【作者】 曾文静；

【导师】 徐玉如；

【作者基本信息】 哈尔滨工程大学 ， 船舶与海洋结构物设计制造， 2009， 硕士

【摘要】 导航技术是AUV实现自治的关键技术之一,高精度的导航和定位对其安全航行和高效率完成任务具有决定性的作用。同时定位与地图构建(SLAM, Simultaneous Localization and Mapping)利用所携带的外部传感器感知环境,并利用提取的信息同时进行水下地图构建和自身定位。由于水下环境极其复杂,可使用的外部传感器仅限于声呐、水下照相机等,且获得的观测信息噪声大、干扰多,所以对SLAM的数据关联提出了很高的要求。本文主要对水下机器人SLAM的数据关联算法进行了深入的研究和分析。论文介绍了AUV的导航方式以及SLAM对于水下定位的重要意义和发展现状；介绍了航行环境的描述方法以及SLAM的实现方法,分析了存在的技术难点；讨论了SLAM算法的性质,阐述了AUV基于扩展卡尔曼滤波的SLAM算法原理,建立了相关的仿真平台；重点研究了几种数据关联方法：最近邻算法、最大可能性算法、连续兼容最近邻算法和联合兼容算法,并提出了一种基于蚁群优化算法改进的最大可能性算法；结合各种数据关联方法,在逐渐增加量测噪声和过程噪声以及变化特征点间间隔的仿真场景中,进行了对比试验；并将某型AUV在水池中做直线运动获得的声呐图像数据融入SLAM仿真平台中,处理得出了相关的结果。试验结果表明：相对于单纯推位方法,SLAM可以提高系统的定位精度,也验证了算法在水下导航应用上的可行性,数据关联算法的优劣受到诸多因素的影响,如地图中特征间的间隔等。提出的新算法在保证实时性的同时,能够有效的提高关联正确率,具有一定的可行性。本论文所研究的工作,对于智能水下机器人自主导航的研究和SLAM技术的应用具有一定的参考意义。更多还原

【Abstract】 Navigation and localization with high precision is vital for the safety of the AUV and its effective completion of missions. Simultaneous Localization and Mapping (SLAM) algorithm allows the vehicle using on-board sensors to sense the environment and extract useful information to construct a feature map while locating itself through the map. However, the available sensors are limited within sonar and TV due to the complexity of underwater environment which brings much disturbance on obtained information. So there is a great demand on performance of data association. The research of this thesis emphasized on data association of SLAM.This paper gives a summary of current development of the worldwide SLAM research and its significance to AUV. The expressions of environment where an AUV will navigate are discussed. The details of SLAM principles realized by Extended Kalman Filter (EKF) are displayed. Relative simulative scenes with increasing measure noise and process noise and different feature separations are built up Four typical data association algorithms are elaborated and a new method based on Ant Colony Optimization is proposed. Comparison simulations for each association algorithm combined with different experiment scenes are researched and analyzed. What’s more, the data of sonar images from some AUV during certain tank test are used in the simulation.The results demonstrated that SLAM has a superior localization precision to pure dead reckoning and the feasibility of SLAM in underwater navigation is verified. The effect of association algorithm is influenced by various factors such as noise and feature separation. The presented algorithm proves to be excellent with high correct association rate and good real-time. The research will be favorable for the application of SLAM and AUV navigation.更多还原