應用自適應性類神經網路控制器於六軸機械手臂

Apply Adaptive Neural Network Controllers for a 6-DOF Robotic Arm

吳孟謙(Wu Meng-Chien)

指導教授 ： 陳美勇

國立臺灣師範大學/科技與工程學院/機電工程學系/碩士(2022年)

https://doi.org/10.6345/NTNU202201106

摘要

本論文提出了一種基於神經網絡框架學習機制的六軸機械臂控制器設計。首先，我們從六軸機械臂的實際構造中得到訓練數據集。其次，神經網絡的訓練方法是基於自適應調整輸入層和隱藏層之間的權重值和誤差。第三，將訓練數據集作為神經網絡的輸入來訓練模型。最後，我們利用李雅普諾夫理論保證了六軸機械臂控制器設計的穩定性，並與PI控制器設計進行了比較。
實現了六軸機械手臂動力學模型推導，以解決運動不穩定性問題。機械臂運動過程中時變不確定擾動引起的現象。詳細動力學模型是藉由Lagrange方程式所推導出來的，計算出六軸機械手臂動力學模型。透過動力學模型，進一步進行模擬驗證。
控制器是以PD為基礎進行設計的，結合自適應徑向基函數神經網絡 (RBFNN)，經由隱藏層與輸出層之間的自適應調整，最終取得所需的輸出結果，再藉由Lyapunov 函數進行穩定性分析，證明整個系統的穩定性，最後實驗分析此控制器對六軸機械手臂的控制穩定性。

關鍵字

類神經網路 ； 適應控制 ； 機械手臂 ； 動力學模型

並列摘要

This paper proposes a design of a six-axis manipulator controller based on the learning mechanism of the neural network framework. First, we get the training dataset from the actual construction of the six-axis manipulator. Second, the training method of the neural network is based on adaptively adjusting the weights and errors between the input layer and the hidden layer. Third, the training dataset is used as the input to the neural network to train the model. Finally, we use Lyapunov theory to ensure the stability of the six-axis manipulator controller design and compare it with the PD controller design.
The dynamic model derivation of the six-axis manipulator is implemented to solve the problem of motion instability. Phenomenon caused by time-varying uncertain disturbance during the movement of the manipulator. The detailed dynamic model is derived by the Lagrange equation, and the dynamic model of the six-axis manipulator is calculated. Through the dynamic model, further simulation verification is carried out.
The controller is designed on the basis of PD, combined with an adaptive radial basis function neural network (RBFNN), through the adaptive adjustment between the hidden layer and the output layer, and finally obtains the desired output result, and then uses the Lyapunov function The stability analysis is carried out to prove the stability of the whole system. Finally, the control stability of the controller to the six-axis robotic arm is analyzed experimentally.

並列關鍵字

Neural network ； adaptive control ； robotic arm ； dynamic model

參考文獻

[1] H. Yang and J. Liu, "An adaptive RBF neural network control method for a class of nonlinear systems, " IEEE/CAA Journal of Automatica Sinica, vol. 5, no. 2, pp. 457-462, Mar. 2018.

[2] Chengxiang Liua, Zhijia Zhaoa, Guilin Wen "Adaptive neural network control with optimal number of hidden nodes for trajectory tracking of robot manipulators, " Elsevier Neurocomputing, Vol 350, Pages 136-145, Jul. 2019.

[3] Z. Tang, M. Yang and Z. Pei, "Self-Adaptive PID Control Strategy Based on RBF Neural Network for Robot Manipulator," 2010 First International Conference on Pervasive Computing, Signal Processing and Applications, 2010, pp. 932-935, doi: 10.1109/PCSPA.2010.230.

[4] X. Huang and Q. Ning, "Active Disturbance Rejection Control Based on Radial Basis Function Neural Network," 2018 IEEE 3rd Advanced Information Technology, Electronic and Automation Control Conference (IAEAC), 2018, pp. 2397-2400, doi: 10.1109/IAEAC.2018.8577937.

[5] Q. Liu, D. Li, S. S. Ge and Z. Ouyang, "Adaptive Feedforward Neural Network Control With an Optimized Hidden Node Distribution," IEEE Transactions on Artificial Intelligence, vol. 2, no. 1, pp. 71-82, Feb. 2021, doi: 10.1109/TAI.2021.3074106.