follows:Considering the problem of the lower estimation accuracy and the slower convergence rate of the relative navigation for two UAVs formation due to the use of the traditional extended Kalman filter algorithm,the INS/Vis Nav relative navigation filter based on cubature Kalman filter is designed.So in this paper,the basic theory of unscented Kalman filter(UKF)and cubature Kalman filter(CKF)are introduced.The precisions of two nonlinear filtering algorithms are compared by using the Taylor series expansion.According to the capture degree of the higher order term of the Taylor function and the numerical stability analysis,the results show that CKF has higher estimation accuracy and better numerical stability for high dimensional nonlinear system.In order to avoid the singularity of the error covariance matrix caused by the limitation of the quaternion normalization,the Rodrigues parameters are introduced to represent the attitude error in the filter design,the simulation results show that this algorithm has obviously fast convergence performance.In order to solve the system modelling bias and the influence of unknown process noise and measurement noise on the filter performance,an adaptive Myers-Tapley filter based on covariance matching method is introduced.But the Myers-Tapley method for adaptively estimating the measurement noise and process noise covariance matrices make use of the sample mean and covariance of the residuals sample,which are nonrobust estimators.This lack of robustness implies that the performance Myers-Tapley adaptive method can degrade in the presence of non-Gaussianity.The Myers–Tapley adaptive tuning method is modified in order to account for non-Gaussianity by means of using the robust covariance estimates based on the projection statistics of the residuals sample.The modified adaptive algorithm can estimate the process noise and measurement noise covariance matrices along with the state estimate and state estimate error covariance matrix,which uses a robust approach to estimating these covariances that can resist the effects of outliers in the residuals sample.In view of the system modelling bias and non-Gaussian measurement noise of visual navigation sensor for the UAV formation relative navigation system,a robust adaptive cubature Kalman filtering algorithm is proposed which combined the improved Myers-Tapley adaptive noise statistical characteristics estimation method and the robust estimation algorithm based on Huber method.This algorithm modifies the measurement update of the standard cubature Kalman filter to a linear regression problem by making use of Huber estimation method,and takes advantage of the modified Myers-Tapley adaptive algorithm to improve uncertainty deviation of the system model and filtering estimation performance under the condition of non-Gauss noise.Then the algorithm is applied to the problem of relative navigation of UAV formation,the INS/Vis Nav relative navigation filter is designed which is based on the robust adaptive CKF.Compared to the tradition CKF and the robust CKF,the robust adaptive CKF has stronger adaptability to the non-Gauss measurement noise and system modeling uncertainty deviation,and can obtain better navigation parameters and better robustness.Aiming at the problem that the vision sensor can not distinguish the different characteristic light spots when the relative distance between formation UAVs is far away,a relative attitude determination method of UAV formation based on mutual survey line of sight vector is proposed.Based on three-vehicle formation model,the sensor measurement model is deduced,and the relative attitude solving problem under the case of mutual line of sight vector observation is studied.Covariance analyses are provided to gain insight on the astochastic properties of the attitude errors.A relative attitude estimation method based on CKF algorithm is proposed to improve the attitude estimation accuracy by combining the CKF algorithm with the attitude solving model.Finally,the estimation accuracy of the two methods is compared,and the performance of the system is verified by simulation.Through the research on the relative navigation problem of UAV formation,the robust CKF,the robust adaptive CKF and the relative attitude determination method based on the line of sight vector,are mainly studied.The effectiveness of this research is verified by both the theoretical analysis and numerical simulation.The related research results can provide an effective reference for the design of UAV relative navigation filter,and have important theoretical research significance and engineering application value.