低成本工業機器人自動校準及其在工業網宇實體系統之實現

Abstract

工業4.0智慧工廠的需求與工業機器人在市場上的廣泛使用, 工業網宇實體系統 (iCPS) 的發展及其擴展應用, 工業機器人的許多功能和應用都需要自我控制改進。 全自動化生產線低成本是全球製造系統的核心要求，自動校準也是其中的一部分。隨著物聯網的發展, 有大量的大型資料可以快速收集和儲存。人工智慧的發展使大資料的有效處理成為可能。 首先，針對工業的需求, 提出了一種在有限工作空間問題上解決機器人 TCP 運動絕對精度的標定方法。採用這種低成本的相機跟蹤方法, 其成本比一般用途的鐳射跟蹤裝置要低得多。採用基於深神經網路的整流線性單元方法, 可以比傳統的經驗補償法更有效地提高誤差偏移計算和標定精度。 其次，本文提出了一種比手工方法更快捷、更簡單、更便宜、更有效的刀具坐標系自動標定系統的獨立方法。所提出的方法需要使用兩個 ' Eye to hand ' 相機和一個 ' Eye in hand ' 相機捕捉圖像。然後, 通過 CamShift 和 MeanShift 演算法對圖像軌跡跟蹤和坐標系統轉換等方法獲取TCP資料, 並利用 PCA、LDA 等多種方式處理視覺資料。最優化的深神經網路方法誤差補償的機器人允許TCP自動運行與校準系統的功能。 我們已經開發了一個6度自由度的工業機器人這個實驗。建立了九種不同的 DNN 模型, 最後對現有機器人進行了適當的機器人座標誤差補償, 可自我調整、高效地實現機器人標定。此外, 還詳細介紹了選擇訓練資料的理由、DNN 模型的設計方法以及改進 DNN 模型的方案。 第三，針對全自動化生產線的工業要求, 提出了一種基於工業網宇實體系統(iCPS)的工業機器人在生產線應用的自動校準方法。對電機的 PID 控制參數進行分段, 以模擬機器在長時間內的校準精確度的狀況。利用九種深度神經網路 (DNN) 方法對大資料趨勢與機器校準評分的關係模型進行了訓練, 證明它比傳統的校準法更有效地提高了機器校準的速度和準確性。 關鍵字: 工業機器人、低成本校準工具、工業網宇實體系統

The requirement of Industry 4.0 for smart factories and the widespread use of industrial robots in market, the development of industrial cyber-physical systems (iCPS) and its extended application, many of the functions and applications of industrial robots will require self-control improvements. Low-Cost of completely automatic production line machine is the core requirement for global manufacturing system. Automated Calibration Approach of Industrial Robot is also part of the requirements. With the development of Internet of things (IoT), there are enormous big data of production line could be collected quickly and stored in large quantities. The development of artificial intelligence makes it possible to deal with big data efficiently. Firstly, Due to the need in the industry, this thesis presents a calibration method to resolve the absolute accuracy of robot TCP movement in a limited workspace problem. We implement this low-cost method with camera tracking which in a much lower cost fashion than general purpose laser tracking device. We use Rectified Linear Unit (ReLU) method based on deep neural network (DNN) which can be more efficient than traditional empirical compensation method in improving the error offset calculation and calibration accuracy. Secondly, this thesis presents a system independent method for automatic calibration of the tool coordinate system which is faster, simpler, cheaper and more effective than the manual method. The proposed method required images to be captured using two “eye to hand” cameras and one “eye in hand” camera. Tool center position (TCP) data is then acquired through CamShift and MeanShift algorithm for image trajectory tracking along with coordinate system conversion, several methods like PCA, LDA can deal with the vision data. Optimal Deep Neural Network (DNN) method error compensation of a robot allows TCP automatically run with the calibration system functions. We have developed 6 degrees of freedom (DoF) industrial robot for this experiment. Nine different kinds of DNN models are built and finally with suitable robot coordinate error compensation for the current robot; robot calibration can be achieved adaptively and efficiently. In addition, the justification for selecting training data, the methods for designing DNN models and the schemes for improving DNN models are also described in details. Thirdly, due to the industrial requirement of completely automatic production line, this thesis presents a method based on industrial Cyber Physical system (iCPS) to automated calibration of robot in production line application using the data produced by the robots. The PID control parameters of motors are segmented to simulate the calibration statue of the robots in a long duration. We use 9 kinds of deep neural network (DNN) methods to train the model of the relationship between the large data trend and the calibration score of the machine, it is demonstrated that it becomes more efficient than traditional calibration method to improve the speed and accuracy for robot calibration. Keywords: Industrial Robot, Low-Cost Calibration Tools, Industrial Cyber Physical System