Research Of The Install-calibration For USBL

This paper is based on the domestic high precise USBL (Ultra-Short Baseline) system toresearch the at-sea installation calibration and ray tracing position algorithm. Afterinstall-calibration, the incidence angle can be calculated to use the tracing positioningalgorithm; the ray-tracing positioning scatter picture can test if the calibration is correct,which is the ultimate goal of the system, therefore the two contents for USBL are inseparable.The core of install-calibration the measurement equation used, as well as the calibrationprocess decided by the equation property and engineering realization, which need the esurientthat the calibration has to have a certain robustness under normal conditions. The robustnessmeans: first the equation must be convergence to obtain a result; secondly single calibrationshould be trusted, which means the uniform and the uncertainty interval must be guaranteed;at last the calibration should be correct, which means the result should be unbiased as far aspossible. Therefore, this paper focuses on the existing observation equations and itsimprovement, which contains the interval uniqueness, the unbiasedness and the nature of thevariance, then the solution form at the least squares rules. At the first, tracks is an importantissue for whether the equation be convergence (interval uniqueness) and the uncertaintyinterval (the nature of the variance), which needs a certain scale and to be symmetry.Secondly the measurement of vertical displacement is needed to ensure the convergence ofequation, and the imprecisement do not affect other estimates under ideal conditions. Then themajor factor of the size of equation partial (estimated accuracy) is underwater acoustic linebending. Because the ray bending compensation can compensate the partial of the equation,the equation compensation do the same, then the practice is that at the first step to estimate theaverage sound velocity and at the second use the partial form. Finally, due to the nonlinearityof the equation and the sea circumstance, the specific numerical solution also influences theestimation precision and variance. While M estimation has better robustness and stability, thispaper presents a weight-form based on outlier eliminating algorithm, which does not need toconsider the specific criterion function but the weight of observation only. To sum up, theimproved observation equation, reasonable tracks and appropriate numerical solution canensure the stability of calibration basically.For ray tracing positioning, the positioning precision is mainly due to positioningalgorithm at the condition of reasonable calibration and measurement. Positioning projectalgorithm of tracking positioning is constant gradient or constant sound speed delimination, then the rationality of delimination will directly influence the accuracy of the algorithm.Apparently the denser gets the higher accuracy, but the criterion for USBL real-time or deeppositioning is not acceptable. In addition, influences of the measurement quality or changes ofthe sound speed profile are difficult to express, which is negative to engineeringimplementation. So the horizontal positioning error propagation model is established, whichis deduced by the error transfer formula and whose parameters are confirmed by simulation,and the concept of equivalent measurement error of sound velocity profile is given. Accordingthe model, an adaptive slicing algorithm is designed to control truncation error: it bases on theadaptive Simpson method with inversed process. And then the model has some guidingsignificance, such as profile processing before position, the equivalent difference. For the next,the velocity measurement method by temperature, salinity and pressure is studied as well asthe profile smoothing or stretching. At last formulas needed by USBL positioning and thebasic module are summarized, and the correctness is verified by the experimental data.At the end of the paper, statistical processing of experimental data shows the improvingof the track and the M estimation. Positioning after calibration using smaller circle data, allthe location have good convergence scatter include the bigger one, which means the scatterhas no radius adaptability and verifies the calibration and tracking accuracy.