UAV Integrated Navigation Technology Based On Multi-model Method

The rotorcraft unmanned aerial vehicle has been widely used in civil and military fields with its advantages of small size, light weight, and high mobility. With the increase of the diversity and complexity of the flight mission and flight environment, higher and higher request for the accuracy, reliability and independence of the UAV navigation system in the practical application was put forward. At present, the single model integrated navigation system based on multiple sensors is the common navigation method for the RUAV. It will bring model errs under the flight maneuver and varying environment conditions, and lead to the inaccurate navigation information. Multi-model estimation is a novel adaptive estimation method, which is suitable for complex systems with uncertain structure or parameters, and it can avoid the influence of inaccuracy of the single model. In recent years, the method was mainly used in the fields like target tracking, process control and so on. But the thorough research and application haven’t fully expanded to multiple information fusion technology of UAV navigation. To improve the accuracy and reliability of RUAV navigation system under complex maneuver and environment conditions, the key technology and application for multi-model integrated navigation system has been researched on.At present, the sensors used in RUAV navigation system mainly include MEMS sensors, GPS, barometric altimeter, sonar system etc. Error models establishment and calibration of the sensors is necessary to obtain more accurate navigation information. In this paper, the error characteristics of the airborne navigation sensors were first analyzed. And the error models of IMU, magnetometer, sonar and barometric altimeter were established respectively. Secondly, the error model parameters were calibrated according to the experimental analysis of the sensors. The sensors error was compensated by that, so that error accumulation of integrated navigation system cause by fixed error of sensors can be avoided.The error characteristics of airborne sensors are the basis of integrated navigation system. Because of the complex maneuver and environment, the accurate mathematical model of RUAV is very difficult to be established. Therefore, the multiple model method for time varying and uncertainty complex system was studied. Aiming at the increasing quantities of models and the inaccuracy of model identification, the Multi-model estimation method based on decentralized state was proposed. The number of models and computational cost can be reduced, and accuracy of the model recognition can be effectively improved by States estimating separately. Besides, for the Interactive multi-model algorithm relying too much on the model prior probability transfer matrix, an adaptive model transition probabilities method was proposed, which can effectively improve the speed and accuracy of the model shift, and provided theoretical basis for algorithm improvement of UAV multi-model integrated navigation system.After research and improvement of MM theory, its application and simulation experiment was carried on for practical problems of RUAV integrated navigation system. Because the bias error of gyro is affected by temperature greatly, bias error estimation based on extended state cannot eliminate attitude calculation error caused by the bias drift. Accordingly, an attitude estimation algorithm based on IMM is proposed. Through the establishment of gyro bias drift models under different environment, the multiple model fusion algorithm was adopted with information aided by magnetometer. It can eliminate the effect of varying bias error of gyro and improve the precision of UAV attitude estimation in complicated environment applications. Because the measurement noise model is unknown and time-varying in INS/GPS/barometric altimeter integrated navigation system of RUAV, a multiple model integrated navigation filtering algorithm with colored noise in time-varying situation on RUAV was finally designed. System model set was built according to the UAV measurement noise characteristics. The probability of each model can be calculated in real time by nonlinear UKF based IMM estimation algorithm. It can approach the real system model, avoid the one-sidedness of single model, and obtain the accurate state estimation to enhance the robustness of UAV navigation system.On the basis of the above theories and methods, in order to make a validation of the error characteristics and fusion method of RUAV integrated navigation system, the indoor and outdoor flight experiments were carried on based on the rotorcraft UAV platform constructed by laboratory autonomously. The effectiveness and feasibility of the UAV MM integrated navigation system scheme designed in this paper was verified.In this paper, the multi-model integrated navigation algorithm in rotorcraft unmanned aerial vehicles application was carried out for research work. And it has good reference value for engineering application.