The Research Of Zero Velocity Update For Laser Gyroscope Strap-down Inertial Navigation System

Zero velocity update (ZUPT) applied in laser gyro strap-down inertial navigation system(SINS) can improve the positioning accuracy of system without no additional cost. ZUPTutilizes speed output of SINS during the stopping intervals as observables to estimate andcorrect the positioning errors and attitude errors. Low-cost and effective, ZUPT is an essentialerror control method for SINS. The ZUPT method based on conic fitting and the ZUPTmethod based on kalman filtering are two frequently-used methods. The ZUPT method basedon conic fitting uses the speed output samples during the stopping points to calculate threeundetermined coefficients of the speed error curve, then estimates and compensates thepositioning errors to improve the positioning accuracy of SINS. The ZUPT method based onkalman filtering utilizes the speed output as observables and establishes the error equationsto estimate and compensate of the velocity errors, position errors and attitude errors, thusreducing various kinds of cumulative errors of SINS.The ZUPT method based on conic fitting is simple, but makes low use of the velocitysamples gained during the stops of SINS, and requires a short stopping interval. The ZUPTmethod based on kalman Filter is very sensitive to the kalman filter parameters, it takes verycarefully adjustment for the filter parameters, and the estimation accuracy would convergeto a good level after a few stops, which makes the performance of the ZUPT not as good forthe beginning period. Based on researching the principle, testing and verifying the effects ofthe two methods mentioned above, a new ZUPT method for SINS based on support vectormachine (SVM) regression curve fitting is proposed. Based on machine learning and theminimum error criterion, the ZUPT method based on SVM regression curve fitting employsthe most of cut-off velocity samples to fit the velocity error curve, then compensates andcorrects the positioning errors. Compared with the ZUPT method based on conic fitting, theZUPT method based on SVM regression curve fitting can make a more accurate velocityerror estimate even if the stopping interval is longer. The ZUPT method based on SVMregression curve fitting works well for the case that the stopping interval of SINS is longer.Identification of the stopping state are very important for ZUPT, therefore three zero-velocity detectors named acceleration moving variance detector,acceleration magnitude detector, and angular rate energy detector for detecting the stopping time intervals of SINSare designed and tested by simulation. Simulation results show that all three kinds ofdetectors can detect the stopping intervals accurately.