隨著機械手臂軸數的不斷增加，近年來學術界對於透過數據驅動方法能夠更有效地應對逆運動學中的冗餘度問題而提升關注度。為了使模型獲得充足地訓練，數據驅動方法需要快速地收集大量數據，通常透過結構化和隨機生成的方式實現。然而，這兩種生成方式均在關節空間均勻取樣後，經由正向運動學將關節配置映射到任務空間的位置卻是不均勻的，導致在機械手臂應用中，任務空間的精確度不足。為了改善這個問題，我們提出了一種整合原有方法的資料生成方式。首先，我們透過結構生成的方法系統地生成部分數據，接著，透過隨機生成出各軸的值，再透過正向運動學轉換到任務空間中。在這個過程中，我們對新生成的資料位置與原本資料內的所有點進行比較，若其附近缺乏數據點，即進行補充，以填補原始數據量不足的。通過實驗結果顯示，在增加40%數據量前，我們的方法與原方法比較後，模型誤差更小。同時在數據集均勻性提高了12%。

With the increase in the number of axes in robotic arms, there has been a growing interest in the data-driven approach to solving inverse kinematics under redundant degrees-of-freedom. The data-driven approach requires the collection of training data. For robotic arms, training data are typically collected through structured generation or random generation under uniform sampling in the joint space. However, when mapped to the task space, the uniformly sampled training data may result in a non-uniform distribution, leading to insufficient precision in the task space. In this thesis, an integrated data generation approach that focuses especially on regions in the task space with insufficient representations is proposed.Experimental results indicate that our method yields smaller model errors compared to the baseline methods and exhibits a 12% improvement in dataset uniformity.

[1] C.-K. Ho, L.-W. Chan, C.-T. King, and T.-Y. Yen, “A deep learning approach to navigating the joint solution space of redundant inverse kinematics and its applications to numerical ik computations,” IEEE Access, vol. 11, pp. 2274–2290, 2023.
[2] T. Asfour and R. Dillmann, “Human-like motion of a humanoid robot arm based on a closed-form solution of the inverse kinematics problem,” in Pro-ceedings 2003 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2003)(Cat. No. 03CH37453), vol. 2, pp. 1407–1412, IEEE, 2003.
[3] I. Zaplana and L. Basanez, “A novel closed-form solution for the inverse kine-matics of redundant manipulators through workspace analysis,” Mechanism and machine theory, vol. 121, pp. 829–843, 2018.
[4] W. A. Wolovich and H. Elliott, “A computational technique for inverse kine-matics,” in The 23rd IEEE Conference on Decision and Control, pp. 1359–1363, IEEE, 1984.
[5] R. Fletcher, Practical methods of optimization. John Wiley & Sons, 2000.
[6] B. Kenwright, “Inverse kinematics–cyclic coordinate descent (ccd),” Journal of Graphics Tools, vol. 16, no. 4, pp. 177–217, 2012.
[7] D. DeMers and K. Kreutz-Delgado, “Learning global direct inverse kinemat-ics,” Advances in Neural Information Processing Systems, vol. 4, 1991.
[8] B. Ames, J. Morgan, and G. Konidaris, “Ikflow: Generating diverse inverse kinematics solutions,” IEEE Robotics and Automation Letters, vol. 7, no. 3, pp. 7177–7184, 2022.
[9] O. Limoyo, F. Mari´c, M. Giamou, P. Alexson, I. Petrovi´c, and J. Kelly, “One network, many robot: Generative graphical inverse kinematics,” arXiv preprint arXiv:2209.08812, 2022.
[10] A. Csiszar, J. Eilers, and A. Verl, “On solving the inverse kinematics problem using neural networks,” in 2017 24th International Conference on Mechatron-ics and Machine Vision in Practice (M2VIP), pp. 1–6, IEEE, 2017.
[11] J. Demby’s, Y. Gao, and G. N. DeSouza, “A study on solving the inverse kinematics of serial robots using artificial neural network and fuzzy neural network,” in 2019 IEEE international conference on fuzzy systems (FUZZ-IEEE), pp. 1–6, IEEE, 2019.
[12] G. Fang, Y. Tian, Z.-X. Yang, J. M. Geraedts, and C. C. Wang, “Efficient jacobian-based inverse kinematics with sim-to-real transfer of soft robots by learning,” IEEE/ASME Transactions on Mechatronics, vol. 27, no. 6, pp. 5296–5306, 2022.
[13] F. Aggogeri, N. Pellegrini, C. Taesi, and F. L. Tagliani, “Inverse kinematic solver based on machine learning sequential procedure for robotic applica-tions,” in Journal of Physics: Conference Series, vol. 2234, p. 012007, IOP Publishing, 2022.
[14] R. Bensadoun, S. Gur, N. Blau, and L. Wolf, “Neural inverse kinematic,” in International Conference on Machine Learning, pp. 1787–1797, PMLR, 2022.
[15] V. Kramar, O. Kramar, and A. Kabanov, “An artificial neural network ap-proach for solving inverse kinematics problem for an anthropomorphic ma-nipulator of robot sar-401,” Machines, vol. 10, no. 4, p. 241, 2022.
[16] J. Lu, T. Zou, and X. Jiang, “A neural network based approach to inverse kine-matics problem for general six-axis robots,” Sensors, vol. 22, no. 22, p. 8909, 2022.
[17] S. Park, M. Schwartz, and J. Park, “Nodeik: Solving inverse kinematics with neural ordinary differential equations for path planning,” in 2022 22nd Inter-national Conference on Control, Automation and Systems (ICCAS), pp. 944–949, IEEE, 2022.
[18] V. B. Semwal, M. Reddy, and A. Narad, “Comparative study of inverse kine-matics using data driven and fabrik approach,” in Advances in Robotics-5th International Conference of The Robotics Society, pp. 1–6, 2021.
[19] V. B. Semwal and Y. Gupta, “Performance analysis of data-driven techniques for solving inverse kinematics problems,” in Intelligent Systems and Appli-cations: Proceedings of the 2021 Intelligent Systems Conference (IntelliSys) Volume 1, pp. 85–99, Springer, 2022.
[20] P. Manceron, “Ikpy-an inverse kinematics library aiming performance and modularity,” 2021.
[21] J. Liang, K. W. Ng, and G. Tian, “A class of uniform tests for goodness-of-fit of the multivariate l p l p-norm spherical distributions and the l p lp-norm symmetric distributions,” Annals of the Institute of Statistical Mathematics, vol. 71, pp. 137–162, 2019.
[22] B. D. Ripley, “Modelling spatial patterns,” Journal of the Royal Statistical Society: Series B (Methodological), vol. 39, no. 2, pp. 172–192, 1977.