【作者】 姜楠；

【导师】 庄严；

【作者基本信息】 大连理工大学 ， 控制理论与控制工程， 2013， 硕士

【摘要】 地面自主移动机器人具有广泛的应用价值,其可在条件恶劣的场景中执行探索救援等多种任务。但由于其自身工作空间的限制,地面机器人只能获取自身周围的局部环境信息。四旋翼飞行器具有良好的灵活性和机动性,非常适合在近地面环境开展监视、侦察等任务。与地面移动机器人相比,它可以在特定高度上对地面环境进行大范围的观测,但是由于其续航能力有限,无法完成长时间的飞行任务。本文以自主研发的空地多机器人协作系统为基础,开展多机器人面向室内场景的空地协作环境感知研究。为完成地面三维环境的认知,本文提出了一种基于三维激光测距的场景认知方法,从而确保移动机器人在存在动态干扰(如运动人体目标)的大范围室内环境中具有自主场景认知的能力。将三维激光点云转换为二维的方向角图,并在其中提取ORB (Oriented FAST and Rotated BRIEF)特征作为场景的局部特征,同时结合从原始三维点云中提取的全局空间特征,从而完成自主场景认知的任务。利用包含30个房间(约1600m2)的大规模室内场景作为测试环境,采用配备有自主研发的三维激光测距系统的SmartROB2移动机器人平台进行实验验证。实验结果表明本文所提出的方法的有效性和对动态环境的适应性。为扩展地面移动机器人的感知范围,采用四旋翼飞行器与其协作来完成大范围、多视角的环境感知。通过对四旋翼飞行器进行硬件改造(配备视觉系统、机载计算机以及通信模块)和软件系统的构建,使其具有基于视觉的实时环境观测和远程监控能力。在此基础上,提出了飞行器针对地面起落标志的检测方法,并将该算法与压缩跟踪算法有效地结合起来,完成了地面起落标志的实时检测与跟踪,并能有效克服相似标志的出现和标志的部分遮挡带来的干扰。采用图像矩和圆形形变理论对起落标志进行分析,完成了四旋翼飞行器六自由度相对位姿的有效估计。一系列的标志干扰实验验证了四旋翼飞行器对地面标志检测与跟踪方法的有效性与实用性。此外进行了四旋翼飞行器与地面移动机器人之间的相对位姿估计实验。实验表明所提出的相对位姿估计算法可以满足目前应用的需求。更多还原

【Abstract】 Ground mobile robots have a wide range of applications, which can implement many tasks in hazardous environments. Local information is acquired because of the limited operation space. The capability to fly flexibility makes quadrotor UAV very suitable for surveillance and other tasks in the near-ground environment. Compared to ground mobile robots, a wide-range scene observation can be obtained at a specified height. They suffer from a limitation related to endurance and mission range. This paper conducts a study on collaborative indoor scene perception with a self-built air-ground multi-robot system.To accomplish3-D scene recognition, this paper presents a3-D-laser-based place recognition method to ensure robots have the ability to recognize the scene in dynamic environment. To complete scene recognition, ORB (Oriented FAST and Rotated BRIEF) is extracted as local features from bearing-angle images generated by a3-D laser data, and global feature is acquired from raw3-D point cloud data. A large-scale environment which owns30rooms (about1600m2) is regarded as testing site, and SmartROB2with a self-built3-D laser, is deployed for implementing experiments. Experimental results show that the approach is effective in dynamic indoor environment.Collaborative perception by combining ground mobile robot with quadrotor UAV is proposed to expand the perception range. Reforming hardware (equipped with vision system, on-board computer and communication module) and designing software for quadroor UAV has been accomplished, which make it have the ability of real-time visual perception and remote monitoring. A landmark recognition approach for quadrotor UAV is proposed, which is integrated with compressive tracking algorithm, to solve the issue with the disturbances of similar landmarks and partial occlusion. Image moments and circular deformation analysis are used for estimating relative6degree of freedom of quadrotor UAV. Landmark interference experiments demonstrate the detection and tracking method is able to recognize the landmark effectively. Besides, pose estimation experiments are conducted to confirm the algorithm of6degree of freedom estimation to meet the application requirements.更多还原