Access the full text.
Sign up today, get DeepDyve free for 14 days.
References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.
Mathematical and Computer Modelling of Dynamical Systems Vol. 11, No. 2, June 2005, 195 – 207 Structured SM Identiﬁcation of Vehicle Vertical Dynamics { { { M. MILANESE* , C. NOVARA AND L. PIVANO Dipartimento di Automatic e Informatica, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy. Department of Engineering Cybernetics, Norwegian University of Science and Technology, O.S.Bragstadz plass 20, No-7491 Trondheim, Norway. In this paper the problem of identifying discrete time nonlinear systems in regression form from ﬁnite and noise corrupted measurements is considered. According to the speciﬁcations about identiﬁcation accuracy that may be needed, a good exploration of the regressor domain of interest has to be ensured by the experimental conditions. This problem becomes very signiﬁcant for growing dimension of the regressor space, leading very easily to computational complexity problems and to inaccurate identiﬁed models. These diﬃculties are signiﬁcantly reduced if, using information about the physical structure of the system to be identiﬁed, this can be decomposed into interacting subsystems. Using this structural information, the high- dimensional identiﬁcation problem may be reduced to the identiﬁcation of lower dimensional subsystems and to the estimation of their interactions. Typical cases considered in the literature are Hammerstein, Wiener and Lur’e systems, but the paper shows that the approach can be extended to more complex structures composed of many subsystems and with nonlinear dynamic blocks, using as an example the identiﬁcation of a half-car model for vehicle vertical dynamics, where nonlinear suspensions and tyres are considered. Assuming that the road proﬁle is given and that front and rear vertical accelerations are measured, an experimental setup easily realizable in actual experiments on real cars, the half-car model, is decomposed as a generalized Lur’e system, consisting of a linear MIMO system, connected in a feedback form with the two nonlinear dynamic systems through non-measured signals. An iterative identiﬁcation scheme is proposed, which makes use of a set membership method for the identiﬁcation of the nonlinear dynamic blocks. This method does not require assumptions on the functional form of the involved nonlinearities, thus circumventing the identiﬁcation accuracy problems that may be generated by considering approximate functional forms. The numerical results demonstrate the eﬀectiveness of the proposed approach. Keywords: Structured identiﬁcation; Set membership identiﬁcation; Semi-active suspensions; Ride comfort models 1. Introduction 1 m Consider a discrete nonlinear dynamic system with m inputs u ; ... ; u e and q outputs t t y ; ... ; y of the form: k k k y ¼ f w ; k ¼ 1; ... ; q tþ1 o t ð1Þ k k k 1 1 m m T n w ¼½y ... y u ... u ... u ... u 2< t t tn þ1 t tn þ1 t tn þ1 y 1 m *Email: mario.milanese@polito.it Mathematical and Computer Modelling of Dynamical Systems ISSN 1387-3954 print/ISSN 1744-5051 online ª 2005 Taylor & Francis Group Ltd http://www.tandf.co.uk/journals DOI: 10.1080/13873950500068849 196 M. Milanese et al. k n Suppose that the functions f : < !< are not known, but a set of noise corrupted k k measurements of ye and we ; k=1, ..., q, t = 1,2,. . .T is available, and it is of interest to t t k k identify the system, i.e. to ﬁnd estimates f of f , possibly giving ‘small’ identiﬁcation error. According to the speciﬁcations about identiﬁcation accuracy that may be needed, a good exploration of the regressor domain of interest has to be ensured by the set of measured regressors we ; k ¼ 1; ... ; q; t ¼ 1; ... ; T. However, exciting a system in order to obtain such a good exploration is still an open problem in nonlinear systems identiﬁcation theory. This problem becomes very signiﬁcant for growing dimension of the regressor space, leading very easily to computational complexity problems and to inaccurate identiﬁed models. These diﬃculties can be signiﬁcantly reduced if, using information about the physical structure of the system to be identiﬁed, this can be decomposed into interacting subsystems. Using this structural information, the high- dimensional identiﬁcation problem may be reduced to the identiﬁcation of lower dimensional subsystems and to the estimation of their interactions. Typical cases considered in the literature are Hammerstein, Wiener and Lur’e systems, consisting of two subsystems, a linear dynamic one and the other nonlinear static, connected in cascade or feedback form. In practical applications more complex structures may be needed, e.g. composed of many subsystems and with nonlinear dynamic blocks. In this paper we propose an approach able to deal with more complex structures. A key feature is that the nonlinear subsystems may also be dynamic and are not supposed to have a given parametric form. Indeed, a method recently developed in [1,2], not requiring assumptions on the functional form of involved regression functions, but assuming only some information on their regularity, is used for nonlinear subsystems identiﬁcation. In this way the complexity/accuracy problems [3,4] posed by the proper choice of the suitable parametrization of the nonlinear subsystems are circumvented. Since the decomposition structure is highly dependent on the speciﬁc system to be studied, we consider here as a paradigmatic example the problem of identifying a half-car model for vehicle vertical dynamics. Identiﬁcation is performed on simulated data obtained by a half-car model where the chassis, the engine and the wheels are simulated as rigid bodies, and the suspensions and the tyres are simulated as static nonlinearities. The paper is organized as follows. In section 2 the nonlinear set membership identiﬁcation method is summarized. In section 3 unstructured identiﬁcation of vehicle vertical dynamics is performed. In section 4 structured identiﬁcation of vehicle vertical dynamics is performed. In section 5 some concluding remarks are made. 2. Nonlinear SM identiﬁcation In this section the nonlinear SM identiﬁcation method developed in [1,2] is summarized. Consider the system (1) for the case k = 1 and omit the index k for simplicity. Suppose that a set of noise corrupted data Y ¼½ye ; t ¼ 1; ... ; T and tþ1 W ¼½we ; t ¼ 1; ... ; T generated by (1) is available. Then: T t ~ e y ¼ f ðw Þþ d ; t ¼ 1; ... ; T ð2Þ o t t tþ1 where the term d accounts for the fact that y and w are not exactly known. t t+1 t Structured SM identiﬁcation of vehicles vertical dynamics 197 e e The aim is to derive an estimate f of f from available measurements ðY ; W Þ, i.e. o T T e e f ¼ fðY ; W Þ. The operator f, called the identiﬁcation algorithm, should be chosen T T ^ ^ to give small (possibly minimal) error eðfÞ¼jjf fjj , where jjjj is the standard L 1 o p p : : norm jjfjj ¼ jj fwðÞ dw , p2[1,?), jjfjj ¼ ess-sup jf(w)j and W is a bounded w2W p W 1 subset of < . This error is not known, since from available data it is only known that e e f 2 FðY ; W Þ, the set of all f that can have generated the data. This set, even in case o T T of exact measurements, is unbounded, since the mapping generating data from given f is not injective. Then, whatever algorithm f is chosen, no information on the identiﬁcation error can be derived, unless some assumptions are made on the function f and the noise d. The typical approach in the literature is to assume a ﬁnitely parametrized functional form for f (linear, bilinear, neural network, etc.) and statistical models on the noise [3,4,5,6]. In the present SM approach, diﬀerent and somewhat weaker assumptions are taken, not requiring the selection of a functional form for f , but related to its rate of variation. Moreover, the noise sequence D =[d ,d , ... d ] is only supposed to be bounded. T 1 2 T Prior assumptions on f : 1 0 f 2 K ¼ f 2 C ðWÞ :kk f ðwÞ g; 8w 2 W Prior assumptions on noise: D 2 D ¼fg ½d ; ... ; d : jd j E ; t ¼ 1; 2; ... ; T T 1 T t t qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ Here, f’(w) denotes the gradient of f(w) and kk x ¼ x is the Euclidean norm. i¼1 i A key role in this set membership framework is played by the feasible systems set, often called the ‘unfalsiﬁed systems set’, i.e. the set of all systems consistent with prior information and measured data. Deﬁnition 1. Feasible systems set The feasible systems set FSS is: FSS ¼ ff 2 K : y ~ fðÞ we E ; t ¼ 1; 2; ... ; Tgð3Þ T t t tþ1 The feasible systems set FSS summarizes all the information on the mechanism generating the data that is available up to time T. If prior assumptions are ‘true’, then f 2 FSS , an important property for evaluating the accuracy of inferences that can be o T done on the system. As typical in any identiﬁcation theory, the problem of checking the validity of prior assumptions arises. The only thing that can actually be done is to check if prior assumptions are invalidated by data, determining if no system exists consistent with data and assumptions, i.e. if FSS is empty. Indeed the fact that the priors are consistent with the present data, i.e. FSS „ Ø, does not exclude the possibility that they may not be consistent with future data. However, it is usual to introduce the concept of prior assumption validation as follows. Deﬁnition 2. Validation of prior assumptions Prior assumptions are considered validated if: FSS 6¼; T 198 M. Milanese et al. Necessary and suﬃcient condition for checking the validity of assumptions are now given. Let us deﬁne the functions: f ðÞ w ¼ min h þgkk w we u t t t¼1;...;T ð4Þ fðÞ w ¼ maxðÞ h gkk w w l t t¼1;...;T : : where h ¼ ye þE and h ¼ ye E . t t t tþ1 t tþ1 Theorem 1. (i) f ðÞ w h ; t ¼ 1; 2; ... ; T, is a necessary condition for prior assumptions to be u t validated. (ii) f ðÞ w 4 h ; t ¼ 1; 2; ... ; T, is a suﬃcient condition for prior assumptions to be u t validated. Proof. See [1]. Note that there is essentially no ‘gap’ between the necessary and suﬃcient conditions, since the condition fðÞ we h þd, t = 1,2,. . .,T, is suﬃcient for any d4 0 arbitrarily small and necessary for d = 0. In this paper it is assumed that the suﬃcient condition holds. If not, values of the constants appearing in the assumptions on the function f and on the noise d are accordingly changed. The validation Theorem 1 can o t be used for assessing the values of such constants such that the suﬃcient condition holds (see [7,8]). An identiﬁcation algorithm f is an operator mapping all available information e e about the function f , noise d,data ðY ; W Þ until time T, summarized by FSS , into o T T T an estimate f 2 L ðWÞ of the function f : fðÞ FSS ¼ f ’ f T o The related L error is: ^ ^ eðfÞ¼ eðÞ fðÞ FSS ¼ f f T o This error cannot be exactly computed, sinceit is only known that f 2 FSS , but its o T ^ ^ tightest bound is given by eðfÞ sup f f . This motivates the following f2FSS deﬁnition of the identiﬁcation error, often indicated as worst-case or guaranteed error. Deﬁnition 3. Identiﬁcation error The identiﬁcation error of f ¼ fðÞ FSS is: ^ ^ E½ fðÞ FSS ¼ EðfÞ¼ sup f f f2FSS Looking for algorithms that minimize the identiﬁcation error leads to the following optimality concepts. Deﬁnition 4. Optimal algorithm An algorithm f* is called optimal if: Structured SM identiﬁcation of vehicles vertical dynamics 199 E½ fðÞ FSS ¼ inf E½ fðÞ FSS T T ¼ inf sup f f f2FSS ¼ r The quantity r , called the radius of information, gives the minimal identiﬁcation error that can be guaranteed by any estimate based on the available information up to time T. An optimal algorithm is provided by the following result. Let the function f be deﬁned as: f ðÞ w ¼½ fðÞ w þ f ðÞ w c l u where f (w)and f (w) are deﬁned in (4). l u Theorem 2. For any L (W) norm, with p2 [1,?]: (i) The identiﬁcation algorithm f (FSS )= f is optimal; c T c (ii) EðÞ f ¼ f f ¼ r ¼ inf E½ fðÞ FSS . c I f T 2 p Proof. See [2]. So far a global bound on f ðwÞ over all W is assumed. However, a local approach can be taken in order to obtain improvements in identiﬁcation accuracy, e.g. by assuming diﬀerent bounds g on suitable partitions W of W. This is similar to what is k k done in identiﬁcation of the piecewise linear model, where partitions W W are looked for, over which f (w) can be considered approximately linear, i.e. f ðwÞ’ const., Vw2 W (see, e.g. [9] and [10]). However, ﬁnding such partitions may not be an easy task. A very simple alternative approach allowing the use of local assumptions on f is based on the evaluation of a function f approximating f (using any desired method) a o and on the application of the method described in this paper to the residue function t+1 tþ1 t f ðÞ w ¼ f ðÞ w f ðÞ w using the set of values Dy = ye f we ; i = 1,2,. . ., T. D o a a Then, the estimate: L c f ðÞ w ¼ f ðÞ w þ f ðÞ w ð5Þ c D t+1 is used, where f ðÞ w is the central estimate of f (w) obtained from data Dy , t = 1,2,. . .,T. 0 0 0 Assuming a global bound f ðÞ w ¼jf ðÞ w f ðwÞjj g on the residue function D o a D 0 0 0 f implies the local bound f ðÞ w g f ðÞ w jjf ðwÞjj þ g for the function f . D o a D o a D If the function f is chosen as the estimate within a parametric model family, the present ‘local’ approach allows us to investigate the eﬀects of neglected dynamics and of possible trapping in local minima during the parameter estimation phase. In fact, if f and f ðÞ w have comparable identiﬁcation accuracy, a conﬁrmation is obtained that the chosen model family is suﬃciently rich to accurately approximate f and that a ‘good’ minimum, if not the global one, is reached. On the other hand, the model f ðÞ w may give accuracy improvements over model f in case the chosen model family is not suﬃciently rich and/or the minimization process got stuck in a local minimum. 200 M. Milanese et al. 3. Unstructured identiﬁcation of vehicle vertical dynamics Models of vehicle vertical dynamics are very important tools in the automotive ﬁeld, especially in view of the increasing diﬀusion of controlled suspension systems [11,12]. Indeed, accurate models may allow eﬃcient tuning of control algorithms in a computer simulation environment, thus signiﬁcantly reducing the expensive in-vehicle tuning eﬀort. Structured and unstructured identiﬁcations are performed on simulated data obtained by the half-car model shown in ﬁgure 1. The chassis, the engine and the wheels are simulated as rigid bodies. Static nonlinearities have been considered for suspensions and tyres. In particular, the force – velocity characteristic of the suspensions and the force – displacement characteristic of tyres are reported in ﬁgure 2. The half-car model, called for short the ‘true system’, has been implemented in Simulink in order to obtain data simulating a possible experimental setup, characterized by type of exciting input, experiment length, variables to be measured and accuracy of sensors. The vehicle is assumed to travel at a constant speed V = 60 km/h. The main variables describing the model are: . p and p : front and rear road proﬁles; rf rr . a and a : front and rear chassis vertical accelerations; cf cr . p and p : front and rear chassis vertical positions; cf cr . p and p : front and rear wheels vertical positions; wf wr . F and F : forces applied to chassis by front and rear suspensions; cf cr . F and F : forces applied to front and rear wheels by tyres. wf wr It is considered that the road proﬁle p (t) is known, that p ðtÞ¼ p ðt ‘=VÞ and that rf rr rf the variables a (t), a (t), can be measured with a precision of 5%. cf cr Two data sets have been generated from ‘true system’ simulation, by recording the values of p , p , a and a with a sampling time of t = 1/512 sec. The ﬁrst set, called rf rr cf cr the estimation set, is composed of 10 240 data values and corresponds to 20 seconds of ‘true system’ simulation on a random road proﬁle with amplitude 4 2.6 cm. The estimation data set, corrupted by a uniformly distributed noise of relative amplitude Figure 1. The half-car model. Structured SM identiﬁcation of vehicles vertical dynamics 201 Figure 2. (a) The force – velocity characteristic of suspensions. (b) The force – displacement characteristic of tyres. 5%, has been used for model identiﬁcation. The second set, called the validation set, composed of 2049 data values, corresponding to 4 seconds of ‘true’ data not used for estimation, has been used to test the simulation accuracy of identiﬁed models. The experimental setup simulated here has been chosen because it is not too complex to be realized in an actual experiment on a real car. A richer setup could give better identiﬁcation results. For example, as will be clear in the sequel, if the forces F and F cf cr are also measured, the identiﬁcation procedure becomes much simpler and more accurate. However, accurately measuring such forces requires more complex instrumentation of the car. Discrete time models, relating front chassis accelerations to the road proﬁle at the sampling times, have been identiﬁed from the estimation data set. In particular, an overall nonlinear model, indicated as NSMU (nonlinear set membership unstructured), not using information on the system structure, has been identiﬁed using the set membership approach presented in [1,13]. 1 2 1 The model is of the form (1) with q=2, m=2, y ¼ a ðttÞ, y ¼ a ðttÞ u ¼ p ðÞ tt , cf cr rf t t t and u ¼ p ðÞ tt . The choice of n ,n , n has been made by considering that the system rr y 1 2 has order 8. Models with values of n ,n , n between 5 and 10 have been identiﬁed and y 1 2 the best one has been chosen, having n =8, n =4, n =4. y 1 2 The simulation results of such a model have been tested on the validation data set. Because of space limitations, results related to the front acceleration are reported, but those related to the rear acceleration are similar. In ﬁgure 3 ‘true’ data and those obtained by the identiﬁed NSMU model are reported. The root mean square simulation error on the validation data set is reported in table 1 below. 202 M. Milanese et al. The simulation accuracy appears to be not quite satisfactory. In order to check that this is not due to poor performance of the identiﬁcation method, alternative models have been identiﬁed using the neural network toolbox of Matlab. Several two-layer neural networks with sigmoidal basis functions and with number of neuron ranging from 3 to 20 have been trained on the estimation set and the best one, having six neurons, has been chosen and is called NNU. The simulation results on the validation data set reported in table 1 suggest that by using only the available data without further information on the system, no more accurate models could be obtained. A structured identiﬁcation approach is then explored. 4. Structured identiﬁcation of vehicle vertical dynamics Considering its physical structure, the half-car model can be represented by the block diagram of ﬁgure 4. The block CE represents the behaviour of the chassis and engine. Since for usual road proﬁles the chassis pitch angles are small (5 5 – 6 degrees), this block can be considered linear. The blocks S and S represent the behaviour of the front and rear f r suspension dampers and springs. These blocks are the main sources of nonlinearities in the system, mainly due to the signiﬁcant nonlinearities of the dampers, see ﬁgure 2(a). The blocks W and W represent the inertial behaviour of the front and rear wheels and f r unsprung masses. These blocks are linear. The blocks T and T represent the behaviour f r of the front and rear tyres. These blocks are also nonlinear, see ﬁgure 2(b). The block diagram of ﬁgure 4 can be represented in a more compact form by the block diagram of ﬁgure 5, which is then used for a structured identiﬁcation procedure. Figure 3. Front chassis accelerations: ‘true’ (bold line); NSMU model (thin line). Structured SM identiﬁcation of vehicles vertical dynamics 203 In this form, the half-car model is represented as a generalized Lur’e system, consisting of the linear MIMO system CE, connected in a feedback form with the two nonlinear dynamic systems SWT and SWT , representing the overall behaviour of the f r front and rear suspensions, wheels and tyres. Note that here we have the same situation occurring in the identiﬁcation of Hammerstein, Wiener or Lur’e systems: only input and output are measured (in the present case p , p , a and a ), while the connecting rf rr cf cr variables are not measured (in the present case F and F ). A widely used method for cf cr dealing with this problem is based on an iterative scheme (see, e.g. [14,15,16,17]), which in the present case can be adapted as follows: 1. Assume an initial model for system CE. 2. Evaluate the forces F and F required to generate measured accelerations a cf cr cf and a through the current model of CE. cr 3. Identify nonlinear models of SWT and SWT using as output data the forces F f r cf and F estimated at step and as input data the measured accelerations a and a cr cf cr and known road proﬁles p and p . rf ir 4. Identify a new model of CE using as output data the measured accelerations a cf and a and as input data the forces F and F simulated by means of the cr cf cr nonlinear models of SWT and SWT identiﬁed at step 3. f r 5. If the simulation accuracy of the structured model NSMS, obtained by using, in the scheme of ﬁgure 5, the models of SWT and SWT identiﬁed at step and of f r CE identiﬁed at step 4, is not satisfactory, return to step 2. Table 1: Root mean squared front chassis acceleration errors on the validation data set. Model NSMU NNU NSMS NSMS NSMS 1 2 3 RMSE 3.27 3.37 2.56 2.41 2.05 Figure 4. Block diagram of half-car model. 204 M. Milanese et al. Figure 5. Generalized Lur’e form of half-car model. The initial model required at step 1 has been obtained from the laws of motion, assuming the chassis and engine as a unique rigid body, leading to a 26 2 transfer matrix with constant elements. Indeed, the ‘true’ 26 2 transfer matrix has elements which are transfer functions of order 4, see ﬁgure 6 where the Bode plots of elements (1,1) of these transfer matrices are reported. This undermodelling is intentionally introduced, in order to verify the eﬀects of errors in the initial guess of the iterative procedure. It can be noted that this initial model error is relevant, giving large errors ( = 100%) in the initial estimates of the forces F and F , see ﬁgure 7. cf cr In step 3, discrete time nonlinear models of SWT and SWT are identiﬁed using the r r set membership approach presented in [1,13]. The models are of the form (1) with 1 1 2 q=1, m=2, y ¼ F ðttÞ, u ¼ p ðÞ tt ,and u ¼ a ðÞ tt , where * stands for ‘f’ or ‘r’. c r c t t t The choice of n n , n has been made by considering that the systems SWT and SWT y, 1 2 f r have order 2. Models with values of n ,n , n between 1 and 6 have been identiﬁed and y 1 2 the best one has been chosen, having n =2, n =2. y 1 In step 4, the models of CE have been obtained by identifying linear output error models by means of the Matlab Identiﬁcation Toolbox, with inputs F , F and outputs cf cf a , a . cf cr The results derived by the NSMS models identiﬁed at iterations j = 1,2,3 are then compared with ‘true’ data on the testing data set. In ﬁgure 6 the Bode plots of the elements (1,1) of the ‘true’ transfer matrix CE and transfer matrices identiﬁed at the ﬁrst and third iterations are reported. In ﬁgure 7 the ‘true’ forces applied to the chassis by the front suspension and those estimated at the ﬁrst and third iteration are shown. In ﬁgure 8 the ‘true’ front acceleration and those estimated by NSMS and NSMS 1 3 are shown. In table 1 the root mean square simulation errors between ‘true’ and estimated front accelerations are reported. It can be noted that after three iterations the estimates of the transfer matrix CE and of the forces F are signiﬁcantly improved, reﬂected in c* Structured SM identiﬁcation of vehicles vertical dynamics 205 Figure 6. Bode plots of CE(1,1): ‘true’ (bold line); initial model (dashed line); model at iteration 3 (thin line). Figure 7. Forces applied to chassis from suspensions: ‘true’ (bold line); initial estimate (dashed line); estimate at iteration 3 (thin line). 206 M. Milanese et al. signiﬁcant improvements of identiﬁcation accuracy evaluated by chassis accelerations errors. Finally, it can be noted that the RMSE errors provided by the unstructured models NNU and NSMU are about 60% greater than the RMSE error provided by the structured model NSMS . 5. CONCLUSIONS Block-oriented or structured identiﬁcation is one of the most common approaches used in the literature to face the computational complexity and accuracy problems in identiﬁcation of complex nonlinear systems. Most of the literature considers decompositions with two subsystems, a linear dynamic and a nonlinear static subsystem. Moreover, the nonlinear static subsystem is supposed to have a given parametric function form (polynomial, piecewise polynomial, etc.). In this paper we have proposed an approach able to deal with more complex structures. A key feature is that the nonlinear subsystems may also be dynamic and are not supposed to have a given parametric form. In this way the complexity/accuracy problems posed by the proper choice of suitable parametrization of the nonlinear subsystems are circumvented. The eﬀectiveness of the proposed approach is tested on a simulated half-car model for vehicle vertical dynamics, where nonlinear suspensions and tyres are considered. The simulated experimental conditions allow a decomposition as a generalized Lur’e system, consisting of a linear MIMO system, connected in a feedback form with two nonlinear dynamic systems. The structured models identiﬁed by the proposed method Figure 8. Front chassis accelerations: ‘true’ (bold line); NSMS model (dashed line); NSMS model (thin 1 3 line). Structured SM identiﬁcation of vehicles vertical dynamics 207 in a few iterations reduce signiﬁcantly the simulation errors, largely improving over unstructured models and reaching quite satisfactory identiﬁcation accuracy. ACKNOWLEDGEMENTS This research was supported in part by Ministero dell’Universita` e della Ricerca Scientiﬁca e Tecnologica under the Project ‘Robustness techniques for control of uncertain systems’ and by GM-FIAT World Wide Purchasing Italia under the Project ‘Modeling of vehicle vertical accelerations’. References [1] Milanese, M. and C. Novara, C., 2002, Set membership estimation of nonlinear regressions. IFAC 2002, Barcelona, Spain. [2] Milanese, M. and C. Novara, C., 2003, Optimality in SM identiﬁcation of nonlinear systems. Proc. 13th IFAC Symposium on System Identiﬁcation SYSID 2003, Rotterdam, The Netherlands. [3] Haber, R. and Unbehauen, H., 1990, Structure identiﬁcation of nonlinear dynamic systems – a survey on input/output approaches. Automatica, 26, 651 – 677. [4] Sjoberg, J., Zhang, Q., Ljung, L., Benveniste, A., Delyon, B., Glorennec, P., Hjalmarsson, H. and A. Juditsky, A., 1995, Nonlinear black-box modeling in system identiﬁcation: a uniﬁed overview. Automatica, 31, 1691 – 1723. [5] Narendra, K. S. and Mukhopadhyay, S., 1997, Neural networks for system identiﬁcation. Sysid 97, vol. 2, pp. 763 – 770. [6] Isermann, R., Ernst, S. and Nelles, O., 1997, Identiﬁcation with dynamic neural networks – architectures, comparisons, applications. Sysid 97, vol. 3, pp. 997 – 1022. [7] Novara C. and Milanese, M., 2001, Set Membership prediction of nonlinear time series. Proc. 40th IEEE Conference on Decision and Control, Orlando, FL, pp. 2131 – 2136. [8] Milanese, M. and C. Novara, C., 2002, Nonlinear Set Membership prediction of river ﬂow. Proc. 41st IEEE Conference on Decision and Control, Las Vegas, Nevada. [9] Sontag, E. D., 1981, Nonlinear regulation. The piecewise linear approach. IEEE Trans. Automatic Control, 26, 346 – 357. [10] Ferrari-Trecate, G., Muselli, M., Liberati, D. and Morari, M., 2001, A clustering technique for the identiﬁcation of piecewise aﬃne systems. In: A. Sangiovanni-Vincentelli and M.D. Di Benedetto (Eds), Hybrid Systems: Computation and Control, Lecture Notes in Computer Science (Berlin: Springer-Verlag). [11] Krtolica, R. and Hrovat, D., 1992, Optimal active suspension control based on a half-car model: An analytical solution. IEEE Transaction on Automatic Control, 37, 2238 – 2243. [12] Lu, J. and DePoyster, M., 2002, Multiobjective optimal suspension control to achieve integrated ride and handling perfomance. IEEE Transactions on Control Systems Technology, 10, 807 – 821. [13] Novara C. and Milanese, M., 2000, Set Membership identiﬁcation of nonlinear systems. Proc. 39th IEEE Conference on Decision and Control, Sydney, Australia, pp. 2831 – 2836. [14] Narendra, K. S. and Gallman, P. G., 1966, An iterative method for the identiﬁcation of nonlinear systems using a Hammerstein model. IEEE Transaction on Automatic Control, 11, 546 – 550. [15] Stoica, P., 1981, On the convergence of an iterative algorithm used for Hammerstein system identiﬁcation. IEEE Transaction on Automatic Control, 26, 967 – 969. [16] Voros, J., 1999, Iterative algorithm for parameter identiﬁcation of Hammerstein systems with two- ¨ ¨ segment nonlinearities. IEEE Transaction on Automatic Control, 44, 967 – 969. [17] Wemhoﬀ, E, Packard, A. and Poolla, K., 1999, On the identiﬁcation of nonlinear maps in a general interconnected system. American Control Conference, San Diego, CA.
Mathematical and Computer Modelling of Dynamical Systems – Taylor & Francis
Published: Jun 1, 2005
Keywords: Structured identification; Set membership identification; Semi-active suspensions; Ride comfort models
You can share this free article with as many people as you like with the url below! We hope you enjoy this feature!
Read and print from thousands of top scholarly journals.
Already have an account? Log in
Bookmark this article. You can see your Bookmarks on your DeepDyve Library.
To save an article, log in first, or sign up for a DeepDyve account if you don’t already have one.
Copy and paste the desired citation format or use the link below to download a file formatted for EndNote
Access the full text.
Sign up today, get DeepDyve free for 14 days.
All DeepDyve websites use cookies to improve your online experience. They were placed on your computer when you launched this website. You can change your cookie settings through your browser.