Robust Image-Servo Control of Robotic Manipulators for Unstructured Environments

尽管机器人控制取得了重大进展，但机器人定位问题仍然需要在构建机器人细胞和机器人教学方面付出巨大努力才能完成新的任务。由此产生的高昂的工程成本阻碍了机器人成为多功能的可重新编程机器，从而快速适应新的任务。这项研究计划的“长期”目标是开发使用图像伺服(“IS”)的高性能控制方法，使机器人能够在非结构化环境中操作。尽管信息系统具有巨大的潜力，但缺乏健壮性已经阻碍了信息系统与许多现实世界应用的集成。另外，以前的工作主要集中在视觉伺服上，没有解决与其他成像传感器集成相关的许多问题。这项提案的短期目标集中在利用图像反馈开发“稳健”控制技术方面尚未解决的问题。除了视觉传感器外，重点还将放在来自超声/回波、计算机断层扫描和透视的图像上。该研究的主要特点包括：i)在IS环境下开发了一种健壮的特征表示方案；ii)开发了第一个全面的IS回路不确定性模型，该模型将被用于设计一种新的用于健壮的多区域运动规划和控制技术的混合框架；以及iii)扩展结果以解决新应用中的困难定位问题，例如连续介质机器人的控制。所提出的研究结果可以省去细胞结构和机器人示教工作，并可能放宽对机器人机构的机械精度和刚度的要求，降低其成本，扩大其应用范围。预计这项研究工作的结果将需要独特和系统的程序来设计健壮的操作系统。这类程序的可获得性可以通过改进当前任务的执行和促进以前被教导用经典技术几乎不可能实现的应用程序而产生重大和直接的影响。此外，这项研究将为建立新的编程范式奠定基础，在该范式中，目标将在整合的任务和图像空间中指定。

Despite significant advances in robot control, robot positioning problem still requires significant effort in structuring the robotic cells and robot teaching for new tasks. The resulting high engineering costs prevent robots from being versatile reprogrammable machines, adapting quickly to new tasks. The "long-term" objective of this research program is on the development of high-performance control methods using image servoing ("IS"), to enable robot operations in unstructured environments. Despite enormous potential of IS, lack of robustness has hindered the integration of IS into many real-world applications. Also the previous work has mainly focused on visual servoing, leaving many issues related to the integration of other imaging sensors unsolved. The short-term objectives of this proposal focus on the unresolved issues for the development of "robust" control techniques using image feedback. In addition to vision sensor, the focus will be on the images from ultrasound/echo, computed tomography, and fluoroscopy. Key features of the proposed research include: i) development of a robust feature presentation scheme in the IS context; ii) development of the first comprehensive IS loop uncertainty model that will be used to design a novel hybrid framework for robust multi-regime motion planning and control techniques; and iii) extending the results to solve difficult positioning problems in new applications such as the control of continuum robots. The proposed research results could eliminate cell structuring and robot teaching efforts, and would provide the potential to relax the mechanical accuracy and stiffness requirements for robotic mechanisms, reducing their cost and extending their applications. The outcome of this research work is expected to entail unique and systematic procedures for the design of robust manipulation systems. The availability of such procedures could have a significant and immediate impact by improving the execution of current tasks and by facilitating applications previously taught to be almost impossible with classical techniques. Also, this research will yield to the foundation of new programming paradigm, in which the goals will be specified in integrated task and image spaces.