Research On Key Technologies Of Aeronautic GNSS/SINS Integration Precision Navigation

The precision and reliability requirement to navigation system in airplane approach step is much more strict than that in other flying steps. The navigation system in approach step should have good real time capability, good reliability and high precision. This requirement is hard for any singal navigation system. The GNSS/SINS integrated navigation is an outstanding resolution because of the complementarity of SINS and GNSS. This integrated navigation with small cubage and weight is applicable for all phases of flying, especially for the phase of approaching and landing.The high precision GNSS/SINS integrated navigation systems generally have poor reliability and real-time performances. GNSS carrier-phase relatively positioning have high precision, but the signal is easy to be interfered, then the integer ambiguities should be resolved. Quick resolution of the integer ambiguities is important for the real-time and reliability performances. The strapdown inertial navigation system should have the ability of initial alignment on the fly. The first goal of this research is to improve reliability of the integrated navigation system by improve the ability of resolving ambiguity and initial alignment on the fly. The stochastic errors of the inertial sensors are one kind of the most crucial errors and the observability is a important factor for the precision of GNSS/SINS integrated navigation systems, so the second goal of this research is to reduce the impact of this two factors.To accomplish these two goals, the following work has been finished in this research:(1)A simple method is introduced to determine the integer ambiguities searching space by analyzing the covariance matrix of float ambiguities. Float ambiguities can furthest satisfy the observation system of equations. It is proved by experimentation that integer ambiguities are almost near float ambiguities if there are not any gross errors. Firstly the primary integer ambiguities searching space was determined by variance of elements of ambiguities. Then an adjustment coefficient was determined according to the covariance. The adjustment coefficient was used to determine the integer ambiguities searching space further. The experiment indicates that the searching space of this method is less than that of Cholesky method, and the two methods have the same success rate almostly.(2) A method of normalized quadrature transform degree of observability analysis used in GNSS/INS integrated navigation system was presented. Firstly the least squares normal equation used to calculate the error states of integrated navigation system was estabilshed according to GNSS measurement information. Then the normal equation coefficient matrix diagonal elements were normalized. And the eigenvalues are the degree of observability of error states after normalized quadrature. And then a degree reduction method was presented. The observable error states were uncorrelated with the unobservable ones. The effects of unobservable error states are considered to be much more less than that of the observable ones. So the unobservable error states can be estimated by extrapolating with history data. So the model degree was reduced.(3)A method of airborne strapdow integrated navigation system coarse alignment in flight with single antenna was presented. This method used the information of location and velocity of GNSS and single epoch visual acceleration of SINS. The attitude quaternion was calculated with the constraint of 0 sideslip velocity. The factors affecting the alignment performance were analyzed. This method can reduce the request of the flying pose and can be carried out in wide dynamic range with high accuracy, high speed and low computer complexity.(4)An improving method of modeling the random errors with ARMA model was researched. Fast Fourier transform was used to stabilize the random error. The truncate property and tailing property was quantitatively identified. The parameters were determend by statistic method. The efficiency of parameters identification is increased by modified recursion least square and the validity of the model is tested. The experimentation showed that the random errors were noise whitened, so that the filtering precision can be improved.