本文为澳大利亚埃迪斯科文大学(作者:Ahmed S. Khusheef)的硕士论文,共152页。
移动机器人可以用于搜索、发现和重新定位各种制造操作和环境中的物体。在这种情况下,目标物体可能以相同的概率驻留在环境中的任何位置,因此,机器人必须自主地导航和搜索整个区域,并配备特定的传感器来检测物体。新的挑战在于开发一个控制系统,帮助移动机器人完成这些任务,包括构建增强的导航系统和基于视觉的目标识别。后者对于执行需要最佳目标识别技术的探测任务非常重要。
在本论文中,对于室内环境来说,这些挑战被分成三个子问题。
首先,导航任务涉及到为整个环境找到一条合适的探测路径,从而对探测的要求最低。采用Bug算法策略对环境进行建模并实现探测路径。
第二个是视觉搜索过程,它包括采用适当的图像处理技术,并选择一个合适视角区域的相机。本文重点研究了颜色分割、模板匹配和快速鲁棒特征(SURF)在目标检测中的应用。
第三个问题是重新定位过程,这涉及到使用机器人的抓手抓住检测到的、所需的对象,然后将其移动到指定的、最终的位置。这还包括使用视觉跟踪技术接近目标和交货地点。所有代码都是用C++和C编程开发的,一些库包括OpenCV和OpenSURF,用于图像处理。分别对各控制系统的功能进行了测试,并将其作为一个整体控制程序进行组合。
系统性能评估使用两种类型的移动机器人:腿和轮式。在这项研究中,有必要开发一个具有高性能处理器的轮式搜索机器人。实验结果表明,只要处理器性能足够,搜索机器人所采用的方法是高效的。结论是,如果在机器人身上有效地定位和使用传感器,就有可能在最少数量的传感器内实现导航系统。视觉搜索过程中的主要挑战是环境条件难以控制,因为搜索机器人是在动态环境中执行任务。此外,还探讨了将这些小型机器人扩展到有用的工业能力方面的其他挑战。
Mobile robots could be used to search, find, and relocate objects in many types of manufacturing operations and environments. In this scenario, the target objects might reside with equal probability at any location in the environment and, therefore, the robot must navigate and search the whole area autonomously, and be equipped with specific sensors to detect objects. Novel challenges exist in developing a control system, which helps a mobile robot achieve such tasks, including constructing enhanced systems for navigation, and vision-based object recognition. The latter is important for undertaking the exploration task that requires an optimal object recognition technique. In this thesis, these challenges, for an indoor environment, were divided into three subproblems. In the first, the navigation task involved discovering an appropriate exploration path for the entire environment, with minimal sensing requirements. The Bug algorithm strategies were adapted for modelling the environment and implementing the exploration path. The second was a visual-search process, which consisted of employing appropriate image-processing techniques, and choosing a suitable viewpoint field for the camera. This study placed more emphasis on colour segmentation, template matching and Speeded-Up Robust Features (SURF) for object detection. The third problem was the relocating process, which involved using a robot’s gripper to grasp the detected, desired object and then move it to the assigned, final location. This also included approaching both the target and the delivery site, using a visual tracking technique. All codes were developed using C++ and C programming, and some libraries that included OpenCV and OpenSURF were utilized for image processing. Each control system function was tested both separately, and then in combination as a whole control program. The system performance was evaluated using two types of mobile robots: legged and wheeled. In this study, it was necessary to develop a wheeled search robot with a high performance processor. The experimental results demonstrated that the methodology used for the search robots was highly efficient provided the processor was adequate. It was concluded that it is possible to implement a navigation system within a minimum number of sensors if they are located and used effectively on the robot’s body. The main challenge within a visual-search process is that the environmental conditions are difficult to control, because the search robot executes its tasks in dynamic environments. The additional challenges of scaling these small robots up to useful industrial capabilities were also explored.
-
引言
- 项目背景与文献回顾
- HEXAPOD移动机器人
- 轮式机器人
- 探索路径与运动控制
- 基于视觉系统的物体检测
- 控制系统
- 结论与展望
更多精彩文章请关注公众号: