應用視圖之經驗轉移於機器手臂夾取隨機堆疊物件

Robot Bin Picking Using View-Based Experience Transfer

王仁蔚 ； 李佳蓮 ； 陳健倫 ； 王俞程 ； 鄭詠珈 ； 李志中

《機械工業雜誌》 461期 (2021/08) Pp. 39-51

摘要

本文提出了一種新的夾取演算流程，可供機器手臂在隨機堆疊的物件中夾取目標物。首先，本文使用實例切割和基於視圖的經驗轉移演算法來獲得候選夾取點；隨後，利用避免碰撞和可進行穩定性分析的演算法，篩選出最佳夾取點，以供機器手臂夾取。基於視圖的經驗轉移策略是先找到物體的姿態，接著再將夾取經驗轉移到切割出的圖像中以獲得夾取點。本文並建立一機器手臂系統進行評估，實驗裝置包含一台六軸機器手臂，一個二指平行夾爪和一台Kinect V2 RGB-D相機。實驗結果顯示，本文所提出的方法可以推廣到具有複雜形狀的物體，並且能夠在密集堆疊的箱中夾取不同類別的物體。

關鍵字

實例切割 ； 夾取經驗轉移 ； 堆疊夾取

並列摘要

In this study, a novel pipeline for grasp detection in robot bin picking was proposed. Firstly, a method combining instance segmentation and view-based experience transfer was developed. Subsequently, collision avoidance and stability analysis to determine the optimal grasp for robot grasping were performed. The approach for the view-based experience transfer was to acquire the object view and then transfer the grasp experience onto the clutter scenario. A robot system consisting of a six-axis robot arm with a two-jaw parallel gripper and a Kinect V2 RGB-D camera was established to validate the approach. The experimental results show that our proposed approach can be implemented to grasp objects with complex shape as well as grasp different types of objects in a bin.

並列關鍵字

Instance segmentation ； Grasp experience transfer ； Random bin picking

參考文獻

Bicchi, A.,Kumar, V.(2000).Robotic grasping and contact: A review.Proc. IEEE Int. Conf. on Robotics and Automation (ICRA).(Proc. IEEE Int. Conf. on Robotics and Automation (ICRA)).

Liu, M.-Y.,Tuzel, O.,Veeraraghavan, A.,Taguchi, Y.,Marks, T. K.,Chellappa, R.(2012).Fast object localization and pose estimation in heavy clutter for robotic bin picking.The Int. J. of Rob. Res..31(8),951-973.

Mahler, J.,Liang, J.,Niyaz, S.,Laskey, M.,Doan, R.,Liu, Xinyu,Aparicio Ojea, J.,Goldberg, K.(2017).Dex-Net 2.0: Deep Learning to Plan Robust Grasps with Synthetic Point Clouds and Analytic Grasp Metrics.Proc. Robotics: Science and Systems XIII (RSS).(Proc. Robotics: Science and Systems XIII (RSS)).

Pinto, L.,Gupta, A.(2016).Supersizing self-supervision: Learning to grasp from 50K tries and 700 robot hours.IEEE Int. Conf. on Robotics and Automation (ICRA).(IEEE Int. Conf. on Robotics and Automation (ICRA)).

Zhou, X.,Lan, X.,Zhang, H.,Tian, Z.,Zhang, Y.,Zheng, N.(2018).Fully convolutional grasp detection network with oriented anchor box.IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS).(IEEE/RSJ Int. Conf. on Intelligent Robots and Systems (IROS)).