Research On Key Technology Of Autonomous Navigation Of Underwater Vehicle

With the competition of the ocean development becoming increasingly fierceramong countries of the world, the scientific researchers have shown more and moreinterest to the underwater vehicle. Because of the marine environment is extremelycomplex, autonomous underwater vehicle has more application advantages, soresearchers all over the world have launched a wide range of research and practiceabout AUV. Autonomous localization and path planning are the two key technologiesto realize the autonomous navigation function of AUV. This paper makes a deepdiscussion about both, analyzes problems and difficulties existing in current methods,then presents innovative and improved methods based on them.As for traditional algorithms of EKF-SLAM, the SLAM observation updateprocess has high computational complexity which isO (n2)even in the best case.Especially, it will increase dramatically when the number of obstacles graduallyincreases. When the system has many environment-observing sensors, equipmenterror and model error are inevitable, as well as clutter information. If clutterinformation is added to SLAM process, not only wrong obstacles are put into theenvironment map, but also error is brought to the SLAM filtering process.In addition, because the far detection distance of sonar and the slow speed ofAUV, obstacles in the same place will repeatedly be observed for many times, and therepeated-observation frequency will be high which will result in increment of thecomputational complexity of the SLAM observation update process. To solve theseproblems, this paper puts forward an improved method called improved SLAMalgorithm based on twice data association. In this method, firstly information ofobstacles collected by sensors is preprocessed which is called the first data association.Only if the number of successful first data association reaches the threshold times,those observation data are regarded as valid obstacles, and then the SLAM data association process which is called the second data association is conducted.Otherwise those observation data are abandoned as invalid obstacles. Through thismethod, clutter information can be effectively abandoned so as to improve theaccuracy of SLAM, at the same time, computational complexity of the SLAM processis reduced.Concerning the path planning technology, the traditional artificial potential fieldpath planning method and some improved methods have been studied in this paper,and some deficiency of them is summarized——that none of them can solve the localminimum point problem and the U-shaped obstacle problem successfully. About theseproblems, this paper proposes a new artificial potential field path planning methodbased on sector scanning and setting up temporary targets to solve local minimumpoint problem and U-shaped obstacle problem.About the improved SLAM algorithm based on twice data association andimproved artificial potential field path planning method, simulations and real AUVexperiments were conducted to verify them. The results show that the two improvedmethods are practicable and effective, the former can effectively filter out clutterobstacle information and lower the computational complexity, and the latter caneffectively solve the local minimum points and U-shaped obstacles.