模擬人類動作訓練關節式機械手臂路徑控制

Machine Learning of Articulated Robot Trajectory Control based upon Human-like Learning Process

謝騰峰(Teng-Feng Hsieh)

指導教授 ： 顏家鈺

國立臺灣大學/工學院/機械工程學研究所/碩士(2022年)

https://doi.org/10.6342/NTU202202172

摘要

本論文目標是將人工智能與機器人結合，提出一個方法利用深度強化學習讓六自由度機械手臂能進行指定任務。藉由設計神經網路與獎勵函數，讓智能體能輸出關節扭矩來實現期望末端點位置與姿態，其訓練演算法分別使用深度確定性策略梯度與雙延遲深度確定性策略梯度，同時進一步分析機械手臂在不同情況下的訓練誤差，包含關節角度限制、軌跡、演算法與神經網路架構。最後從中找出最佳控制器，使手臂完成精密控制。

關鍵字

機器學習 ； 深度強化學習 ； 神經網路 ； 軌跡規劃 ； 關節式機器人

並列摘要

The objective of the thesis is to combine artificial intelligence with the robot. The presented method applies deep reinforcement learning to 6-DoF manipulator to perform the specific task. By designing the neural network and reward function, the joint torque produced by an agent can move the end-effector to the desired position and orientation during the process. Its training algorithm implements deep deterministic policy gradient (DDPG) and twin delayed DDPG (TD3) respectively. Furthermore, the training error of manipulator is analyzed in various situations, such as joint angle constraints, trajectories, algorithms and neural network architectures. Ultimately, the optimal controller is obtained to achieve precise control over robot arm.

並列關鍵字

Machine Learning ； Deep Reinforcement Learning ； Neural Network ； Trajectory Planning ； Articulated Robot

參考文獻

Y. Li, "Deep reinforcement learning: An overview," arXiv preprint arXiv:1701.07274, 2017.

D. Silver et al., "Mastering the game of Go with deep neural networks and tree search," Nature, vol. 529, no. 7587, pp. 484-489, 2016/01/01 2016, doi: 10.1038/nature16961.

B. R. Kiran et al., "Deep reinforcement learning for autonomous driving: A survey," IEEE Transactions on Intelligent Transportation Systems, 2021.

N. Justesen, P. Bontrager, J. Togelius, and S. Risi, "Deep learning for video game playing," IEEE Transactions on Games, vol. 12, no. 1, pp. 1-20, 2019.

T. Johannink et al., "Residual reinforcement learning for robot control," in 2019 International Conference on Robotics and Automation (ICRA), 2019: IEEE, pp. 6023-6029.