Integrated Visual/Force Servoing Control of Robotic Manufacturing Systems

本申请中提出的未来五年的工作建立在我以前的研究机器人制造系统的视觉伺服控制，旨在提高机器人定位的精度和速度。在机器人制造系统中，为了实现这一目标，已经开发了几种先进的控制策略。 然而，除了高精度定位之外，该行业还在寻找安全有效的控制策略，以允许机器人系统在非结构化环境中执行跟踪，定位和抓取任务等。 这样的控制策略可以保持机器人系统长时间运行，并且可以避免机器人碰撞发生时的任何损坏。目前，运营商不得不关闭生产，并试图修复损坏。对于高端制造系统，特别是航空航天工业，碰撞将导致模具损坏和生产时间的重大损失。 例如，在先进的纤维铺放（AFP）机中，纤维铺放头与芯轴之间的碰撞在修复纤维铺放头中的辊或其他部件时将花费高达数万美元。虽然一些被动的解决方案已经在工业中使用，例如在末端执行器和工具之间安装碰撞保护装置或设计良好的运动规划控制器，但工业界正在寻求集成视觉，力和位置信号并为机器人制造系统提供安全有效的控制策略的主动预防解决方案。 在机器人控制领域，视觉和力控制已经得到了大量的研究。然而，很少有方法专注于处理碰撞和碰撞在真实的工业制造过程中。因此，一个新的实用的基于多传感器的避碰控制框架是建立，以满足机器人制造业的需求。

The work proposed in this application for the next five years builds on my previous research on visual servoing control of robotic manufacturing systems which aims at increasing the accuracy and speed of robot positioning. Several advanced control strategies have been developed to achieve the goal in the robotic manufacturing systems. However, in addition to high precision positioning, the industry is also looking for safe and efficient control strategies that allow the robotic system performing tracking, positioning and grasping tasks etc. in an unstructured environment. Such control strategies can keep the robotic system up and running for the long haul and can avoid any damage when the robot crash occurred. Currently, the operator has to shut down production and try to fix the damages. For the high end manufacturing systems especially in aerospace industry, the crash will result in tooling damage and significant loss of the production time. For instance, in the advanced fiber placement (AFP) machine, the collision between the fiber place head and the mandrel will cost up to ten thousands dollar in repairing the roller or other parts in the fiber placement head. Although some passive solutions have been used in the industry such as installing a crash protection device between the end effector and the tool or designing a good motion planning controller, an active prevention solution that integrates the vision, force and position signals and provides the safe and efficient control strategies to the robotic manufacturing systems are sought by the industry. A lot of research has been devoted to vision and force control in the robotic control community. However, few approaches focus on dealing with the collision and crash in the real industrial manufacturing processes. Therefore, a new practical multi-sensor based control framework for collision avoidance is to be built to meet the demand from robotic manufacturing industry.