| Abstract: | 工業機械手臂具有全天候工作，維護成本低，生產效率高等優點，已經被廣泛的應用於人類的工廠自動化生產，它在提高企業的生產力同時也可以降低生產所需成本，但如何精確的控制機械手臂追蹤期望軌跡，是目前機械手臂控制過程中不可被忽視的問題。為了保證追蹤過程中的精確度和穩定性，一個强大的控制器是不可或缺的，同時在控制器的設計過程中，如果能夠得知機械手臂的精確模型，就可以降低控制器的設計難度。因此，如何建立一個精確的機械手臂模型也是極需解決的重要議題。在本論文中，將從模型和控制器兩個方面同時著手，來克服二軸機械手臂的追蹤精度和穩定性問題。首先，我們基於二軸機械手臂的標稱動態模型，綜合考慮各種干擾因素，包含模型不確定性、摩擦力、齒輪背隙等，建立出一個增強型的二軸機械手臂動態模型，可在減小控制器設計難度的同時也可以降低未知干擾的大小問題。其次，在所建立的增強型二軸機械臂動態模型的基礎上，將干擾因素分為兩部分考慮，負載部分使用適應性徑向基神經網路控制器處理，負載之外的其他干擾則由時間延遲估計器來估測，藉此減少對控制器參數的依賴，更進一步通過非奇異快速終端滑模控制器來調整輸出控制訊號，藉此來降低外部干擾可能產生的影響。最後，我們設計了兩種不同的模擬環境，並透過多種典型的控制方法比較，來證明此論文所提出的控制器之有效性和優越性。;Industrial robotic manipulator has the advantages of all-weather work, low maintenance costs, high production efficiency, has been widely used in the human factory automation production, it can improve the productivity of enterprises as well as reduce the cost of production required. However, how to precisely control the manipulator to track the desired trajectory is a crucial problem that cannot be ignored during the operation of the manipulator. In order to guarantee the accuracy and stability of the tracking process, a powerful controller is indispensable. Also, during the design process of the controller, if we can know the precise model of the robot arm, the difficulty of the controller design can be significantly reduced. Therefore, a precise model of the robotic manipulator is also an important issue to be solved. In this thesis, we will focus on both the model and the controller to overcome the tracking accuracy and stability problems of the dual-axis manipulator. Firstly, based on the nominal dynamic model of the dual-axis manipulator, we integrated various interference factors, including model uncertainties, frictions, backlash of gears, etc., to build an enhanced dynamic model of the dual-axis manipulator, which can reduce the size of unknown interferences while reducing the difficulty of controller design. Secondly, based on the developed enhanced dual-axis manipulator dynamic model, the interference factors is separated into two parts, the payload part is handled by an adaptive radial basis neural network (ARBFNN) controller, and the other interferences except the payload is estimated by the time delay estimator (TDE) to reduce the dependence on the controller parameters, and furthermore, the control signal is output by the non-singular fast terminal sliding mode controller (NFTSMC) to reduce the possible effect of external interference. Finally, we design two different simulations and demonstrate the effectiveness and superiority of the controller proposed in this paper by comparing it with various traditional control methods. |
