Research On Key Technology Of AUV Integrated Navigation System Based On SINS/DVL/GPS

As the exploration and development of ocean progresses, the demand of underwater vehicle with autonomous navigation capabilities is growing. In the complex environments of a deep ocean, single sensor device is unable to meet the requirements of high-prcision autonomous navigation. So the method of multi-sensor information fusion and improving the accuracy of filtering algorithms become an inevitable choice. Through research on key technology of AUV integrated navigation system, this thesis strongly guarantees a broader, more long-range, more complex work area, and then will significantly advance our country and all the world AUV navigation technology.This thesis combines the current AUV navigation technique, describes technical difficulties that autonomous navigation technology often encountered. We propose an"Underwater Navigation - water surface correction - Underwater Navigation"navigation mode, in which SINS/DVL integrated navigation system is used as the underwater navigation system and SINS/GPS integrated navigation system is used as the water surface correction system. In the long-range AUV navigation, AUV surfaces after underwater navigation in a period of time, and GPS is introduced for correcting the speed and location information, then the divergence problem of positioning errors accumulating over time in the long-range AUV is solved. Then we research the two popular filtering methods, such as EKF and NN. Analyze their strengths and weaknesses, and propose appropriate improvements.EKF is the optimizing algorithm of Kalman filter for non-linear environment. We design state equation and observation equation of EKF by SINS and DVL error equations, which can calculate the various errors. Because it is dicky that the real running environment of AUV integrated navigation system, the precise mathematical model of system and the transcendent characteristic of the yawp are difficult to get, and cause EKF to get divergence. Its advantage is higher accuracy and real-time, but to be worse in reliability.On the premise of ensuring the accuracy and real-time that the precise navigation of the AUV needs, this thesis proposes a filter algorithm of NN based on improved GA for the poor reliability of EKF. Signal processing with the characteristics of NN can eliminate processes of modeling and feature extraction, and thus reduce errors caused by uncertain model and improper characteristics selection. Real-time identification in order to improve the performance of the system can then be realized. GA is good at global searching. It can in a complex, multi-peak, non-linear and non-differentiable space to achieve global search. However, the process of the NN optimization, we found that conventional GA can be inadequate, such as prematurity, poor stability algorithm, convergence effectiveness caused by the crossover and mutation probability etc. To solve these problems, traditional GA is improved from encoding, selecting options, best individual reservatons, the introduction of"migration"mechanism, crossover and mutation operators in this thesis. NN is trained with the improved GA and applied to AUV integrated navigation system.In the"Underwater Navigation - water surface correction - Underwater Navigation"navigation mode, respectively, these two kinds of filtering algorithm are applied to AUV integrated navigation experiment. The experiment's results show that the improved algorithm can be good enough to achieve the desired objectives.The two kinds of filtering algorithms all improve the filtering accuracy. And NN based on improved GA on the premise of ensuring the accuracy and real-time is more reliable than EKF, so enhances the usefulness of the filter.