立體視覺擲球機械手臂之研究

A Study on a Ball-Pitching Robotic Arm Based on Stereo Vision

林駿青(Jun-Qing Lin)

指導教授 ： 陳冠宇

中原大學/工學院/機械工程研究所/碩士(2012年)

https://doi.org/10.6840/cycu201200751

摘要

本文的研究目的在改善基於立體視覺擲球機械手臂的投擲命中率。此機械手臂具有4個自由度及4指機械手掌，能夠抓球投擲。投擲目標固定於XZ雙軸機械平台上，能在空間中任意變換位置，其與機械手臂之相對位置可由一組雙攝影機組成之立體視覺系統計算而得。首先，本文從改善機械手臂振動過大的問題著手，並利用XZ機械平台以精準設定投擲目標，盡可能降低硬體造成的誤差。其次，本文在三維空間中設定投擲目標的網格點，依序移動投擲目標，藉由立體視覺系統估算其空間位置，再代入逆向運動學轉換求得機械手臂各軸的控制參數，進行數次投擲，然後利用立體視覺系統計算球的平均落點並記錄之。此後，任意變換投擲目標時，即可利用前述蒐集之所有網格參數進行雲形線內插，可提高投擲精確度。然而攝影機的筒狀扭曲現象，會造成某些位置的命中率偏低，本文針對此問題最後進行控制參數的補償。經過上述程序後，本文完成的實驗結果顯示，在距離機械手臂前方150至180 cm、截面範圍20 cm×20 cm之三維空間中，平均誤差為6 cm，將擲球命中率提高至60%以上。

關鍵字

雲形線內插 ； 立體視覺 ； 逆向運動學 ； 擲球機械手臂

並列摘要

The purpose of this study is to increase the hit rate of a ball-pitching robotic arm based on stereo vision. The robotic arm has four degrees of freedom and a four fingers robot hand designed for grabbing and throwing a ball. The target is fixed on an XZ two-axis motorized mechanical stage, so the position of the target can be changed arbitrarily. The robotic arm is equipped with a pair of CCD cameras to calculate the relative position between the target and cameras in three-dimensional space. The work in this study is firstly to reduce the vibration phenomenon of the robotic arm and to accurately set the target for ball throwing using an XZ two-axis motorized mechanical stage. The objective of the first step is to minimize errors caused by hardware. Secondly, we set some grid points in three-dimensional space as pitching targets and move these grid points in sequence. When the target location is calculated by the stereo vision system and converted to control parameters of the robotic arm by inverse dynamics, the robotic arm can throw a ball toward the target. After the robotic arm throwing several times, we can estimate these hitting locations by using the stereo vision system and then to record the average value. Therefore, as the pitching target is moved arbitrarily, we can use these previously collected grid parameters for spline interpolation to improve throwing accuracy. Due to the presence of camera barrel distortion, the hit rate is low in some positions. So we finally do a compensation of these control parameters to solve this problem. After the above procedures, the experimental results in this study show that the hit rate is increased to about 60% and average error is 6 cm from the front of the robot 150 to 180 cm and cross-section range 20 cm × 20 cm of three-dimensional space.

並列關鍵字

cubic spline interpolation ； inverse dynamics. ； stereo vision ； ball-pitching robotic arm

參考文獻

[24] 謝奇均，具遠端控制服務型移動機器人之研究，中原大學機械工程學系碩士學位論文，2011

[1] M. T. Mason and J. K. Salisbury, Robot Hands and the Mechanics of Manipulation, Cambridge, M.A.: The MIT Press, 1985.

[2] G. A. Bekey, H. Liu, R. Tomovic, and W. J. Karplus, “Knowledge-based control of grasping in robot hands using heuristics from human motor skills,” IEEE Transactions on Robotics and Automation, Vol. 9, No. 6, pp. 709-722, Dec. 1993.

[4] L. R. Lin and H. P. Huang, “Integration fuzzy control of the dexterous National Taiwan University (NTU) hand,” IEEE/ASME Transactions on Mechatronics, Vol. 1, No. 3, pp. 216-229, Sep. 1996.

[8] S. L. Werner, G. S. Fleisig, C. J. Dillman, and J. R. Andrews, “Biomechanics of the elbow during baseball pitching,” JOSPT, Vol. 17, No. 6, pp. 274-278, 1993.