Application Of Different Filtering Algorithm In Autonomous Orbit Determination Of Navigation Constellation

The system model of autonomous orbit determination for navigation constellation is nonlinearsystem. Density distributions of State variables and error do not satisfy Guassian distribution strictly.Nonlinear system needs to be linearized firstly when Extended Kalman Filter (EKF) applied in it.Moreover, state and observation must satisfy Guassian distribution. EKF must bring in rounding errorin practical application for the reason above. Although UKF is a kind of linear algorithm that avoidslinearization error, it asks for Guassian distribution of input variable yet, and the number of sigmapoints is constant which leads to the bad flexibility. In addition, UKF algorithm takes too much timeand real-time is poor when applied in high dimension system.In order to solve this problem, this paper considers Particle Filter (PF) which is the most suitablefor nonlinear and non-Guassian system. Reserch and improvement are made just on the problem thatPF depends on system model seriously and large quantity calculation leads to poor real-time. Themain study is as follows:1. Based on the basic theory of autonomous orbit determination for navigation constellation, thispaper summarizes the filter model including the perturbation equation of satellite orbit which is regardas state model, the main disturbing force, the crosslink range observation model and inter-satelliterorientation observation model. Furthermore, the condition of getting observation variable and theobservation equation are introduced.2. Under the circumstance that only crosslink range observation is adopted, this paper uses PFalgorithm to overcome the limitation of nonlinear and non-Guassian to improve the precision oflong-term autonomous orbit determination. PF is combined with EKF to overcome the seriousdependence to system model. Improved EKPF algorithm is proposed to improve the poor real-time ofstandard EKPF algorithm when used in high dimension system. Improved EKPF algorithm overcomesnot only the linear limitation, but also improves the real-time.3. Only crosslink range observation adopted can only measure relative position of satellites inconstellation, but can not overcome the barrier of constellation rotation error. Some scholarsintroduced the inter-satellite orientation observation to solve the problem above. However,obsevations increasing costs with the loss of real-time. On this basis, this paper introduces FederatedFilter (FF) algorithm. Crosslink range observation and inter-satellite orientation observation are usedas two different local filters which adopts improved EKPF as filter algorithm, respectively. Especially,biology evolution theory is introduced in information fusion process. Then Federated Particle Filter based on evlution (EFPF) algorithm is proposed that its obtaining of optimal estimation is viewed asthe result that “survival of the fittest in natural selection”.4. Finally, many groups of simulation experiments are designed to verify the validity of improvedEKPF algorithm and EFPF algorithm, respectively. Simulation data are from IGS ephemeris of GPSconstellation. The simulation results manifest their advantages of the above two algorithms atprecision of orbit determination and real-time.