面向复杂工艺的拟人化装配技巧研究

与机器人相比，人类在执行轴孔装配任务时具有更高的灵活性和效率，因此 本文主要研究了基于模仿学习和深度学习并利用现实世界人类示教数据实现机器 人轴孔装配任务的综合方法。 本论文分析了轴孔装配中关键的姿态调整阶段，并建立了一个高斯混合模型 (GMM)，利用人类示教数据构建控制变量和状态变量之间的非线性映射关系，并 通过期望最大化（EM）算法求解模型最优参数，最终应用到真是机械臂操作平台 使得机器人能够实现拟人化的柔性装配过程。 本研究引入了一种基于深度学习的数据驱动控制方法，首先收集轴孔装配过 程中双向插入和拉拔数据，构建了一种由宽度学习系统（Broad Learning System, BLS）和双向长短期记忆（Bidirectional Long Short-Term Memory, Bi-LSTM）组成， 我们称之为 BLS-biLSTM 的深度学习模型。该模型将 BLS 的强大特征提取和增强 能力与 Bi-LSTM 的深度序列建模能力相结合，从复杂人类示教数据中实现了全面 的装配特征提取和装配轨迹预测。 本文搭建了虚拟-现实仿真与实验平台，通过虚拟到现实的迁移方法并使用真 实机械臂臂操作平台开展了应用实验，充分验证了所提两种创新型方法的有效性。 结果表明，我们的方法能够提高机器人轴孔装配任务的效率和适应性。总的来说， 本文通过学习人类并研究先进的模仿学习和深度学习方法，推动了装配任务的机 器人学习策略的发展。

[1] 黄群慧, 贺俊. 中国制造业的核心能力、功能定位与发展战略——兼评《中国制造 2025》[Z]. 2015: 13.
[2] 毕胜. 国内外工业机器人的发展现状[J]. 机械工程师, 2008(7): 4.
[3] NAGARAJAN P, SARAVANA PERUMAAL S, YOGAMEENA B. Vision Based Pose Estimation of Multiple Peg-in-Hole for Robotic Assembly[M]//Computer Vision, Graphics, andImage Processing. Cham: Springer International Publishing, 2017: 50-62.
[4] LIU S, LIU C, LIU Z, et al. Laser tracker-based control for peg-in-hole assembly robot[J/OL].The 4th Annual IEEE International Conference on Cyber Technology in Automation, Controland Intelligent, 2014: 569-573. https://api.semanticscholar.org/CorpusID:18849643.
[5] LIU Z, XIE Y, XU J, et al. Laser tracker based robotic assembly system for large scale peg-holeparts[M/OL]//The 4th Annual IEEE International Conference on Cyber Technology in Automation, Control and Intelligent. 2014: 574-578. DOI: 10.1109/CYBER.2014.6917527.
[6] PARK H, PARK J, LEE D H, et al. Compliance-Based Robotic Peg-in-Hole Assembly Strategy Without Force Feedback[J/OL]. IEEE Transactions on Industrial Electronics, 2017, 64(8):6299-6309. DOI: 10.1109/TIE.2017.2682002.
[7] PARK H, KIM P K, BAE J H, et al. Dual arm peg-in-hole assembly with a programmed compliant system[M/OL]//2014 11th International Conference on Ubiquitous Robots and AmbientIntelligence (URAI). 2014: 431-433. DOI: 10.1109/URAI.2014.7057477.
[8] CHANG W C. Robotic assembly of smartphone back shells with eye-in-hand visual servoing[J/OL]. Robotics and Computer-Integrated Manufacturing, 2018, 50: 102-113. https://www.sciencedirect.com/science/article/pii/S073658451630271X. DOI: https://doi.org/10.1016/j.rcim.2017.09.010.
[9] WANG R, LIANG C, PAN D, et al. Research on a Visual Servo Method of a ManipulatorBased on Velocity Feedforward[J/OL]. Space: Science &; Technology, 2021, 2021. DOI:10.34133/2021/9763179.
[10] NIU X, PU J, ZHANG C. An Improved SIFT Algorithm for Monocular Vision Positioning[J/OL]. IOP Conference Series: Materials Science and Engineering, 2019, 612(3): 032124.https://dx.doi.org/10.1088/1757-899X/612/3/032124.
[11] DING G, LIU Y, ZANG X, et al. A Task-Learning Strategy for Robotic Assembly Tasks fromHuman Demonstrations[J/OL]. Sensors, 2020, 20(19). https://www.mdpi.com/1424-8220/20/19/5505.
[12] KANG H, ZANG Y, WANG X, et al. Uncertainty-Driven Spiral Trajectory for Robotic Peg-inHole Assembly[J/OL]. IEEE Robotics and Automation Letters, 2022, 7(3): 6661-6668. DOI:10.1109/LRA.2022.3176718.52参考文献
[13] GU J, ZHU M, CAO L, et al. Improved Uncalibrated Visual Servo Strategy for HyperRedundant Manipulators in On-Orbit Automatic Assembly[J/OL]. Applied Sciences, 2020,10(19). https://www.mdpi.com/2076-3417/10/19/6968.
[14] MOL N, SMISEK J, BABUšKA R, et al. Nested compliant admittance control for roboticmechanical assembly of misaligned and tightly toleranced parts[M/OL]//2016 IEEE International Conference on Systems, Man, and Cybernetics (SMC). 2016: 002717-002722. DOI:10.1109/SMC.2016.7844650.
[15] LISTMANN K D, HANS F, WAHRBURG A. Compliance Control for Robust Assembly withRedundant Manipulators[J/OL]. IFAC-PapersOnLine, 2020, 53(2): 8538-8545. https://www.sciencedirect.com/science/article/pii/S2405896320318231. DOI: https://doi.org/10.1016/j.ifacol.2020.12.1413.
[16] CAO H, CHEN X, HE Y, et al. Dynamic Adaptive Hybrid Impedance Control for Dynamic Contact Force Tracking in Uncertain Environments[J/OL]. IEEE Access, 2019, 7: 83162-83174.DOI: 10.1109/ACCESS.2019.2924696.
[17] XU J, HOU Z, LIU Z, et al. Compare contact model-based control and contact model-freelearning: A survey of robotic peg-in-hole assembly strategies[A]. 2019.
[18] JASIM I F, PLAPPER P W, VOOS H. Contact-state modelling in force-controlled roboticpeg-in-hole assembly processes of flexible objects using optimised Gaussian mixtures[J/OL].Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 2017, 231(8): 1448-1463. DOI: 10.1177/0954405415598945.
[19] HOU Z, FEI J, DENG Y, et al. Data-Efficient Hierarchical Reinforcement Learning for RoboticAssembly Control Applications[J/OL]. IEEE Transactions on Industrial Electronics, 2021, 68(11): 11565-11575. DOI: 10.1109/TIE.2020.3038072.
[20] HOU Z, LI Z, HSU C, et al. Fuzzy Logic-Driven Variable Time-Scale Prediction-Based Reinforcement Learning for Robotic Multiple Peg-in-Hole Assembly[J/OL]. IEEE Transactions onAutomation Science and Engineering, 2022, 19(1): 218-229. DOI: 10.1109/TASE.2020.3024725.
[21] WANG Y, BELTRAN-HERNANDEZ C C, WAN W, et al. Robotic Imitation of Human Assembly Skills Using Hybrid Trajectory and Force Learning[M/OL]//2021 IEEE InternationalConference on Robotics and Automation (ICRA). 2021: 11278-11284. DOI: 10.1109/ICRA48506.2021.9561619.
[22] CALINON S. Robot programming by demonstration[M]. EPFL Press, 2009.
[23] ARGALL B D, CHERNOVA S, VELOSO M, et al. A survey of robot learning from demonstration[J]. Robotics and Autonomous Systems, 2009, 57(5): 469-483.
[24] ZHU Z, HU H. Robot learning from demonstration in robotic assembly: A survey[J]. Robotics,2018, 7(2): 17.
[25] CALINON S, GUENTER F, BILLARD A. On learning, representing, and generalizing a taskin a humanoid robot[J]. IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), 2007, 37(2): 286-298.53参考文献
[26] CALINON S, BILLARD A. A probabilistic programming by demonstration framework handling constraints in joint space and task space[M]//2008 IEEE/RSJ International Conference onIntelligent Robots and Systems. IEEE, 2008: 367-372.
[27] CALINON S, ALIZADEH T, CALDWELL D G. On improving the extrapolation capability of task-parameterized movement models[M]//2013 IEEE/RSJ International Conference onIntelligent Robots and Systems. IEEE, 2013: 610-616.
[28] ALIZADEH T, CALINON S, CALDWELL D G. Learning from demonstrations with partiallyobservable task parameters[M]//2014 IEEE International Conference on Robotics and Automation (ICRA). IEEE, 2014: 3309-3314.
[29] OSA T, ESFAHANI A M G, STOLKIN R, et al. Guiding trajectory optimization by demonstrated distributions[J]. IEEE Robotics and Automation Letters, 2017, 2(2): 819-826.
[30] ZHU Z, HU H. Robot learning from demonstration in robotic assembly: A survey[J]. Robotics,2018, 7(2): 17.
[31] YANG C, ZENG C, CONG Y, et al. A learning framework of adaptive manipulative skills fromhuman to robot[J]. IEEE Transactions on Industrial Informatics, 2018, 15(2): 1153-1161.
[32] DING L, LI S, GAO H, et al. Adaptive partial reinforcement learning neural network-basedtracking control for wheeled mobile robotic systems[J]. IEEE Transactions on Systems, Man,and Cybernetics: Systems, 2018, 50(7): 2512-2523.
[33] ZHANG M Y, TIAN G H, LI C C, et al. Learning to transform service instructions into actionswith reinforcement learning and knowledge base[J]. International Journal of Automation andComputing, 2018, 15: 582-592.
[34] WANG Y, JIAO Y, XIONG R, et al. MASD: A multimodal assembly skill decoding systemfor robot programming by demonstration[J]. IEEE Transactions on Automation Science andEngineering, 2018, 15(4): 1722-1734.
[35] WAN A, XU J, CHEN H, et al. Optimal path planning and control of assembly robots for hardmeasuring easy-deformation assemblies[J]. IEEE/ASME Transactions on Mechatronics, 2017,22(4): 1600-1609.
[36] GU S, HOLLY E, LILLICRAP T, et al. Deep reinforcement learning for robotic manipulation with asynchronous off-policy updates[M/OL]//2017 IEEE International Conference onRobotics and Automation (ICRA). 2017: 3389-3396. DOI: 10.1109/ICRA.2017.7989385.
[37] HOPPE S, LOU Z, HENNES D, et al. Planning approximate exploration trajectories for modelfree reinforcement learning in contact-rich manipulation[J]. IEEE Robotics and AutomationLetters, 2019, 4(4): 4042-4047.
[38] HOU Z, FEI J, DENG Y, et al. Data-efficient hierarchical reinforcement learning for roboticassembly control applications[J]. IEEE Transactions on Industrial Electronics, 2020, 68(11):11565-11575.
[39] MNIH V, BADIA A P, MIRZA M, et al. Asynchronous methods for deep reinforcement learning[M]//International Conference on Machine Learning. PMLR, 2016: 1928-1937.
[40] XU J, HOU Z, WANG W, et al. Feedback deep deterministic policy gradient with fuzzy rewardfor robotic multiple peg-in-hole assembly tasks[J]. IEEE Transactions on Industrial Informatics,2018, 15(3): 1658-1667.54参考文献
[41] HOU Z, LI Z, HSU C, et al. Fuzzy logic-driven variable time-scale prediction-based reinforcement learning for robotic multiple peg-in-hole assembly[J]. IEEE Transactions on AutomationScience and Engineering, 2020, 19(1): 218-229.
[42] SAKATA N, KINOSHITA Y, KATO Y. Predicting a pedestrian trajectory using seq2seq formobile robot navigation[M]//IECON 2018-44th Annual Conference of the IEEE Industrial Electronics Society. IEEE, 2018: 4300-4305.
[43] ALAHI A, GOEL K, RAMANATHAN V, et al. Social lstm: Human trajectory predictionin crowded spaces[M]//Proceedings of the IEEE Conference on Computer Vision and PatternRecognition. 2016: 961-971.
[44] DEO N, TRIVEDI M M. Multi-modal trajectory prediction of surrounding vehicles with maneuver based lstms[M]//2018 IEEE Intelligent Vehicles Symposium (IV). IEEE, 2018: 1179-1184.
[45] BAHDANAU D, CHO K, BENGIO Y. Neural machine translation by jointly learning to alignand translate[A]. 2014.
[46] CHEN J, JIANG D, ZHANG Y. A hierarchical bidirectional GRU model with attention forEEG-based emotion classification[J]. IEEE Access, 2019, 7: 118530-118540.
[47] LU S, ZHANG Q, CHEN G, et al. A combined method for short-term traffic flow predictionbased on recurrent neural network[J]. Alexandria Engineering Journal, 2021, 60(1): 87-94.
[48] LIU S, SHENG J, OU Y. Learning Compliant Assembly Strategy From Demonstration[M]//2023 IEEE International Conference on Real-time Computing and Robotics (RCAR). IEEE,2023: 929-934.
[49] SHU W, CAI K, XIONG N N. A short-term traffic flow prediction model based on an improvedgate recurrent unit neural network[J]. IEEE Transactions on Intelligent Transportation Systems,2021, 23(9): 16654-16665.
[50] XUE H, HUYNH D Q, REYNOLDS M. Bi-prediction: Pedestrian trajectory predictionbased on bidirectional LSTM classification[M]//2017 International Conference on Digital Image Computing: Techniques and Applications (DICTA). IEEE, 2017: 1-8.
[51] CHEN C P, LIU Z. Broad learning system: An effective and efficient incremental learningsystem without the need for deep architecture[J]. IEEE Transactions on Neural Networks andLearning Systems, 2017, 29(1): 10-24.
[52] GONG X, ZHANG T, CHEN C P, et al. Research review for broad learning system: Algorithms,theory, and applications[J]. IEEE Transactions on Cybernetics, 2021, 52(9): 8922-8950.
[53] XU S, LIU J, YANG C, et al. A learning-based stable servo control strategy using broad learningsystem applied for microrobotic control[J]. IEEE Transactions on Cybernetics, 2021, 52(12):13727-13737.
[54] GONG X, ZHANG T, CHEN C L P, et al. Research Review for Broad Learning System: Algorithms, Theory, and Applications[J/OL]. IEEE Transactions on Cybernetics, 2022, 52(9):8922-8950. DOI: 10.1109/TCYB.2021.3061094.
[55] ZHOU P, SHI W, TIAN J, et al. Attention-based bidirectional long short-term memory networksfor relation classification[M]//Proceedings of the 54th annual meeting of the association forcomputational linguistics (volume 2: Short papers). 2016: 207-212.