Abstract
This paper focuses on the parameter identification of a class of nonlinear mechanical dynamic systems subject to feedback quantization errors. A second-order integral method for the parameter identification of a nonlinear two-degrees-of-freedom robotic manipulator is studied. To obtain accurate parameter estimates, an optimal excitation trajectory subject to state constraints is proposed. The designated excitation trajectory is able to minimize estimation uncertainties, providing the controlled system trajectory reaches the desired optimal excitation trajectory. Moreover, for most of the nonlinear robotic systems, the output measurements are acquired from the encoders subject to apparent quantization error. Directly differentiating the position to estimate the velocity and acceleration trajectories is unrealistic and unsuitable due to the amplified quantization errors will cause considerable estimation errors. Instead, the finite Fourier series is introduced to perform optimal estimation of the position, velocity, and acceleration histories to capture the dominant frequencies of the system response. A weighted least-squares solution is further proposed to improve parameter identification precision. The comparative studies reveal that the integral method has superior parameter estimation performance than the direct difference model and the filtered regression model in the existing literature. Experimental simulations illustrate the effectiveness of the presented method.
Similar content being viewed by others
Data availability
The manuscript has no associated data.
References
Bahloul, A., Tliba, S., Chitour, Y.: Dynamic parameters identification of an industrial robot with and without payload. Ifac-Papersonline 51(15), 443–448 (2018)
Calafiore, G., Indri, M., Bona, B.: Robot dynamic calibration: optimal excitation trajectories and experimental parameter estimation. J. Robot. Syst. 18(2), 55–68 (2001)
Calanca, A., Capisani, L.M., Ferrara, A., Magnani, L.: Mimo closed loop identification of an industrial robot. IEEE Trans. Control Syst. Technol. 19(5), 1214–1224 (2010)
Chatzilygeroudis, K., Rama, R., Kaushik, R., Goepp, D., Vassiliades, V., Mouret, J.B.: Black-box data-efficient policy search for robotics. In: 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 51–58. IEEE (2017)
Chen, C.L., Peng, C.C.: Control of a perturbed chaotic system by using a trajectory trapping strategy. Nonlinear Dyn. 69(4), 2105–2115 (2012)
Chen, L.H., Peng, C.C.: Extended backstepping sliding controller design for chattering attenuation and its application for servo motor control. Appl. Sci. 7(3), 220 (2017)
Deng, J., Shang, W., Zhang, B., Zhen, S., Cong, S.: Dynamic model identification of collaborative robots using a three-loop iterative method. In: 2021 6th IEEE International Conference on Advanced Robotics and Mechatronics (ICARM), pp. 937–942. IEEE (2021)
Dolinskỳ, K., Čelikovskỳ, S.: Application of the method of maximum likelihood to identification of bipedal walking robots. IEEE Trans. Control Syst. Technol. 26(4), 1500–1507 (2017)
Gautier, M.: Dynamic identification of robots with power model. In: Proceedings of International Conference on Robotics and Automation, vol. 3, pp. 1922–1927. IEEE (1997)
Gautier, M., Janot, A., Vandanjon, P.: A new closed-loop output error method for parameter identification of robot dynamics. IEEE Trans. Control Syst. Technol. 21(2), 428–444 (2012)
Gautier, M., Janot, A., Vandanjon, P.O.: Didim: A new method for the dynamic identification of robots from only torque data. In: 2008 IEEE International Conference on Robotics and Automation, pp. 2122–2127. IEEE (2008)
Gautier, M., Khalil, W., Restrepo, P.: Identification of the dynamic parameters of a closed loop robot. In: Proceedings of 1995 IEEE International Conference on Robotics and Automation, vol. 3, pp. 3045–3050. IEEE (1995)
Guého, D., Singla, P., Majji, M., Melton, R.G.: Filtered integral formulation of the sparse model identification problem. J. Guid. Control Dyn. 45(2), 232–247 (2022)
Han, S.H., Yoo, A., Yoon, S.W., Yoon, Y.D.: A simple method for identifying mechanical parameters based on integral calculation. J. Power Electron. 16(4), 1387–1395 (2016)
Han, Y., Wu, J., Liu, C., Xiong, Z.: An iterative approach for accurate dynamic model identification of industrial robots. IEEE Trans. Robot. 36(5), 1577–1594 (2020)
He, W., Ge, W., Li, Y., Liu, Y.J., Yang, C., Sun, C.: Model identification and control design for a humanoid robot. IEEE Trans. Syst. Man Cybern. Syst. 47(1), 45–57 (2016)
Horvath, S., Neuner, H.: System identification of a robot arm with extended kalman filter and artificial neural networks. J. Appl. Geod. 13(2), 135–150 (2019)
Jia, J., Zhang, M., Zang, X., Zhang, H., Zhao, J.: Dynamic parameter identification for a manipulator with joint torque sensors based on an improved experimental design. Sensors 19(10), 2248 (2019)
Jin, J., Gans, N.: Parameter identification for industrial robots with a fast and robust trajectory design approach. Robot. Comput. Integr. Manuf. 31, 21–29 (2015)
Juang, J.N.: Applied System Identification. Prentice-Hall Inc., Hoboken (1994)
Juang, J.N.: Continuous-time bilinear system identification. Nonlinear Dyn. 39(1), 79–94 (2005)
Juang, J.N., Pappa, R.S.: An eigensystem realization algorithm for modal parameter identification and model reduction. J. Guid. Control Dyn. 8(5), 620–627 (1985)
Juang, J.N., Phan, M.Q.: Identification and Control of Mechanical Systems. Cambridge University Press, Cambridge (2001)
Lee, T., Wensing, P.M., Park, F.C.: Geometric robot dynamic identification: a convex programming approach. IEEE Trans. Robot. 36(2), 348–365 (2019)
Li, Y., Li, Y., Zhu, M., Xu, Z., Mu, D.: A nonlinear momentum observer for sensorless robot collision detection under model uncertainties. Mechatronics 78, 102603 (2021)
Li, Y.R., Peng, C.C.: Encoder position feedback based indirect integral method for motor parameter identification subject to asymmetric friction. Int. J. Non-Linear Mech. 152, 104386 (2023)
Liu, G., Iagnemma, K., Dubowsky, S., Morel, G.: A base force/torque sensor approach to robot manipulator inertial parameter estimation. In: Proceedings. 1998 IEEE International Conference on Robotics and Automation (Cat. No. 98CH36146), vol. 4, pp. 3316–3321. IEEE (1998)
Liu, X., Yang, X., Yin, S.: Nonlinear system identification with robust multiple model approach. IEEE Trans. Control Syst. Technol. 28(6), 2728–2735 (2019)
Lopez-Sanchez, I., Montoya-Cháirez, J., Pérez-Alcocer, R., Moreno-Valenzuela, J.: Experimental parameter identifications of a quadrotor by using an optimized trajectory. IEEE Access 8, 167355–167370 (2020)
Mablekos-Alexiou, A., Da Cruz, L., Bergeles, C.: Friction-inclusive modeling of sliding contact transmission systems in robotics. IEEE Trans. Robot. 37(4), 1252–1267 (2021)
Majji, M., Juang, J.N., Junkins, J.L.: Observer/kalman-filter time-varying system identification. J. Guid. Control Dyn. 33(3), 887–900 (2010)
Majji, M., Juang, J.N., Junkins, J.L.: Time-varying eigensystem realization algorithm. J. Guid. Control Dyn. 33(1), 13–28 (2010)
MATLAB: version 9.6 (R2019a). The MathWorks Inc., Natick, MA, USA (2019)
Moreno-Valenzuela, J., Aguilar-Avelar, C.: Motion Control of Underactuated Mechanical Systems, vol. 1. Springer, Berlin (2018)
Moreno-Valenzuela, J., Miranda-Colorado, R., Aguilar-Avelar, C.: A matlab-based identification procedure applied to a two-degrees-of-freedom robot manipulator for engineering students. Int. J. Electr. Eng. Educ. 54(4), 319–340 (2017)
Naozuka, G.T., Rocha, H.L., Silva, R.S., Almeida, R.C.: Sindy-sa framework: enhancing nonlinear system identification with sensitivity analysis. Nonlinear Dyn. 110, 2589–2609 (2022)
Peng, C.C., Li, Y.R.: Parameters identification of nonlinear Lorenz chaotic system for high-precision model reference synchronization. Nonlinear Dyn. 108(2), 1733–1754 (2022)
Peng, C.C., Lin, Y.I.: Dynamics modeling and parameter identification of a cooling fan system. In: 2018 IEEE International Conference on Advanced Manufacturing (ICAM), pp. 257–260. IEEE (2018)
Reyes, F., Kelly, R.: Experimental evaluation of identification schemes on a direct drive robot. Robotica 15(5), 563–571 (1997)
Swevers, J., Ganseman, C., De Schutter, J., Van Brussel, H.: Experimental robot identification using optimised periodic trajectories. Mech. Syst. Signal Process. 10(5), 561–577 (1996)
Swevers, J., Ganseman, C., Tukel, D.B., De Schutter, J., Van Brussel, H.: Optimal robot excitation and identification. IEEE Trans. Robot. Autom. 13(5), 730–740 (1997)
Swevers, J., Verdonck, W., De Schutter, J.: Dynamic model identification for industrial robots. IEEE Control Syst. Mag. 27(5), 58–71 (2007)
Vicente, B.A.H., James, S.S., Anderson, S.R.: Linear system identification versus physical modeling of lateral-longitudinal vehicle dynamics. IEEE Trans. Control Syst. Technol. 29(3), 1380–1387 (2020)
Wu, W., Zhu, S., Wang, X., Liu, H.: Closed-loop dynamic parameter identification of robot manipulators using modified Fourier series. Int. J. Adv. Robot. Syst. 9(1), 29 (2012)
Funding
The authors have not disclosed any funding.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The authors declare that they have no conflict of interest.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by the Ministry of Science and Technology under Grant No. MOST 111-2221-E-006-170.
Appendices
Appendix A Design strategy of integral factors
To illustrate how the proposed integral operator based regression model suppresses the influence of the initial condition, the first element of the matrix differential equation \(\ddot{{\textbf {Y}}}_{2,2}(t) = -(\lambda _1+\lambda _2)\dot{{\textbf {Y}}}_{2,2}(t) - \lambda _1\lambda _2{\textbf {Y}}_{2,2}(t) + \lambda _1^2\varvec{\varphi }_2(t)\) in (42) is considered:
where \(k_1=\lambda _1\lambda _2\), \(k_2=\lambda _1+\lambda _2\), and \(y_{22,1}(t)\) is the first element of the matrix \({\textbf {Y}}_{2,2}(t)\). The associated initial conditions are \(y_{22,1}(0)=0\) and \({\dot{y}}_{22,1}(0)=\lambda _1\theta _1(0)-{\dot{\theta }}_1(0)\). Define the auxiliary state variables \(\eta _1(t)=y_{22,1}(t)\) and \(\eta _2(t)={\dot{y}}_{22,1}(t)\), the state-space equation in the controllable form can be expressed by
where the initial conditions are \(\eta _1(0)=0\) and \(\eta _2(0)=\lambda _1\theta _1(0)-{\dot{\theta }}_1(0)\). The analytic solution of (96) is
where \(\varvec{\eta }(0)=[\eta _1(0),\eta _2(0)]^T\), and
For the analytic solution (97), the term \({\textbf {C}}e^{{\textbf {A}}t}\varvec{\eta }(0)\) represents the transient response of the auxiliary system (96). Another convolution term denotes the force response. If the initial condition of (96) is unknown, applying larger values of \(\lambda _{1}\) and \(\lambda _{2}\) could make the transient response \({\textbf {C}}e^{{\textbf {A}}t}\varvec{\eta }(0)\) decay faster, which is able to attenuate integration bias.
Appendix B Feasibility analysis of robust stability
The feasibility of the positive definite and symmetric matrices \({\textbf {P}}\) and \({\textbf {Q}}\) which satisfy (74) is discussed by introducing the linear matrix inequality (LMI) [6] as follows.
Factorizing \(\Delta {\textbf {A}}\) in (70) yields
where
and
Under the assumption that the uncertain matrix \(\Delta {\textbf {M}}\) is bounded, there exists a known positive constant matrix \(\varvec{\mu }\) such that
where \(\tilde{{\textbf {M}}}\tilde{{\textbf {M}}}^{T} < \varvec{\mu }^{+}\). Based on (99), the term \(\Delta {\textbf {A}}{^T}{} {\textbf {P}} + {\textbf {P}}\Delta {\textbf {A}}\) can be factorized by
Consider the well-known inequality
where \(\varepsilon >0\), and \((\textbf{X},\textbf{Y})\) are square matrices with proper dimensions. On the basis of (102) and (104), (103) satisfies the following inequality [5]
where \(\varepsilon >0\) is a decision variable to be solved.
Thus, the robust stability requirement (74) is achieved if the following inequality is guaranteed:
By Schur complement, (106) can be reformulated as the feasibility problem of the following LMIs:
As a result, given \(({\textbf {K}}_{P},{\textbf {K}}_{I},{\textbf {K}}_{D})\) and \(\varvec{\mu }\), the robust stability of the closed-loop error dynamics (64) is guaranteed if the LMIs in (107) are satisfied.
Appendix C Analysis of quantization error
The quantization error \(e_{Q}(t)\) is the error caused by using the quantized signal \(x_{Q}(t)\) instead of the true signal x(t), defined as
The quantization interval is denoted as Q. The probability density function of the quantization error is given by
By definition, the expectation of \(e_{Q}\) is
and the variance is
Hence, due to the zero-mean property, the root-mean-square error of \(e_{Q}\) can be obtained by taking the square root of \({{\,\textrm{var}\,}}{(e_{Q})}\):
The property (112) will be used in the cost function of the optimal trajectory estimation presented in (85).
If the quantized signal is used in the traditional velocity estimation, it will induce significant estimation errors. For example, estimate the velocity \({\dot{x}}(t)\) at \(t=kT\) by applying the backward difference formula:
where T is the sample interval and
It can be seen that velocity estimation error \({\dot{e}}(kT)\) is amplified by a factor of 1/T due to the numerical differentiation. Hence, if the quantized signal is used in the acceleration estimation, the estimation error will be affected by a factor of \(1/T^{2}\).
Appendix D Derivation of low-pass filter-based filtered regression model
The studies [27, 29, 34, 35] introduce the first-order low-pass filter to establish a filtered regression model to avoid differentiating the acceleration. The derivation is discussed as follows.
Rewrite (20) for the regression model as
where
and \(\phi _{1,14}\), \(\phi _{0,14}\), \(\phi _{0,24}\), \(\phi _{0,18}\), \(\phi _{0,29}\) are defined in (28). Taking the Laplace transform of (115) yields
where s is the Laplace operator and
Introduce the first-order low-pass filter as
where \(\lambda >0\) denotes the cut-off frequency of the filter. Multiplying both sides of (117) by (119) gives
where
Reformulating \(\varvec{\varphi }_1^{1f}(s)\) to eliminate \(s\varvec{\varphi }_1(s)\) produces
where
Taking the inverse Laplace transform of (120) gives the following filtered regression model:
where
The integral quantities \(\varvec{\uptau }^{1f}(t)\), \({\textbf {y}}_1(t)\), and \(\varvec{\varphi }_0^{1f}(t)\) are obtained by numerically integrating the following matrix differential equations:
The parameter estimation can be conducted by using (20) shown in Sect. 3.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Li, YR., Peng, CC. & Juang, JN. An integral method for parameter identification of a nonlinear robot subject to quantization error. Nonlinear Dyn 111, 22419–22441 (2023). https://doi.org/10.1007/s11071-023-09027-z
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11071-023-09027-z