For the identification of strongly coupled, nonlinear robotic systems, a modeling method multioutput support vector machine (MSVM) which based on feature vectors selection (FVS) was proposed.This method maps the input data to a high-dimensional feature space through kernel trick, considering the geometric features, selecting the relevant data subset to form the base vector of the subspace, which is called feature vector selection (FVS).Secondly, the input data was projected onto the feature subspace, and the final regression model was established by using the MSVM method.MSVM reserves the advantages of compact and sparse under the ε insensitive loss function.In order to verify the effectiveness of the FVS modeling method, it was applied to the oil pressure identification of the hydraulic drive robot, the inverse kinematics identification of the PUMA 560 industrial robot, and the inverse kinetic modeling of the SARCOS humanoid robot.Under the same conditions, the FVS-MSVM method was compared with the SVM, the KPCA-MSVM and the FVS-linear regression (LR) method etc.The experimental results show that the FVS-MSVM method can not only reduce the computational complexity, but also has good identification accuracy and good model promotion.Among them, the FVS-MSVM method has higher accuracy.
李 军,张观东. FVS-MSVM方法在机器人建模与辨识中的应用[J]. 振动与冲击, 2018, 37(20): 67-74.
LI Jun,ZHANG Guandong. Robot modeling and identification based on the FVS-MSVM methodmethod. JOURNAL OF VIBRATION AND SHOCK, 2018, 37(20): 67-74.
[1] 刘金琨. 机器人控制系统的设计与MATLAB仿真[M].清华大学出版社, 2008.
[2] Lai G, Liu Z, Zhang Y, et al. Adaptive position/attitude tracking control of aerial robot with unknown i-nertial matrix based on a new robust neural identifier[J]. IEEE transactions on neural networks and learning systems, 2016, 27(1): 18-31.
[3] He W, Chen Y, Yin Z. Adaptive neural network co-ntrol of an uncertain robot with full-state constraints[J].IEEE Transactions on Cybernetics, 2016, 46(3): 620-629.
[4] 梁捷;陈力 . 关节柔性的漂浮基空间机器人基于奇异摄动法的轨迹跟踪非奇异模糊Terminal滑模控制及柔性振动抑制[J].振动与冲击 , 2013, 32(23): 6-12.
LIANG Jie;CHEN Li . Nonsingular fuzzy terminal sliding mode control based on singular suppression and elastic vibration suppressing of free-floating space robot with flexible joints.Journal of vibration and shock , 2013, 32(23):6-12.
[5] Laurain V, Toth R, Piga D, et al. An instrumental l-east squares support vector machine for nonlinear system identification[J]. Automatica, 2015, 54: 340-347.
[6] Mao W, Xu J, Wang C, et al. A fast and robust m-odel selection algorithm for multi-input multi-output support vector machine[J]. Neurocomputing, 2014, 130:10-19.
[7] 杨扬, 宋亦旭, 梁伟,等. 基于SVM的机器人高精度磨削建模[J].机器人,2010,32(2):278-282.
YANG Yang, SONG Yixu, LIANG Wei, et al. Modeli-ng for Robot High Precision Grinding Based on SVM. ROBOT, 2010, 32(2): 278-282.
[8] Mozaffari A, Behzadipour S. A modular extreme lea-rning machine with linguistic interpreter and accelerated chaotic distributor for evaluating the safety of robot maneuvers in laparoscopic surgery[J]. Neurocomputing, 2015, 151: 913-932.
[9] 李军 乃永强. 基于ELM的一类MIMO仿射非线性系统的鲁棒自适应控制[J]. 控制与决策, 2015, 30(09): 1559-1566.
LI Jun NAI Yong-qiang. Robust adaptive control for a class of MIMO affine nonlinear systems using extreme learning machine. Control and Decision, 2015, 30(09): 1559-1566.
[10] Scholkopf B,Smola A J. Learning with kernel[M].Ca-mbridge, MA: MIT Press, 2000.
[11] Baudat G, Anouar F. Feature vector selection and pr-ojection using kernels[J]. Neurocomputing, 2003, 55(1): 21-38.
[12] 张少捷, 王振雷, 钱锋. 基于LTSA的FS-SVDD方法及其在化工过程监控中的应用[J].化工学报,2010,61(8):1894-1900.
Zhang S, Wang Z, Feng Q. FS-SVDD based on LT-SA and its application to chemical process monitoring[J]. Ciesc Journal,2010,61(8):1894-1900.
[13] Tuia D, Verrelst J, Alonso L, et al. Multioutput Sup-portVector Regression for Remote Sensing Biophysi-cal Parameter Estimation[J]. IEEE Geoscience & Remote Sensing Letters, 2011, 8(4):804-808.
[14] Perez-Cruz F, Navia-Vazquez A, Alarcon-Diana P L, et al. An IRWLS procedure for SVR[C]// IEEE Signal ProcessingConference, European, 2000:1-4.
[15] Civicioglu P. Backtracking Search Optimization Algo-rithm for numerical optimization problems[J]. AppliedMathematics & Computation, 2013, 219(15):8121-8144.
[16] Sjoberg J, Zhang Q, Ljung L, et al. Nonlinear black-boxmodeling in system identification: a unified overview[J]. Automatica, 1995, 31(12): 1691-1724.
[17] 李军, 黄杰. 基于自组织映射神经网络和局部自回归模型的网络流量预测[J]. 信息与控制, 2016,45(1): 120-128
LI Jun, HUANG Jie. Prediction of Network Traffic Using Local Auto-regressive Methods Based on Self-organizing Map Neural Network. Information and control, 2016, 45(1): 120-128.
[18] Elgazzar S. Efficient kinematic transformations for t-he PUMA 560 robot[J]. IEEE Journal on Robotics &Automation, 1985, 1(3):142-151.
[19] Corke P. A robotics toolbox for MATLAB[J]. intern-ational conference on robotics and automation, 1996,3(1): 24-32.
[20] Vijayakumar S, D'souza A, Schaal S. Incremental on-line learning in high dimensions[J]. Neural computat-ion, 2005, 17(12): 2602-26.