Robot Skill Learning Based on Dynamic Motion Primitives and Adaptive Control

doi:10.16183/j.cnki.jsjtu.2021.379

Abstract

Abstract:

A novel robot skill learning method using dynamic movement primitive (DMP) and adaptive control is proposed. The existing DMP method learns actions from a single teaching trajectory, and its Gaussian basis function distribution mode is fixed, which is not suitable for multiple movement trajectories with different characteristics. Therefore, the Gaussian mixture model (GMM) and Gaussian mixture regression are introduced into DMP to enable the robot to learn skills from multi-teaching trajectory. Moreover, radial basis function neural network (RBFNN) is introduced into DMP to establish the RBF-DMP method, which is able to learn the central position and weight of Gaussian basis through gradient descent and improves the accuracy of skill modeling. Furthermore, an adaptive neural network controller is designed to control the learned actions of the manipulator in redemonstration. Finally, experiments on Franka Emika Panda manipulator prove the effectiveness of the proposed method.

Key words: dynamic movement primitive (DMP), Gaussian mixture model (GMM), radial basis function neural network (RBFNN), robot learning

CLC Number:

TP242

ZHANG Wenan, GAO Weizhan, LIU Andong. Robot Skill Learning Based on Dynamic Motion Primitives and Adaptive Control[J]. Journal of Shanghai Jiao Tong University, 2023, 57(3): 354-365.

Figures/Tables 14

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Tab.1

Fig.8

Fig.9

Fig.10

Fig.11

Fig.12

Fig.13

References 21

[1]	马良, 徐晓兰, 丁汉, 等. 2019年中国机器人产业发展报告[R]. 北京: 中国电子学会, 2019.
	MA Liang, XU Xiaolan, DING Han, et al. China robot industry development report 2019[R]. Beijing: China Electronics Association, 2019.
[2]	EWERTON M, MAEDA G, KOLLEGGER G, et al. Incremental imitation learning of context-dependent motor skills[C]// 2016 IEEE-RAS 16th International Conference on Humanoid Robots (Humanoids). Cancun, Mexico: IEEE, 2016: 351-358.
[3]	BROCK O, KHATIB O. Elastic strips: A framework for integrated planning and execution[C]// Experimental Robotics VI. London: Springer, 2000: 329-338.
[4]	ZHANG H W, LENG Y Q. Motor skills learning and generalization with adapted curvilinear Gaussian mixture model[J]. Journal of Intelligent & Robotic Systems, 2019, 96(3/4): 457-475.
[5]	CHANG G T, KULIĆ D. Motion learning from observation using affinity propagation clustering[C]// 2013 IEEE RO-MAN. Gyeongju, Korea (South): IEEE, 2013: 662-667.
[6]	IJSPEERT A J, NAKANISHI J, HOFFMANN H, et al. Dynamical movement primitives: Learning attractor models for motor behaviors[J]. Neural Computation, 2013, 25(2): 328-373. doi: 10.1162/NECO_a_00393 pmid: 23148415
[7]	MEIER F, SCHAAL S. A probabilistic representation for dynamic movement primitives[EB/OL]. (2016-12-18) [2021-06-05]. https://arxiv.org/abs/1612.05932.
[8]	GAŠPAR T, NEMEC B, MORIMOTO J, et al. Skill learning and action recognition by arc-length dynamic movement primitives[J]. Robotics & Autonomous Systems, 2018, 100: 225-235.
[9]	LI J J, LI Z J, LI X D, et al. Skill learning strategy based on dynamic motion primitives for human-robot cooperative manipulation[J]. IEEE Transactions on Cognitive & Developmental Systems, 2021, 13(1): 105-117.
[10]	YANG C G, JIANG Y M, LI Z J, et al. Neural control of bimanual robots with guaranteed global stability and motion precision[J]. IEEE Transactions on Industrial Informatics, 2017, 13(3): 1162-1171. doi: 10.1109/TII.2016.2612646 URL
[11]	BECKERS T, UMLAUFT J, HIRCHE S. Stable model-based control with Gaussian process regression for robot manipulators[J]. IFAC-PapersOnLine, 2017, 50(1): 3877-3884. doi: 10.1016/j.ifacol.2017.08.359 URL
[12]	YANG C G, JIANG Y M, NA J, et al. Finite-time convergence adaptive fuzzy control for dual-arm robot with unknown kinematics and dynamics[J]. IEEE Transactions on Fuzzy Systems, 2019, 27(3): 574-588. doi: 10.1109/TFUZZ.91 URL
[13]	ZHANG S, DONG Y T, OUYANG Y C, et al. Adaptive neural control for robotic manipulators with output constraints and uncertainties[J]. IEEE Transactions on Neural Networks & Learning Systems, 2018, 29(11): 5554-5564.
[14]	ZHANG Y Y, CHEN S Y, LI S, et al. Adaptive projection neural network for kinematic control of redundant manipulators with unknown physical parameters[J]. IEEE Transactions on Industrial Electronics, 2018, 65(6): 4909-4920. doi: 10.1109/TIE.41 URL
[15]	XU Z H, LI S, ZHOU X F, et al. Dynamic neural networks based kinematic control for redundant manipulators with model uncertainties[J]. Neurocomputing, 2019, 329: 255-266. doi: 10.1016/j.neucom.2018.11.001 URL
[16]	MAO Z Y, ZHAO F L. Structure optimization of a vibration suppression device for underwater moored platforms using CFD and neural network[J]. Complexity, 2017, 2017: 5392539.
[17]	YIN X C, CHEN Q J. Learning nonlinear dynamical system for movement primitives[C]// 2014 IEEE International Conference on Systems, Man, and Cybernetics. San Diego, USA: IEEE, 2014: 3761-3766.
[18]	YANG C G, CHEN C Z, HE W, et al. Robot learning system based on adaptive neural control and dynamic movement primitives[J]. IEEE Transactions on Neural Networks & Learning Systems, 2019, 30(3): 777-787.
[19]	YANG C G, WANG X Y, CHENG L, et al. Neural-learning-based telerobot control with guaranteed performance[J]. IEEE Transactions on Cybernetics, 2017, 47(10): 3148-3159. doi: 10.1109/TCYB.2016.2573837 pmid: 28113610
[20]	YANG C G, WANG X J, LI Z J, et al. Teleoperation control based on combination of wave variable and neural networks[J]. IEEE Transactions on Systems, Man, & Cybernetics: Systems, 2017, 47(8): 2125-2136.
[21]	GHORBEL F, SRINIVASAN B, SPONG M. On the positive definiteness and uniform boundedness of the inertia matrix of robot manipulators[C]// Proceedings of 32nd IEEE Conference on Decision and Control. San Antonio, USA: IEEE, 1993: 1103-1108.

方法	总MSE	总建模时间/s
DMP	1.628	12.91
RBF-DMP	1.011	16.24
拓展框架	1.193	13.98