Visual Servo Control for a Paddle Juggling Manipulator

• Main Authors: Po-Yi Kuo, 郭柏毅
• Other Authors: Chi-Cheng Cheng
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/09989712807295785486

碩士 === 國立中山大學 === 機械與機電工程學系研究所 === 104 === The main objective of this thesis is to focus on improvement and modification of the bouncing system developed in previous researches including “Intelligent Hand-Eye Coordination Control on Ball Bouncing” and “Hand-Eye Coordination Control on an Adaptive Ball Bouncing System”. A series of researches aims to develop human-like dexterous manipulation techniques to lay foundation for control of future intelligent robots. An Xtion PRO LIVE camera is applied to capture images with depth information at the same time to track a ping-pong ball as a target. The task of paddle juggling by bouncing the pong-pong ball is achieved by a robotic manipulator for realization of dexterous operational capability of humans. In order to enhance the position accuracy on the image plane, the HSV color space less affected by illumination is selected. Firstly, two-dimensional image information is obtained by the background separation technique. Incorporating with the depth information, the ball’s position in three-dimensional space can be determined. Then the ball’s trajectory is predicted based on dynamic equations of the ball. Therefore, the timing for the computer to give control command to the control board for the purpose of achieving the mission of paddle juggling can be calculated. By analyzing videos of bouncing ball experiments, the emphasis is on improvement on two key factors: the ball’s position in three-dimensional space and the reaction time of the bouncing mechanism. As a result, the bouncing mechanism is able to play the ball upwards and the ball will stay in a region to allow for another play to accomplish the goal of continuous bouncing. Through experiments and modification, accurate three-dimensional spatial information of the ping-pong ball can be acquired in real-time. According to prediction of the remaining fall time and the motor reaction time, the better instant to hit the ball can be computed. Experimental results demonstrate that the ping-pong ball reaches at twenty-five centimeters height in continuous bouncing condition and the average bouncing time is about four times.