基于动力学模型辨识的全臂柔顺控制

doi:10.3901/JME.2022.03.045

机械工程学报 ›› 2022, Vol. 58 ›› Issue (3): 45-54.doi: 10.3901/JME.2022.03.045

扫码分享

基于动力学模型辨识的全臂柔顺控制

李洋^1,2,3, 朱立爽⁴, 刘今越^1,2,3, 郭士杰^1,2,3

1. 河北工业大学省部共建电工装备可靠性与智能化国家重点实验室天津 300130;
2. 河北工业大学河北省机器人感知与人机融合重点实验室天津 300130;
3. 河北工业大学机械工程学院天津 300130;
4. 中国汽车技术研究中心有限公司天津 300300

收稿日期:2021-02-09 修回日期:2021-06-16 出版日期:2022-02-05 发布日期:2022-03-19
通讯作者: 郭士杰(通信作者),男,1963年出生,博士,教授,博士研究生导师。主要研究方向为智能护理机器人、柔性步行助力机器人、无束缚生理信息检测、机器人智能传感技术。E-mail:guoshijie@hebut.edu.cn
作者简介:李洋,男,1991年出生,博士研究生。主要研究方向为智能控制。E-mail:lyyliyang@126.com
基金资助:
国家重点研发计划（2016YFE0128700，2017YFB1301002）、河北省重点研发计划（18211816D）和河北省自然科学基金（E2017202270）资助项目。

Dynamic Model Identification for Whole-arm Compliance Control

LI Yang^1,2,3, ZHU Lishuang⁴, LIU Jinyue^1,2,3, GUO Shijie^1,2,3

1. State Key Laboratory of Reliability and Intelligence of Electrical Equipment, Hebei University of Technology, Tianjin 300130;
2. Hebei Key Laboratory of Robot Sensing and Human-robot Interaction, Hebei University of Technology, Tianjin 300130;
3. School of Mechanical Engineering, Hebei University of Technology, Tianjin 300130;
4. China Automotive Technology & Research Center Co. Ltd., Tianjin 300300

Received:2021-02-09 Revised:2021-06-16 Online:2022-02-05 Published:2022-03-19

摘要/Abstract

摘要： 基于动力学模型的阻抗控制无须力传感器，可降低系统复杂性，实现机械臂的全臂柔顺，但现实系统的动力学模型往往难以精确确定。监督学习可以通过关节状态回归辨识动力学模型，但辨识精度取决于观测数据的数量和质量，且难以泛化到未观测空间。提出一种基于先验动力学知识的递归参数辨识方法，可提高数据效率及泛化能力。辨识过程结合递推牛顿-欧拉动力学算法，逐一递归辨识关节参数，减少鞍点数量，克服辨识结果对初值的依赖性。在此基础上，设计了柔顺控制器；其外环为阻抗控制，通过辨识模型实现了无力传感器的全臂柔顺；内环采用滑模控制器，以辨识力矩作为动力学前馈，并通过径向基函数神经网络补偿系统的动态不确定性。实验结果表明，所提出的递归辨识算法可通过少量观测数据辨识完备的动力学模型，并实现全臂柔顺控制。

关键词: 全臂柔顺, 先验动力学知识, 递归参数辨识, 辨识模型

Abstract: Although the impedance control based on the dynamic model is independent of force sensors, simplifying the robot's complexity and realizing the whole-arm compliance, the real system's accurate dynamic model is hard to formulate. Supervised learning can identify the model through joint state regression, but the accuracy depends on the observed data's quantity and quality, and the identification model was challenging generalization in unobserved space. A recursive parameter identification method with prior dynamics knowledge are proposed to improve data efficiency and model's generalization. The joint's parameters are identified recursively with the iterative Newton-Euler dynamics algorithm, which reduces the number of saddle points and overcomes the dependence on the initial values. On this foundation, the whole-arm compliance controller is designed. The outer loop is impedance control with the identified model to realize the whole arm compliance without force sensors. The inner loop is sliding mode control with the identified model as the dynamic feedforward, and the radial basis function neural network is used to compensate for the system's dynamic uncertainty. The results show that the recursive identification algorithm can identify the complete dynamic model with a few observation data and realize the robot's whole-arm compliance.

Key words: whole-arm compliance, prior dynamics knowledge, recursive parameter identification, identification model

中图分类号:

TP242

李洋, 朱立爽, 刘今越, 郭士杰. 基于动力学模型辨识的全臂柔顺控制[J]. 机械工程学报, 2022, 58(3): 45-54.

LI Yang, ZHU Lishuang, LIU Jinyue, GUO Shijie. Dynamic Model Identification for Whole-arm Compliance Control[J]. Journal of Mechanical Engineering, 2022, 58(3): 45-54.

参考文献

[1] ZHANG W X,ZHANG X L,SHI D. A review on lower limb rehabilitation exoskeleton robots[J]. Chinese Journal of Mechanical Engineering,2019(4):2-12.
[2] OKE A O,AFOLABI A. Development of a robotic arm for dangerous object disposal[C]//International Conference on Computer Science and Information Technology. Bangalore,India:AIRCC Publishing Corporation,2014:153-160.
[3] LI Z,CHEN C L P,HE W. Guest editorial for special issue on human-centered intelligent robots:Issues and challenges[J]. Journal of Automatica Sinica,2017,4(4):599-601.
[4] HERRERA D,ROBERTI F,TOIBERO M,et al. Human interaction dynamics for its use in mobile robotics:Impedance control for leader-follower formation[J]. IEEE/CAA Journal of Automatica Sinica,2017,4(4):696-703.
[5] HUANG J,XIE Z,JIN M,et al. Adaptive impedance-controlled manipulator based on collision detection[J]. Chinese Journal of Aeronautics,2009,22(1):105-112.
[6] TAKAHASHI C D,SCHEIDT R A,REINKENSMEYER D J. Impedance control and internal model formation when reaching in a randomly varying dynamical environment[J]. Journal of Neurophysiology,2001,86(2):1047-1051.
[7] YUAN J,QIAN Y,GAO L,et al. Position-based impedance force controller with sensorless force estimation[J]. Assembly Automation,2019,39(3):489-496.
[8] 邢宏军,丁亮,高海波,等. 基于阻抗控制的机器人旋拧阀门轴向位置自适应跟踪[J]. 机械工程学报,201955(15):124-134. XING Hongjun,DING Liang,GAO Haibo,et al. Adaptive tracking of axial position for valve-turning with a robot based on impedance control[J]. Journal of Mechanical Engineering,2019,55(15):124-134.
[9] LUCA A D,FLACCO F. Integrated control for pHRI:collision avoidance,detection,reaction and collaboration[C]//Proceedings of the IEEE/RAS-EMBS International Conference on Biomedical Robotics and Biomechatronics. Rome,Italy:IEEE,2012:288-295.
[10] 邱澜,曹建国,周建辉,等. 机器人柔弹性仿生电子皮肤研究进展[J]. 中南大学学报,2019,50(5):1065-1074. QIU Lan,CAO Jianguo,ZHOU Jianhui,et al. Research progress of flexible and elastic bionic e-skin for robot[J]. Journal of Central South University,2019,50(5):1065-1074.
[11] CALANDRA R,IVALDI S,DEISENROTH M P,et al. Learning inverse dynamics models with contact[C]//Proceedings of the IEEE International Conference on Robotics and Automation. Barcelona,Spain:IEEE, 2005:3186-3191.
[12] STOLT A,LINDEROTH M,ROBERTSSON A,et al. Force controlled robotic assembly without a force sensor[C]//Proceedings of the IEEE International Conference on Robotics and Automation. St. Paul,Minnesota,USA:IEEE,2012:1538-1543.
[13] NGUYEN-TUONG D,SEEGER M W,PETERS J. Model learning with local gaussian process regression[J]. Advanced Robotics,2009,23(15):2015-2034.
[14] KARIMI A,BUTCHER M,LONGCHAMP R. Model-free precompensator tuning based on the correlation approach[J]. IEEE Transactions on Control Systems Technology,2008,16(5):1013-1020.
[15] FLIESS M,JOIN C. Model-free control and intelligent PID controllers:Towards a possible trivialization of nonlinear control?[J]. IFAC Proceedings Volumes,2009,42(10):1531-1550.
[16] CHEN Z,KAZANZIDES P. Force control of a non-backdrivable robot without a force sensor[C]//Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems. Tokyo,Japan:IEEE,2013:3570-3575.
[17] ZEMITI N,ORTMAIER T,VITRANI M A,et al. A force controlled laparoscopic surgical robot without distal force sensing[C]//Proceedings of the International Symposium on Experimental Robotics (ISER). Rio de Janeiro,Brazil:Springer,2004:18-21.
[18] LEVINE S,FINN C,DARRELL T,et al. End-to-end training of deep visuomotor policies[J]. Machine Learning Research,2016,17(1):1334-1373.
[19] FINN C,XIN Y T,DUAN Y,et al. Deep spatial autoencoders for visuomotor learning[C]//Proceedings of the IEEE International Conference on Robotics and Automation. Qingdao,China:IEEE,2016:512-519.
[20] CHEBOTAR Y,HAUSMAN K,ZHANG M,et al. Combining model-based and model-free updates for trajectory-centric reinforcement learning[C]//Proceedings of the International Conference on Machine Learning. NSW,Australia:PMLR,2017:703-711.
[21] DEISENROTH M,RASMUSSEN C E. PILCO:A model-based and data-efficient approach to policy search[C]//Proceedings of the International Conference on machine learning. Washington,USA:PMLR,2011:465-472.
[22] HA S,YAMANE K. Reducing hardware experiments for model learning and policy optimization[C]//Proceedings of the IEEE International Conference on Robotics and Automation. FL,USA:IEEE,2015:2620-2626.
[23] SNELSON E,GHAHRAMANI Z. Sparse gaussian processes using pseudo-inputs[J]. Advances in neural information processing systems,2006,18(1):1257-1264.
[24] LAZARO-GREDILLA M,QUINONERO-CANDELA J,RASMUSSEN C E,et al. Sparse spectrum gaussian process regression[J]. Journal of Machine Learning Research,2010,11:1865-1881.
[25] LESHNO M,LIN V Y,PINKUS A,et al. Multilayer feedforward networks with a nonpolynomial activation function can approximate any function[J]. Neural networks,1993,6(6):861-867.
[26] HORNIK K,STINCHCOMBE M,WHITE H. Multilayer feedforward networks are universal approximators[J]. Neural networks,1989,2(5):359-366.
[27] 陈贵亮,李长鹏,赵月,等. 下肢外骨骼康复机器人的动力学建模及神经网络辨识仿真[J]. 机械设计与制造,2013(11):197-200. CHEN Guiliang,LI Changpeng,ZHAO Yue,et al. Dynamic modeling and neural network identification simulation for lower limbs exoskeletons rehabilitation robot[J]. Machinery Design & Manufacture,2013(11):197-200.
[28] CHOROMANSKA A,HENAFF M,MATHIEU M,et al. The loss surfaces of multilayer networks[J]. Eprint Arxiv,2014:192-204.
[29] DAUPHIN Y N,PASCANU R,GULCEHRE C,et al. Identifying and attacking the saddle point problem in high-dimensional non-convex optimization[C]//Proceedings of the International Conference on Neural Information Processing Systems. Kuching,Malaysia:MIT,2014:2933-2941.
[30] KRZYWINSKI M,ALTMAN N. Points of significance:importance of being uncertain[J]. Nature Methods,2013,10(9):809-810.
[31] CHUA K,CALANDRA R,MCALLISTER R,et al. Deep reinforcement learning in a handful of trials using probabilistic dynamics models[J]. Advance Neural Information Process System,2018:4754-4765.
[32] DEISENROTH M P,FOX D,RASMUSSEN C E. Gaussian processes for data-efficient learning in robotics and control[J]. IEEE Transactions on Pattern Analysis & Machine Intelligence,2015,37(2):408-423.
[33] ROMERES D,ZORZI M,CAMORIANO R,et al. Derivative-free online learning of inverse dynamics models[J]. IEEE Transactions on Control Systems Technology,2019:1-15.
[34] LU H,TSAI C,CHANG M. Radial basis function neural network with sliding mode control for robotic manipulators[C]//Proceedings of the IEEE International Conference on Systems Man & Cybernetics. Istanbul,Turkey:IEEE,2010:1209-1215.
[35] SWEVERS J,GANSEMAN C. Optimal robot excitation and identification[J]. IEEE Transactions on Robotics and Automation,1997,13(5):730-740.

基于动力学模型辨识的全臂柔顺控制

Dynamic Model Identification for Whole-arm Compliance Control

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 1

编辑推荐

Metrics

本文评价