Vision-Based Imitation Learning of Needle Reaching Skill for Robotic Precision Manipulation
- PDF / 3,140,487 Bytes
- 13 Pages / 595.224 x 790.955 pts Page_size
- 45 Downloads / 168 Views
Vision-Based Imitation Learning of Needle Reaching Skill for Robotic Precision Manipulation Ying Li1,2
· Fangbo Qin1,2 · Shaofeng Du3 · De Xu1,2 · Jianqiang Zhang3
Received: 7 April 2020 / Accepted: 5 November 2020 © Springer Nature B.V. 2020
Abstract In this paper, an imitation learning approach of vision guided reaching skill is proposed for robotic precision manipulation, which enables the robot to adapt its end-effector’s nonlinear motion with the awareness of collision-avoidance. The reaching skill model firstly uses the raw images of objects as inputs, and generates the incremental motion command to guide the lower-level vision-based controller. The needle’s tip is detected in image space and the obstacle region is extracted by image segmentation. A neighborhood-sampling method is designed for needle component collision perception, which includes a neural networks based attention module. The neural network based policy module infers the desired motion in the image space according to the neighborhood-sampling result, goal and current positions of the needle’s tip. A refinement module is developed to further improve the performance of the policy module. In three dimensional (3D) manipulation tasks, typically two cameras are used for image-based vision control. Therefore, considering the epipolar constraint, the relative movements in two cameras’ views are refined by optimization. Experimental are conducted to validate the effectiveness of the proposed methods. Keywords Imitation learning · Skill learning · Visual control · Robotic precision manipulation · Neural networks
1 Introduction Precision manipulation and assembly have been widely used in biotechnology, micro-electromechanical system (MEMS), and medical science, which address the manipulation problems of objects with size ranging from tens of microns to several millimeters [1–3]. There are four classical actions in robotic assembly: reaching, grasp, alignment and insertion. Reaching action means that one component is moved to near another component or a goal position, which is usually conducted before precise alignment and glue dispensing tasks. The adaptive motion considering collision avoidance is important for reaching. In [4], the components were manually moved to an appropriate position before pose alignment tasks. In glue dispensing tasks, the reaching trajectories of needle tip should be smooth and collisionfree. In [5], the needle should be moved to align with
Ying Li
[email protected]
Extended author information available on the last page of the article.
each of 16 holes on the upper surface of a cylindrical component. Then the glue is dispensed to fix the cylindrical component with other components. At first, the needle should be moved to the upper space from the side space of the cylindrical component for the following needle alignment process. And this process was conducted manually in the preparation stage. However, manual reaching operation limits the automation level of the assembly system. Besides, the perception of environmen
Data Loading...
