Development of Automated Cell Manipulation System

在畜牧业、医学和食品供应领域，对用于克隆和人工授精的生物细胞的操作的需求日益增加。然而，大多数商业化的工具是手动或半自动的。操作尺寸为10 - 100μm的微小物体需要操作员经历长时间的培训，以便在显微镜的帮助下执行任务，例如保持，穿透，注射和抽吸。操作繁琐且耗时，但产率非常低。本研究的目的是开发一个高性能和可重复性的自动操作系统的生物应用。在本文中，视觉控制方案已被提出并实施，以指导运动的工具臂。目标和工具的位置，移动的路径和速度通过使用具有CCD（电荷耦合器件）图像装置的光学显微镜在像素坐标处计算。这种视觉控制方案能够消除机械误差，包括错位，滞后和漂移，通常存在于机械阶段，但在精细运动控制中至关重要。像素坐标也通常由多个工具臂和目标对象共享，并且对于实现平滑的协作操作是必不可少的。结果表明，在跟踪直径为3μm的小循环的试验中，刀具轨迹在二维空间内具有很高的精度和重复性。在力/触觉传感器的辅助下，该系统还能够执行3D操作。

There are increasing demands of manipulation on bio-cells for cloning and artificial insemination in the fields of animal husbandry, medical science and food supply. However, most commercialized tools are manual or semiautomatic. Manipulation of tiny objects sized at 10〜100μm requires the operators to experience a long time training to implement the tasks, such as holding, penetration, injection and suction, with the aid of a microscope. The operations are tedious and time consuming, yet with very low yield rate. The objective of this research is to develop an automated manipulation system with high performance and repeatability for the biological applications. In this paper, a vision control scheme has been proposed and implemented to steer the movements of the tool arms. The positions of the targets and tools, the paths and speed of the movement are calculated at the pixel coordinate by use of an optical microscope with a CCD (charge-coupled device) image device. This vision control scheme is able to eliminate mechanical inaccuracy including misalignment, hysteresis and drift which normally exist at mechanical stage but are crucial in fine movement control. The pixel coordinate is also commonly shared by multiple tool arms and targeted objects, and essential to achieve smooth cooperative operations. As the result, the great accuracy and repeatability of tool path have been achieved in 2D at a test of following a cycle as small as 3μm in diameter. With an assistance of force/tactile sensor for sensing at third dimension, the system is also able to perform 3D manipulation.