Abstract
Typical neural-adaptive control approaches update neural-network weights as though they were adaptive parameters in a continuous-time adaptive control. However, requiring fast digital rates usually restricts the size of the neural network. In this paper we analyze a delta-rule update for the weights, applied at a relatively slow digital rate. We show that digital weight update causes the neural network to estimate a discrete-time model of the system, assuming that state feedback is still applied in continuous time. A Lyapunov analysis shows uniformly ultimately bounded signals. Furthermore, slowing the update frequency and using the extra computational time to increase the size/accuracy of the neural network results in better performance. Experimental results achieving link tracking of a two-link flexible-joint robot verify the improved performance.
Similar content being viewed by others
References
Albus J (1975) Data storage in the cerebellar model articulation controller (CMAC). J Dyn Syst Meas Contr 97:228–233
Albus J (1975) A new approach to manipulator control: the cerebellar model articulation controller (CMAC). J Dyn Syst Measur Contr 97:220–227
Chaturvedi DK, Malik OP (2007) Experimental studies of a generalized neuron based adaptive power system stabilizer. Soft Comput 11:149–155
Chiu CH (2010) Adaptive output recurrent cerebellar model articulation controller for nonlinear system control. Soft Comput 14:627–638
Chiu CH, Peng YF, Lin YW (2011) Robust intelligent backstepping tracking control for wheeled inverted pendulum. Soft Comput 15:2029–2040
Dixon W, Zergeroglu E, Dawson D, Hannan M (2000) Global adaptive partial state feedback tracking control of rigid- link flexible-joint robots. Robotica 18:325–336
Frayman Y, Wang L (2002) A dynamically constructed fuzzy neural controller for direct model reference adaptive control of multi-input-multi-output nonlinear processes. Soft Computing 6:244–253
Huang AC, Chen YC (2004) Adaptive sliding control for single-link flexible-joint robot with mismatched uncertainties. IEEE Trans. Contr. Syst. Technol 12(5):770–775
Huang AC, Chien MC (2009) Design of a regressor-free adaptive impedance controller for flexible-joint electrically driven robots. In: Proceedings of IEEE International Conference on Industrial Electronics and Applications, pp 17–22. Xi’an, China
Ider S, zgren M (2000) Trajectory tracking control of flexible-joint robots. Comput Struct 76(6):757–763
Ioannuou P, Kokotovic P (1984) Instability analysis and improvement of robustness of adaptive control. Automatica 20(5):583–594
Jagannathan S (1999) Discrete-time CMAC NN control of feedback linearizable nonlinear systems under a persistence of excitation. IEEE Trans Neural Netw 10:128–137
Kanellakopoulos I (1994) A discrete-time adaptive nonlinear system. IEEE Trans Automat Contr 39:2362–2365
Kim MS, Lee JS (2004) Adaptive tracking control of flexible-joint manipulators without overparametrization. J Robotic Syst 21(7):369–379
Kim N, Calise A, Hovakimyan N (2004) Several extensions in methods for adaptive output feedback control. In: Proceedings of American Control Conference, pp 2421–2426, Boston
Kim Y, Lewis F (2000) Optimal design of CMAC neural-network controller for robot manipulators. IEEE Trans Syst Man Cybern C 30(1):22–30
Kuo F, Sloan I (2005) Lifting the curse of dimensionality. Am Math Soc 52:1320–1328
Kwan C, Lewis F (2000) Robust backstepping control of nonlinear systems using neural networks. IEEE Trans Syst Man Cybern A 30:753–766
Lee Y, Zak SH (2004) Uniformly ultimately bounded fuzzy adaptive tracking controllers for uncertain systems. IEEE Trans Fuzzy Syst 12:797–811
Lei Y, Wu H (2006) Tracking control of robotic manipulators based on the all-coefficient adaptive control method. Int J Control Automat Syst 4(2):139–145
Macnab C (2007) A new robust weight update for multilayer-perceptron adaptive control. Control Intell Syst 35(3):279–288
Macnab C (2010) Improved output tracking of a flexible-joint arm using neural networks. Neural Process Lett 32(2):201–218
Macnab C, D’Eleuterio G (2000) Discrete-time Lyapunov design for neuroadaptive control of elastic-joint robots. Int J Robot Res 19:511–525
Miller III W, Glanz F, Kraft L (1990) CMAC: an associative neural network alternative to backpropagation. Proc IEEE 78(10):1561–1567
Miller III W, Hewes P, Glanz F, Kraft L (1990) Real-time dynamic control of an industrial manipulator using a neural-network-based learning controller. IEEE Trans Robot Automat 6(1):1–9
Nakanishi J, Schaal S (2004) Feedback error learning and nonlinear adaptive control. Neural Netw 17:1453–1465
Ozgoli S, Taghirad HD (2006) A survey on the control of flexible joint robots. Asian J Control 8(4):1–15
Rokui M, Khorasani K (1997) An indirect adaptive control for fully feedback linearizable discrete-time non-linear systems. Int J Adapt Contr Sig Proc 11:665–680
Rokui MR, Khorasani K (2000) Experimental results on discrete-time nonlinear adaptive tracking control of a flexible link manipulator. IEEE Trans Syst Man Cybern B 30(1):151–164
Spong M, Vidyasagar M (1989) Robot dynamics and control. Wiley, New York
Suna FC, Lib HX, Lic L (2002) Robot discrete adaptive control based on dynamic inversion using dynamical neural networks. Automatica 38:1977–1983
Tian L, Wang J, Mao Z (2004) Constrained motion control of flexible robot manipulators based on recurrent neural networks. IEEE Trans Syst Man Cybern B 34(3):1541–1552
Wang HR, Yang L, Wei LX (2007) Fuzzy-neuro position/force control for robotic manipulators with uncertainties. Soft Comput 11:311–315
Yeh P, Kokotovic P (1995) Adaptive control of a class of nonlinear discrete-time systems. Int J Contr 62:303–324
Author information
Authors and Affiliations
Corresponding author
Appendix: discrete-time transformation
Appendix: discrete-time transformation
We are interested in the discrete-time transformation of a continuous nonlinear system given by
We want to integrate the expression from time t = t 1 to t = t 2, where t 2 − t 1 = T. Let us denote terms like \({\bf A}({\varvec{\theta}}(t))\) simply as A(t) for ease of presentation. Now in anticipation of handling the time-varying matrix A(t), let us rewrite the equation as
Premultiply by \(\exp(-{\bf A}(t_1) t), \) and let us define \(\Updelta {\bf A} = {\bf A}(t)-{\bf A}(t_1); \) then
Let us integrate the left-hand side of (85).
By integrating and multiplying both sides by \(\exp({\bf A}(t_1) t_2)\) we have
Choose t 1 = kT, t 2 = (k + 1)T, and denote z k = z(kT), A k = A(kT), and B k = B(kT)
By defining \(\Updelta {\bf B}(t) = {\bf B}(t) - {\bf B}_{k}, \) then we can write
where (Fig. 8)
Rights and permissions
About this article
Cite this article
Richert, D., Masaud, K. & Macnab, C.J.B. Discrete-time weight updates in neural-adaptive control. Soft Comput 17, 431–444 (2013). https://doi.org/10.1007/s00500-012-0918-1
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00500-012-0918-1