Abstract In this paper we present an optimised 3 degrees- of-freedom non-linear dynamic model of a four-wheel- steering (4WS) vehicle. As variables, we retain the lateral velocity V, the rolling velocity p and yaw velocity r. The front steer angle d f and rear steer angle d r are considered to be linear functions of the steering wheel angle h s and of dh s =dt, the proportionality parameters being k 1f , k 2f for d f and k 1r , k 2r for d r . The parameters k 1f , k 2f , k 1r , k 2r are optimised by use of the BOX mathematical algorithm. In a rst optimisation loop we minimise the sideslip angle b of the vehicle and in a second optimisation loop we assure, that the resultant (taken in the centre of gravity of the vehicle) of all the transversal forces F y applied on the wheels of the vehicle (reaction forces contained in the road plane), give a component F yx along the longitudinal axis of the vehicle, that takes a non negative value. This assures, that the motor of the vehicle will not waste fuel to over- come resistance forces originating from the steering system of the vehicle. A numerical application is also presented for a 4WS vehicle negotiating a curve at con- stant velocity. The results are compared to those obtained by two models frequently used in the literature. The comparison testies on the superiority of our model for the application presented here. Dynamik der vierradgelenkten Fahrzeuge Zusammenfassung In diesem Beitrag wird ein optimiertes nicht-lineares dynamisches Model eines vierradgelenkten Fahrzeuges mit 3 Freiheitsgraden prasentiert. Als Variable wird die Seitenwindgeschwindigkeit V, die Nicken- winkelgeschwindigkeit p und die Gierenwinkelgeschwin- digkeit r beibehalten. Der Vorderradeinschlagwinkel d f und der Hinterradeinschlagwinkel d r werden als lineare Funktionen des Lenkradwinkels h s und dh s =dt, ange- nommen, mit Proportionalitatsparameter k 1f , k 2f fur d f und k 1r , k 2r fur d r . Die Parameter k 1f , k 2f , k 1r , k 2r sind fur den mathematischen Algorithmus BOX optimiert. In einer ersten Optimierungsschleife wird der Seitenrutschwinkel b minimiert und in einer zweiten Optimierungsschleife wird gesichert, die Resultierende (am Schwerpunkt des Fahr- zeuges angenommen), aller auf den Fahrzeugradern wir- kenden Querkraften F y (Reaktionskrafte sind in der Straenebene enthalten) eine Komponente F yx entlang der Fahrzeuglangsachse ergeben, die keinen negativen Wert annimmt. Damit wird gewahrleistet, das der Fahrzeug- motor nicht Kraftstoff verschwendet, um die Wider- standskrafte des Fahrzeuglenksystems zu uberwinden. Eine numerische Applikation fur ein Fahrzeug, das eine Kurve mit konstanter Geschwindigkeit fahrt, wird pra- sentiert. Die Resultate werden mit zwei Modellen vergli- chen, die haug in der Literatur benutzt werden. Der Vergleich bezeugt die U
berlegenheit unseres Modells fur
die hier prasentierte Applikation. List of symbols a; b distance of the front and rear axle to the centre of gravity of the sprung mass (absolute value) A x , A y longitudinal and lateral acceleration/decelera- tion respectively C f , C r aligning stiffness of a single front and of a single rear wheel respectively D roll total roll damping of the suspension F kj force acting along the k direction on the jth wheel (k = x; y; z and j = fo,,ro,ri) F yx projection on the vehicle's longitudinal axis of the resultant of all the forces that are contained in the tire-road plane and are applied on the wheels g acceleration of the gravity h ra roll axis distance from the centre of gravity of the sprung mass I xs roll moment of inertia of the sprung mass I z yaw moment of inertia of the vehicle K f , K r roll stiffness of the front and rear suspension respectively K roll total roll stiffness of the suspension k ratio of the rear steer angle to the front steer angle ` wheelbase of the vehicle m mass n transmission ratio of the steering box p roll angular velocity of the sprung mass Forschung im Ingenieurwesen 66 (2001) 260266 Springer-Verlag 2001 260 Received: 9. January 2001 K. N. Spentzas (&) Mechanical Engineering Department, National Technical University of Athens, Polytechnioupolis, Building M, 15780 Athens, Greece I. Alkhazali National Technical University of Athens, Polytechnioupolis, Building M, 15780 Athens, Greece M. Demic Faculty of Mechanical Engineering, University of Kragujevac, 34000 Kragujevac, Yugoslavia q pitch angular velocity of the sprung mass r yaw angular velocity of the vehicle t time 2t f , 2t r absolute value of the front and of the rear track width respectively U forward velocity of the vehicle V lateral velocity of the vehicle W L longitudinal weight transfer W bf , W br front and rear transversal weight transfer due to body roll W rf , W rf front and rear transversal weight transfer due to roll centre height W Tf , W Tr front and rear transversal weight transfer W uf , W ur front and rear transversal weight transfer due to the height of the un-sprung mass a j sideslip angle of the jth wheel of the vehicle b sideslip angle of the vehicle d j steer angle of the jth wheel of the vehicle h s rotation angle of the steering wheel u roll angle of the sprung mass w heading angle of the vehicle Subscripts f front wheel G centre of mass i inner wheel to the negotiated curve j jth wheel of the vehicle (j = f for front wheels, j = r for rear wheels or j = 1 for the front left wheel, j = 2 for the front right wheel, j = 3 for the rear left wheel and j = 4 for the rear right wheel) o outer wheel to the negotiated curve r rear wheel s sprung u un-sprung w wheel x; y; z along the longitudinal, the transversal and the vertical axis of the vehicle or of the wheel 1 Introduction Considerable efforts have been invested recently in the development of four-wheel-steering (4WS) vehicles, aim- ing to improve the manoeuvring ability and the safety of operation of vehicles. The following are the typical models of 4WS vehi- cles that are most frequently used in the literature [10, 8, 6, 7, 1]: Model I. A 2 degrees-of-freedom linear model (bicycle model) that retains the lateral velocity V and the yaw angular velocity r as degrees-of-freedom. The equations of motion for this model are written as follows m(U r dV=dt) = F yf F yr (1) I z (dr=dt) = a F yf b F yr : (2) Each one of the lateral forces appearing in the above equations is a linear function of the sideslip angle of the respective wheel, the linearity coefcient being the cor- nering stiffness of the tire F yf = C f a f (3) F yr = C r a r : (4) Each one of the sideslip angles is a functions of the steer angle of the respective wheel and is computed as follows a f = d f (V a r)=U (5) a r = d r (V b r)=U : (6) The following relations dene the steer angles of the wheels d f = h s =n (7) d r = k d f : (8) h s being the steering wheel angle, n being the transmission ratio of the steering box and k the proportionality coef- cient between the front steer angle and rear steer angle. From the above relations we can compute the propor- tionality coefcient k k = (b L a m V 2 =C r )=(a L b m V 2 =C f ) : (9) Model II. A 3 degrees-of-freedom non-linear model that retains as variables the lateral velocity V, the rolling ve- locity p and yaw velocity r and considers the rear wheels steer angle d r proportional to the front wheels steer angle d f . The equations of motion for model II are m(U r dV=dt) m s h ra (dp=dt) = X [F y [ (10) I z (dr=dt) = X [M z [ (11) I xs (dp=dt) m s h ra (U r dV=dt) = (m s g h ra K roll ) u D roll p : (12) A non-linear denition of the sideslip angles is retained a f = d f arctan[(V a r)=U[ (13) a r = d r arctan[(V b r)=U[ : (14) Model II makes use of the assumptions d 1 = d 2 = d f (15) d 3 = d 4 = d r (16) d r = k d f : (17) The above two models neglect two important phenomena: (a) The front wheel steer angle d f and the rear wheel steer angle d r do not depend only on the steering wheel angle h s but also on the speed with which the driver of a vehicle changes the angle h s , that is, they also depend on dh s =dt. (b) The resultant F y (taken in the centre of gravity of the vehicle) of all the transversal forces F yj applied on the wheels of the vehicle (reaction forces contained in the road plan), gives a component F yx along the longitu- dinal axis of the vehicle that may take a negative value and has an impact on the fuel consumption of the vehicle. The authors of this paper elaborated a model more so- phisticated than either of the previous ones, that takes into 261 K. N. Spentzas et al.: Dynamics of four-wheel-steering vehicles consideration the above two phenomena. This model will be called ``Model III'' throughout this paper and will be presented in the following paragraphs. 2 Dynamic model III 2.1 Assumptions Model III was elaborated on the basis of the following general principles and assumptions: (a) The dynamic model of a 4WS vehicle must be able to take into consideration all the physical phenomena that are related to the motion of the vehicle, included the phenomenon of the lateral and the longitudinal weight transfer. (b) The dynamic model must be as simple as possible, retaining no more degrees of freedom than it is nec- essary. This is favourable to a fast solution of the equations of motion. (c) The chassis of the vehicle is considered to be rigid. (d) The wheels of the vehicle are considered to remain all time in contact with the ground. (e) Without sacricing the accuracy of the model, we consider that the body of the vehicle is kinetically equivalent to the sprung mass and the two un-sprung masses of the front and rear suspensions. These three masses are interconnected by the roll axis of the ve- hicle, the later being dened as the axis that joins the front roll centre to the rear roll centre. (f) The centre of mass G of the vehicle is coincident with the origin of the axis system O; the vehicle is sym- metric about the xz plane. Consequently the inertia product I xy = 0. (g) The vehicle is moving on a smooth road. Consequently the heave velocity and the pitch velocity are zero, W = q = 0, and also P [F z [ = P [M y [ = 0. (h) The total mass of the vehicle is the sum of the sprung mass m s and the un-sprung mass m u . The co- ordinates of the centre of the sprung mass are (0, 0, h s ) and, obviously, the rolling velocity p and the pitching velocity q have a meaning only for the sprung mass. 2.2 Equations of motion Referring to the table dening the symbols used in this paper, we will write, in the time domain, the equations of motion for model III as follows m (dU=dt V r) m s h s p r = X [F x [ (18) m (dV=dt U r) m s h ra (dp=dt) = X [F y [ (19) I xs (dp=dt) m s h ra (dV=dt U r) = X [M x [ (20) I z r = X [M z [ : (21) This system of equations is derived from the general equations of motion given in bibliography [12, 4, 11] by applying the above assumptions. Given the fact, that the driver of the vehicle chooses the value of the forward velocity U, U is not an unknown. Consequently the unknowns in the system of equations of motion are reduced to three, the lateral velocity V, the rolling velocity p and the yaw velocity r. 2.3 Total forces and total moments The total forces and total moments that gure in the second members of the equations of motion are computed by use of the following relations X [F x [ = X [F j sin d j [ (22) X [F y [ = X [F j cos d j [ (23) X [M x [ = (m s g h ra K roll ) u D roll p (24) X [M z [ = a (F 1 cos d 1 F 2 cos d 2 ) b (F 3 cos d 3 F 4 cos d 4 ) t f (F 2 sin d 2 F 4 sin d 4 ) t r (F 1 sin d 1 F 3 sin d 3 ) : (25) 2.4 Sideslip angles of the wheels The sideslip angle of each wheel is considered to be a function of its steer angle. We have the following four relations a 1 = d 1 arctan[(V a r)=(U t f r)[ (26) a 2 = d 2 arctan[(V a r)=(U t f r)[ (27) a 3 = d 3 arctan[(b r V)=(U t r r)[ (28) a 4 = d 4 arctan[(b r V)=(U t r r)[ (29) Given the denition of our model, we can assume, as in model II, that d 1 = d 2 = d f (15) d 3 = d 4 = d r : (16) 2.5 Tire model In our model we make use either of the CALSPAN tire model or of the magic formula tire model [2, 9]. The magic formula tire model is the more accurate tire model now available [5]. 2.6 Weight transfers The longitudinal and the transversal weight transfers are taken into account in the computation of the normal reaction forces of the road on a wheel. The longitudinal weight transfer is due to the acceler- ation or to the deceleration of the vehicle and is given by W L = (m s g A x h m uf g A x h uf M ur g A x h ur )=2L : (30) 262 Forsch Ingenieurwes 66 (2001) The value of A x in the above relation must be entered as positive for an accelerating vehicle and negative for a de- celerating vehicle. The transversal weight transfer consists of three com- ponents, the transversal weight transfer due to body roll W b , the transversal weight transfer due to the height of the roll centre W r and the transversal weight transfer due to the height of the un-sprung mass W u . Consequently the front and rear transversal weight transfers are W Tf = W bf W rf W uf (31) W Tr = W br W rr W ur : (32) Each one of the terms guring in the last two equations can be computed from the following relations W bf = m s (dV=dt U r) h ra cos u (K f =K roll ) (1=2t f ) (33) W br = m s (dV=dt U r) h ra cos u (K f =K roll ) (1=2t r ) (34) W rf = m s (dV=dt U r) (b=`) (h f =2t f ) (35) W rr = m s (dV=dt U r) (a=`) (h r =2t r ) (36) W uf = m uf (dV=dt U r) (h f =2t f ) (37) W ur = m ur (dV=dt U r) (h r =2t r ) : (38) 2.7 Normal reaction forces on the wheels Given the computation of the weight transfers, we can compute the normal reaction forces on the wheels as fol- lows F zfo = m g b=2` W L W Tf (39) F zfi = m g b=2` W L W Tf (40) F zro = m g a=2` W L W Tr (41) F zri = m g a=2` W L W Tr : (42) 2.8 Steer angles of the wheels In the light of the results presented in a previous paper [13], the steering wheel angles are considered to be linear functions not only of the steering wheel angle h s , but also of dh s =dt d f = k 1f h s k 2f (dh s =dt) (43) d r = k 1r h s k 2r (dh s =dt) : (44) In the traditional method of steering a vehicle by a steering wheel, the driver decides on the values of k 1f and k 2f . The designer of the 4WS vehicle decides on the values of k 1r and k 2r , either by giving a x value to them or by providing a adequate algorithm (materialised by a steering system controller) that provides the optimum values for k 1r and k 2r as functions of k 1f and k 2f . 3 Numerical evaluation of model III by comparison to models I and II In the following numerical application we consider a ve- hicle moving along a given trajectory with a forward ve- locity U and a tolerance of 1:5 m. The trajectory used here is dened by the following generalised kinematical relation [14] tan(d f a f ) tan(d r a r ) = `=Rcos b : (45) We solve the equations of motion by the 4th order Runge Kutta algorithm and obtain the following functions of time: the lateral acceleration A y , the lateral velocity V, the yaw velocity r, the sideslip angle b and the projection F yx , on the longitudinal axis x of the vehicle, of the resultant (taken in the centre of gravity of the vehicle) of all the transversal forces F y applied on the wheels of the vehicle (reaction forces contained in the road plane). Moreover we make use of the Box optimisation algo- rithm and determine the optimum values of the four co- efcients k 1f , k 2f , k 1r , k 2r that minimise the sideslip angle b Fig. 1. Variation with time of the lateral velocity of the vehicle (simulation results by model I in mixed line, by model II in dashed line, by model III in full line) 263 K. N. Spentzas et al.: Dynamics of four-wheel-steering vehicles of the vehicle and also render the value of the resultant F yx non-negative. In Figs. 15, we present the functions of time V, A y , r, b and F yx computed by our model (model III) and also the respective functions computed by models I and II. By comparing these results we can conclude on the superi- ority of model III for the application presented here. 4 Conclusions In this paper we presented an optimised 3 degrees-of- freedom non-linear model of four-wheel-steering vehicles that retains as variables the lateral velocity V, the rolling velocity p and yaw velocity r. This model considers the steer angle of the front wheels d f and the steer angle of the rear wheels d r as linear functions of the steering wheel angle h s and of dh s =dt d f = k 1f h s k 2f (dh s =dt) (43) d r = k 1r h s k 2r (dh s =dt) : (44) The parameters k 1f , k 2f , k 1r , k 2r are optimised by a math- ematical algorithm that renders minimum the sideslip angle b of the vehicle and also assures that, the resultant (taken in the centre of gravity of the vehicle) of all the transversal forces applied on the wheels of the vehicle (reaction forces contained in the road plane), give a component F yx along the longitudinal axis of the vehicle that takes a non negative value. We considered a vehicle negotiating a curve at con- stant speed and computed all the motion functions by our model (model III), by the bicycle model (model I) and by a non-linear three degrees-of-freedom model (model II) that considers the rear wheel steer angle as proportional to the front wheel steer angle. Comparison Fig. 2. Variation with time of the lateral acceleration of the vehicle (simulation results by model I in mixed line, by model II in dashed line, by model III in full line) Fig. 3. Variation with time of the yaw velocity of the vehicle (simulation results by model I in mixed line, by model II in dashed line, by model III in full line) 264 Forsch Ingenieurwes 66 (2001) of the results for that case testies for the superiority of model III. References 1. Ackermann J, Sienel W (1993) Robust yaw damping of cars with front and rear wheel steering. IEEE Trans. Con. Syst. Technol. 1(1) 1520 2. Bakker E, Nyborg L, Pacejka HB (1987) Tire modelling for use in vehicle dynamics studies. Trans. SAE 2, 2.1902.204 3. Clark SK (1975) Mechanics of pneumatic tires. US Depart- ment of Transportation, National Highway Trafc Safety Administration, US Government Printing Ofce, Washington, DC 20402 4. Ellis JR (1994) Vehicle handling dynamics. MEP Ltd., London 5. Maalej AY, Guenther DA, Ellis JR (1989) Experimental de- velopment of tire force and moments models. Int. J. Vehicle Design 10(1) 3540 6. Nalecz G, Bindemann AC (1988) Investigation into the sta- bility of four-wheel steering vehicles. Int. J. Vehicle Design 9(2) 159178 7. Nalecz G, Bindemann AC (1989) Handling properties of four wheel steering vehicles. SAE technical paper No. 890080, Autotechnologies Conference and Exposition, January 2325, Monte Carlo, Monaco 8. Nagai M, Ohki M (1989) Theoretical study on active four- wheel-steering systems by virtual vehicle model following control. Int. J. Vehicle Design 10(1) 1633 9. Pacejka HB, Bakker E (1992) The magic formula tire model. Proceedings of the International Colloquium on Tire Models for Vehicle Dynamics, Delft, The Netherlands, October 2122, 1991, Edited as a supplement to Vehicle System Dynamics, vol. 21 10. Sano S, Furukawa Y, Shiraishi S (1986) Four wheel steering system with rear wheel steer angle controlled as function of steering wheel angle. SAE technical paper No. 860625 Fig. 4. Variation with time of the sideslip angle of the vehicle (simulation results by model I in mixed line, by model II in dashed line, by model III in full line) Fig. 5. Variation with time of the projection, on the vehicle's longitudinal axis, of the resultant of all the forces that are contained in the tire-road plane and are applied on the wheels (simulation results by model I in mixed line, by model II in dashed line, by model III in full line 265 K. N. Spentzas et al.: Dynamics of four-wheel-steering vehicles 11. Segel L (195657) Theoretical prediction and experimental substantiation of the response of the automobile to steering control. Proceedings of the Institution of Mechanical Engi- neers, Automotive Division 7, 310330 12. Spentzas CN (1998) Vehicle dynamics. Lecture notes, National Technical University of Athens, Athens 13. Spentzas KN, Alkhazali I (1999) Inuence of the rate of change of the steering wheel angle on the dynamic behaviour of four-wheel-steering vehicles. Proceedings of the 32nd IS- ATA, on Automotive Mechatronics, Design & Engineering 271279 14. Spentzas KN, Alkhazali I, Demic M (2001) Kinematics of four-wheel-steering vehicles. Forschung in Ingenieurwesen 66(5) 211216 266 Forsch Ingenieurwes 66 (2001)