zbMATH — the first resource for mathematics

A robust training algorithm of discrete-time MIMO RNN and application in fault tolerant control of robotic system. (English) Zbl 1327.93288
Summary: In this paper, a novel robust training algorithm of multi-input multi-output recurrent neural network and its application in the fault tolerant control of a robotic system are investigated. The proposed scheme optimizes the gradient type training on basis of three new adaptive parameters, namely, dead-zone learning rate, hybrid learning rate, and normalization factor. The adaptive dead-zone learning rate is employed to improve the steady state response. The normalization factor is used to maximize the gradient depth in the training, so as to improve the transient response. The hybrid learning rate switches the training between the back-propagation and the real-time recurrent learning mode, such that the training is robust stable. The weight convergence and \(L_{2}\) stability of the algorithm are proved via Lyapunov function and the Cluett’s law, respectively. Based upon the theoretical results, we carry out simulation studies of a two-link robot arm position tracking control system. A computed torque controller is designed to provide a specified closed-loop performance in a fault-free condition, and then the RNN compensator and the robust training algorithm are employed to recover the performance in case that fault occurs. Comparisons are given to demonstrate the advantages of the control method and the proposed training algorithm.
93C85 Automated systems (robots, etc.) in control theory
93C55 Discrete-time control/observation systems
Full Text: DOI
[1] Lewis FL, Liu K, Yesildirek A (1995) Neural net robot controller with guaranteed tracking performance. IEEE Trans Neural Netw 16(3):703–715 · doi:10.1109/72.377975
[2] Vemuri AT, Polycarpou MM, Diakourtis SA (1998) Neural network based fault detection in robotic manipulators. IEEE Trans Rob Autom 14:342–348 · doi:10.1109/70.681254
[3] Trunov AB, Polycarpou MM (2000) Automated fault diagnosis in nonlinear multivariable systems using a learning methodology. IEEE Trans Neural Netw 11:91–101 · doi:10.1109/72.822513
[4] Song Q, Yin L (2001) Robust adaptive fault accommodation for a robot system using a radial basis function neural network. Int J Syst Sci 32(2):195–204 · Zbl 1002.93535
[5] Song Q, Hu WJ, Soh (2002) Robust adaptive dead zone technology for fault-tolerant control of robot manipulators using neural networks. J Intell Robot Syst 33(2):113–137 · Zbl 1047.93546 · doi:10.1023/A:1014603028024
[6] Mandic DP, Chambers JA (2001) Recurrent neural networks for prediction: learning algorithms, architecture and stability. Wiley, Chichester
[7] Liang X, Chen RC, Yang J (2008) An architecture-adaptive neural network online control system. Neural Comput Appl 17: 413–423 · doi:10.1007/s00521-007-0137-3
[8] Blanke M (1997) Fault-tolerant control systems–a history view. Control Eng Pract 5(5):693–702 · doi:10.1016/S0967-0661(97)00051-8
[9] Liu Z, Li CW (2003) Fuzzy neural network quadratic stabilization output feedback control for biped robots via happroach. IEEE Trans Syst Man Cybern B 33(1):67–84
[10] Lin CM, Chen CH (2007) Robust fault-tolerant control for a biped robot using a recurrent cerebellar model articulation controller. IEEE Trans Syst Man Cybern B 37:110–123 · doi:10.1109/TSMCB.2006.881905
[11] Wu YL, Song Q, Yang XL (2007) Robust recurrent neural network control of biped robot. J Intell Robot Syst 49:151–169 · Zbl 05193650 · doi:10.1007/s10846-007-9133-1
[12] Williams R, Zipser D (1989) A learning algorithm for continually running fully recurrent neural networks. Neural Comput 1:270–280 · doi:10.1162/neco.1989.1.2.270
[13] Rumelhart D, Hinton GE, Williams R (1986) Learning internal representation by error propagation. Parallel Distrib Process 1:319–361
[14] Wu YL, Song Q, Liu S (2008) A normalized adaptive training of recurrent neural networks with augmented error gradient. IEEE Trans Neural Netw 19:351–356 · doi:10.1109/TNN.2008.2003271
[15] Song Q, Wu YL, Soh YC (2008) Robust adaptive gradient descent training algorithm for recurrent neural networks in discrete time domain. IEEE Trans Neural Netw 19:1841–1853 · doi:10.1109/TNN.2008.2001923
[16] Man ZH, Yu XH, Wu HR (1998) An rbf neural network-based adaptive control for siso linearisable nonlinear systems. Neural Comput Appl 7:71–77 · Zbl 0925.93476 · doi:10.1007/BF01413711
[17] Song Q (2003) Robust neural network controller for variable airflow volume system. IEE Proc Contr Theor Appl 150:112–118 · doi:10.1049/ip-cta:20030065
[18] Mandic DP, Chambers JA (2000) A normalised real time recurrent learning algorithm. Elsevier Signal Process 80(9):1909–1916 · Zbl 1035.94513
[19] Cluett VR, Shah L, Fisher G (1988) Robustness analysis of discrete-time adaptive control systems using input-output stability theory: a tutorial. IEE Proc 135:133–141 · Zbl 0639.93005
[20] Nelles O (2001) Nonlinear system identification: from classic approaches to neural network and fuzzy models. Springer-Velag, Berlin
[21] Song Q, Xiao J, Soh YC (1999) Robust backpropagation training algorithm for multi-layered neural tracking controller. IEEE Trans Neural Netw 10(5):1133–1141 · doi:10.1109/72.788652
[22] Haykin S (1999) Neural networks. Prentice-Hall, Upper saddle River, NJ · Zbl 0934.68076
[23] Tzafestas SG, Krikochoritis AE, Tzafestas CS (1997) A robust-adaptive locomotion controller for 9-link with rapidly varying unkown parameters In: Proc Mediterranean Conf Contr Syst, pp 21–23
[24] Lewis FL, Abdallah CT, Dawson D (1993) Control of robot manipulators. Macmillan Press, New York
[25] Lin CM, Chen CH, Hus CF, Fan WZ (2005) Robust fault-tolerant control for robotic system using recurrent cerebellar model articulation controller In: Proc. IEEE Int Conf Ind Technol, pp 1006–1011
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.