【作者】 陈树娟；

【导师】 何波；

【作者基本信息】 中国海洋大学 ， 通信与信息系统， 2011， 硕士

【摘要】 自主式水下机器人AUV(Autonomous Underwater Vehicle)代表了未来水下机器人技术的研究方向,是目前研究工作的热点。而导航问题仍是AUV所面临的主要技术挑战之一,在导航问题中,定位问题又是移动机器人的基本问题,是指移动机器人通过携带的传感器完成对内部状态的检测或对外部环境的感知,从而估算其自身位置和姿态的过程。在未知的环境中,机器人的定位与构图是融为一体相互关联的,即同时定位和地图构建SLAM (Simultaneous Localization and Mapping)。问题可描述为:一个自主移动机器人从未知环境中的未知位置出发,在机器人航行过程中根据自身携带的传感器采集的数据实现自身位置的确定,同时增量式地构建全局环境地图。本文首先介绍了SLAM算法的基础即卡尔曼滤波器理论,然后阐述了在单纯声纳更新的SLAM算法基础上引入了其他传感器更新环节的SLAM算法,即多传感器更新的SLAM算法,详细介绍了其实现流程,并从理论上说明其能够提高了机器人定位和构图精度。接着介绍了试验的前期准备,即AUV平台携带的传感器以及SLAM算法中用于构建环境地图的特征的提取。提取精确可靠的环境特征是SLAM算法准确度的保障,而环境特征的表示则依据机器人航行的环境。然而直接提取的环境特征可能比较密集,影响了SLAM算法的效率和精度,因此需对特征点进行去噪声和稀疏化处理,噪声是针对声纳自身的噪声及其环境背景噪声,而稀疏化处理则是针对声纳发射的单个波束及其多个波束间的冗余信息的。最后通过湖试和海试试验验证了多传感器更新的SLAM算法的有效性。湖试试验的结果表明了该算法在机器人的定位和构图精度方面优于仅声纳更新的SLAM算法,而较长距离的海试试验进一步验证了该算法的有效性,且说明了该算法的定位和构图精度满足自主式水下机器人的航行要求。更多还原

【Abstract】 Autonomous underwater vehicle (AUV) is the current research focus, which represents the future development direction of underwater vehicle technology. However, navigation is still the one of the major technical challenges of AUV, in which localization is the fundamental problem, which refers to a process of estimating robot’s pose by detecting the internal state of robot and perceiving the external surrounding environment through carried sensors.Localization and mapping is integrated and interrelated in unknown environment, which is simultaneous localization and mapping (SLAM). SLAM can be described as: an autonomous mobile robot leaves from an unknown location in an unknown environment, relies on sensors to build a global environment map gradually, and then calculates its location by this map.This article first introduces the theory basis of SLAM algorithm, which is EKF, then add other sensors update to only sonar update SLAM, namely SLAM with multi-sensor update, illustrates its flow and certifies it can improve accuracy of localization and mapping for the robot from theory. Next, experimental preparation is introduced, which includes sensors carried in AUV platform and features extraction for map construction in SLAM. Extracting reliable and accurate environmental features is prerequisite for the precision of SLAM algorithm, while features representing is based on environment where the robot navigates. But the original features of direct extraction from environment are generally too dense to affect the efficiency and accuracy of SLAM algorithm, and then denoising and sparseness of features are adapted. Noise involves noise of sonar itself and environmental background noise, and sparseness aims in redundant information of each beam and adjacent beams of sonar launches.Finally, validity of SLAM with multi-sensor update is verified by lake and sea experiments. Lake experiment result proves that accuracy of localization and mapping of this algorithm is superior to SLAM only with sonar update, and longer sea trial experiment further verifies validity of SLAM algorithm. In addition, it also shows the accuracy of localization and mapping meets the requirements of autonomous navigation of robot.更多还原