EP1697189A1

EP1697189A1 - Method and device for determining a vehicle state

Info

Publication number: EP1697189A1
Application number: EP04804126A
Authority: EP
Inventors: Markus Raab; Alexander Stein
Original assignee: DaimlerChrysler AG
Current assignee: Mercedes Benz Group AG
Priority date: 2003-12-23
Filing date: 2004-12-21
Publication date: 2006-09-06
Also published as: DE112004002473D2; DE10360728A1; JP2007534534A; US20070156315A1; WO2005063536A1

Abstract

The invention relates to a method for determining the state of a vehicle, comprising the following steps: estimation of a first state of a vehicle (F) by means of a first vehicle model using predefined parameters (γ, γ, ay, ax ); estimation of a second state of the vehicle (F) by means of a second vehicle model using said predefined parameters (γ, γ, ay, ax ); weighted switching from the first vehicle model to the second vehicle model as the vehicle (F) moves from the first state into the second state, in accordance with at least one estimated parameter (ζ). The invention also relates to a device for determining the state of a vehicle (F).

Description

Method and device for determining a vehicle condition

The invention relates to a method and an apparatus for determining a vehicle state, and in particular to a method and an apparatus for determining vehicle states, the knowledge of which is required to stabilize a vehicle when a tilt angle is reached.

In modern motor vehicles, the influence of electrical and electronic driving safety systems, such as ESP (Electronic Stability Program), which is intended to prevent a vehicle from skidding within the framework of fixed physical limits, is increasing steadily. The yaw rate of the vehicle is regulated with said ESP system. Since, for reasons of cost, critical driving states or movement states of the vehicle are to be detected with as few sensor means as possible, efforts are made to be able to determine movement quantities or movement states on the basis of fewer measured parameters.

DE 41 23 053 discloses a method for determining at least one movement variable of a vehicle. The measured variables of a lateral acceleration and a steering angle on both vehicle axles become a transverse speed and / or a yaw rate of the vehicle or a movement quantity dependent thereon. A combination of two adaptive, equivalent Kaiman filter pairs is provided for evaluating the measured variables, the one filter pair being supplied with a measured quantity sum and the other filter pair with a measured variable difference.

DE 195 15 055 describes a driving stability control circuit with a speed-dependent change of the vehicle model, in which a setpoint value of a yaw rate is calculated with the aid of a vehicle model. In order to be able to calculate a value that is as accurate as possible both at very high and at very low speeds with the aid of the vehicle model circuit, at least two vehicle models are provided within the vehicle model circuit, to which suitable speed ranges are assigned, depending on the speed range currently used between the two models is switched. A hysteresis of the two speed threshold values at which the switch is made, and means for avoiding jumps in the output signal of the vehicle model circuit when there is a corresponding switch between the models are described.

However, the two known methods and devices just mentioned are not suitable for determining the transition from a first vehicle state to another vehicle state or movement state of the vehicle, in particular from a rolling movement into a tilting movement, in order to take appropriate countermeasures, for example by Brake intervention for stabilization, especially immanent to the system. The object on which the present invention is based is to provide a method and a device for determining a vehicle state, in particular a vehicle movement state, with which a tilting movement of a vehicle can be identified in a safe and as unambiguous way as possible.

According to the invention, this object is achieved by a method having the features of patent claim 1 and by a device for determining a vehicle state having the features of patent claim 12.

Accordingly, it is provided:

A method for determining a vehicle state with the method steps: estimating a first state of a vehicle by means of a first vehicle model on the basis of predetermined parameters; Estimating a second state of the vehicle by means of a second vehicle model based on the predetermined parameters; weighted switching from the first vehicle model to the second vehicle model during the transition of the vehicle from the first state to the second state as a function of at least one estimated parameter. (Claim 1)

- A device for determining a vehicle state, with a first estimation device for estimating a first state of a vehicle by means of a first vehicle model on the basis of predetermined parameters; with a second estimation device for estimating a second state of the vehicle by means of a second vehicle model on the basis of the predetermined parameters; with egg- ner switching device for weighted switching from the first vehicle model to the second vehicle model when the vehicle changes from the first state to the second state as a function of at least one estimated parameter. (Claim 12)

The idea on which the present invention is based essentially consists in estimating the movement states of a vehicle, in particular a roll or tilt angle, over an entire roll or tilt movement, with different vehicle models, in particular different caimans, in each case with respect to the roll movement and the tilt movement. Filters are used. The states estimated by the vehicle models are weighted and superimposed as a function of the roll or tipping behavior present, so that the transition from the estimates of the vehicle model intended for the roll movement to the estimates of the vehicle model intended for the tilt movement takes place smoothly. Above all, the aim is to ensure that there is no jump in the estimates. In other words: The roll angle or the tilt angle should be continuous over the considered range of motion of the vehicle, i.e. starting from a roll motion and transitioning into the tilting motion.

The wording “predetermined parameters” used above is to be understood as follows: these variables are those variables, depending on which the states of the vehicle are determined. These variables represent the input variables for the vehicle models or Kaiman filters. These quantities can be measurement quantities or quantities derived from measurement quantities by simple conversions. Both the vehicle model provided with respect to the rolling movement and with respect to the tilting movement use the same variables in each case to determine the states of the vehicle.

Advantageous refinements and developments of the invention can be found in the subclaims and the description with reference to the drawing.

According to a preferred development, the first vehicle model uses a first Kalman filter and the second vehicle model uses a second Kalman filter to simulate movement states of the vehicle.

According to a further preferred development, the first state of the vehicle stands for a rolling movement of the vehicle and the second state of the vehicle for a tilting movement of the vehicle, wherein a rolling movement describes a rotary movement about a longitudinal axis of the vehicle with ground contact of all wheels and a tilting movement of a rotary movement following the rolling movement corresponds to a lane when the wheels lose contact with the ground. The rolling movement and / or the tilting movement can take place about the longitudinal axis of the vehicle or about an axis oriented in the longitudinal direction of the vehicle.

According to a further preferred development, the weighted switchover from the first vehicle model to the second vehicle model initializes the second vehicle model with parameters of the state of the first vehicle model.

According to a further preferred development, the weighting takes place during the weighted switching as a function of nes estimated angle, preferably a roll or tilt angle of the vehicle. It is particularly advantageous if the weighting during the switchover takes place with a linear increase in the weighting of the second vehicle model for increasing values of the estimated angle (φ) with a simultaneous linear decrease in the weighting of the first vehicle model.

According to a further preferred development, the switchover takes place when the angle lies between a first predetermined angle value and a second predetermined angle value, the first predetermined angle value preferably describing a vehicle angle at which a first relieved wheel of a lane lifts off, and the second predetermined angle value describes a vehicle angle at which a second relieved wheel of the same lane loses contact with the ground.

According to a further preferred development, when the first state is estimated, a longitudinal inclination of the road, a cross-road inclination, a cross-road inclination rate and / or a coefficient of friction on the road are simulated and taken into account, taking into account the longitudinal inclination of the road preferably in conjunction with a detected longitudinal acceleration of the vehicle he follows.

According to a further preferred development, the longitudinal inclination of the roadway and the rate of crossroad slope are simulated by a Markov process. The roadway friction is advantageously modeled as a quasi-constant variable.

According to a further preferred development, when a tilting of the vehicle is detected as a state of motion Individual wheel brakes of the vehicle are selectively activated in order to stabilize the vehicle.

According to a further preferred development, the vehicle mass, the position of the center of gravity of the vehicle, the wheelbase, the track width and / or the roll characteristic, in particular the roll stiffness, and / or the vehicle damping are taken into account in the vehicle modeling.

According to a further preferred development, circumferential forces of individual wheels are estimated using the brake pressures per wheel provided by the vehicle and the number of wheel revolutions provided, preferably by means of a deterministic Luenberger observer system, from which a vehicle longitudinal acceleration is estimated.

According to a further preferred development, a yaw acceleration measuring device, a transverse acceleration measuring device and preferably a longitudinal acceleration measuring device and / or a roll rate measuring device are provided for providing the predetermined parameters.

The invention is explained in more detail below with reference to the exemplary embodiments given in the schematic figures of the drawing. It shows:

Fig. 1 is a schematic block diagram for explaining the operation of an embodiment of the present invention;

2 shows a schematic weighting diagram for explaining the functioning of an embodiment of the present invention;

3 shows a schematic side view of a motor vehicle; 4 shows a schematic plan view of a motor vehicle; and FIG. 5 shows a schematic rear view of a motor vehicle, in each case to explain an embodiment of the present invention.

In the figures of the drawing, the same or functionally identical elements and features have been provided with the same reference numerals, unless stated otherwise.

1 schematically shows a block diagram of a method sequence for determining a vehicle state to explain a preferred embodiment. A transverse acceleration a _y, preferably measured by an acceleration sensor, in the transverse direction of a vehicle, that is to say in the y direction, is fed to a first estimating device 10 and a second estimating device 11. Likewise, a determined yaw acceleration Ψ is also fed to a first and second estimation device 10, 11. In the estimating devices 10, 11, separate state estimates are made using a first vehicle model in the first estimating device 10 and a second vehicle model in the second estimating device 11. For vehicle modeling, different caimans are preferably used in both the first and second estimating devices 10, 11. Filters used. Both the mass m of the vehicle F and the position of the center of gravity S in the vehicle F, the wheelbase of the vehicle, the track width front and rear and the roll characteristic, that is to say in particular the roll rigidity and damping, flow into the vehicle models using the preferably individual Cayman filters of the vehicle in terms of Roll, with a. With the first vehicle model, the condition is estimated using a roll monitor.

In the second vehicle model, a tilt observer is used to estimate the state of motion in the second estimation device 11. This is followed by a weighting 12 of the state estimated by the roll observer and a separate weighting 13 of the state estimated by the tipping observer. Both correspondingly weighted movement state estimates are then added in an adding device Σ, so there is a combined state estimate 13 which corresponds to that of a combined observer. The weighting 12 of the roll observer and the weighting 13 of the tilt observer 13 in the state estimation are shown by way of example in FIG. 2.

FIG. 2 schematically shows a weighting diagram over the roll or estimated in the estimation devices 10, 11

tilt angle shown. The ordinate has a factor between 0 and 1 as a weighting factor for multiplication with the corresponding state estimate of the roll monitor or tilt monitor, that is to say of the first vehicle model or of the second vehicle model. According to FIG. 2, the weighting 12 of the roll observer runs by a factor of 1 up to that

angle value and then falls linearly between the angle value and the angle value down to 0. Accordingly, the weighting 13 of the tilt observer rises from the value 0 at the angle value | ζp, | linear to the value 1 at the angle value) <j? ₂ | on. Both weighting functions 12, 13 according to FIG. 2 can both increase as well as towards smaller | <j? | run through become. The angle values Ψ and Ψi stand for alternative

Angular values, which results in a less steep increase or decrease in the weighting functions 12, 13. So, if necessary, during a roll or tilt movement via the left wheels, i.e. another predetermined one via the left lane

Angle value Ψx Ψτ to choose than for a corresponding one

Movement over the right wheels, ie over the right lane of a vehicle. The angle is a roll angle estimated by the observer systems, where stands for an angle value at which a wheel of a track loses contact with the ground and where | ζ £> ₂ | stands for an angle value at which both wheels of a lane no longer have ground contact.

In order to stabilize a tipping movement of vehicles F with a high center of gravity, targeted brake interventions on individual wheels R of such a vehicle F, such as, for example, a truck or a transporter, can prevent these vehicles from rolling over or tipping over or rolling over within predetermined physical limits. In order for such a controller concept to work effectively, it is necessary that various vehicle states are made available to this system for analysis. However, such states can only be partially recorded or measured directly by existing sensors. It is therefore advisable to estimate the additional vehicle conditions using an observer method. The basic equation of various observer methods is: x = f {x, u) + κ (x, u) - (yy) y = h {x, u)

The difference between different observer methods lies in the calculation of the feedback matrix K (x, u), whereby according to the present preferred embodiment a Cayman filter is used which takes into account the stochastic properties of the system for the calculation of the feedback matrix K (x, u). The different Kaiman filters differ in the model equations f (x, u) and h (x, u), so that there are different feedbacks. In order to stabilize a vehicle when a tilt angle φ occurs, knowledge of the following vehicle conditions is generally required: speed in the vehicle longitudinal direction v _x , speed in the vehicle transverse direction v _y the roll or tilt angle φ, and the roll or tilt rate φ. Rolling movement is understood in the following to mean a rotary movement about a longitudinal axis of the vehicle, that is to say the x-axis, which results from the deflection of a vehicle F on one side of the track. During a roll movement, all wheels R are in contact with the ground. If a track of the vehicle is lifted off the ground, ie if all wheels on one side of the vehicle are lifted off the ground, the rotary movement about the longitudinal axis of the vehicle is referred to below as tilting movement or tilting. At this point it should be noted that the rolling movement and / or the tilting movement can take place not only about the vehicle's longitudinal axis or x-axis, but also about an axis oriented in the longitudinal direction of the vehicle.

In order to be able to observe the above-mentioned, necessary vehicle states over the entire rolling and tilting movement of a vehicle, according to a preferred embodiment two different Kalman filters are used for vehicle modeling. The first Kalman filter takes over the estimation of the driving conditions during the rolling movement, whereas the second Kalman filter estimates the states during the tilting movement for vehicle modeling. In principle, it is also possible to estimate the required vehicle conditions when using a suitable model using a single Kalman filter. The following equations of motion of the horizontal speeds form the basis for the filter device used to estimate the roll motion:

A speed change v _y in the y direction thus corresponds to the negative product of a yaw angle speed Ψ and a vehicle longitudinal speed v _x in addition to an acceleration a _y in the y direction. In addition, a change in speed i ^ in the x direction equals the product of the

Yaw rate Ψ and the vehicle speed v _y in the transverse direction plus an acceleration a _x in the longitudinal direction. If the horizontal accelerations a _y , a _x measured by sensors are used as input signals within these two equations, the following linearized system equations for the roll filter result after a transformation from a vehicle-specific to a roadway-based coordinate system or reference system: v _x = Ψv _y + g {θ + Θ) + a _x v _y = -Ψv _x - g {φ + φ) + a: ^e ' ^{, sor)}

In comparison to the system of equations (2), the product of the acceleration due to gravity g with the sum of a vehicle pitch angle θ and a road gradient nick is added for the term in the vehicle longitudinal direction. With the equation of motion in the y direction, there is a subtractive additional term as the product of the acceleration due to gravity g with the sum of the roll angle φ measured over the road surface plus the road surface slope Φ. As a further basic equation, a differential equation of the roll dynamics is used, which applies to small roll angles and results from the swirl set around the vehicle's longitudinal axis:

φ = - _k (F _V + Fsi, + m (a _z + g) φ) + M w (4)

J xx

where φ for the roll angle acceleration _Λ hs for a shift in the center of gravity, F _Sv for the front lateral force of the wheels, F _sh for the lateral force of the wheels R of the rear axle A _h , m for the vehicle mass, a _z for the acceleration in the Z direction, which is the Vertical axis in vehicle F corresponds to M _w corresponds to a rolling moment and Jx corresponds to an moment of inertia about the vehicle longitudinal axis. If the rolling moment M _w is included in this equation with:

M _w = -c _φ - φ - d _φ - φ (5)

where c _φ and d _{φ represent} predetermined constant values or, if appropriate, also dependent on the roll or tilt angle, the lateral forces of the wheels F _Sv , F _{S are expressed} by the lateral acceleration accordingly:

^F + ^F Sh = ^m ( ^a y + gΦ) (6) the linearized system equation for the roll dynamics within the vehicle model, preferably within the Kalman filter, results in:

where the term wt) stands for a time-dependent disturbance variable term, corresponding to a stochastic noise. As disturbances of the other the longitudinal roadway inclination Θ, the lateral inclination Φ, the lateral inclination rate Φ and the road friction coefficient μ are modeled ^'. The longitudinal inclination of the roadway Θ and the roadway gradient rate Φ are preferably simulated by a Markov process in accordance with a colored noise, which can be attributed to a white noise, since these two variables are stochastic, correlated variables. The road coefficient of friction μ is modeled in particular as a quasi-constant variable.

The directions or angles of the various sizes are illustrated schematically on the basis of FIGS. 3, 4, 5a and 5b. FIG. 3 shows a vehicle speed v _x in the longitudinal direction of the vehicle, which acts as an example on the center of gravity S of the vehicle, on which the weight force m - g acts radially to the center of the earth. The vehicle movement in the v _x direction is counteracted by a tire friction force which is exemplarily illustrated by the road friction coefficient μ.

A possible longitudinal inclination of the roadway with the inclination angle Θ is also apparent from the schematic side view according to FIG. 3. The schematic plan view according to FIG. 4 again shows the vehicle speed v _x in the longitudinal direction of the vehicle and a speed v _y in the transverse direction of the vehicle. In addition, a yaw rate Ψ and a yaw acceleration Ψ are exemplarily illustrated at the center of gravity S. 5a and 5b illustrate the vehicle inclination angle φ as well as the inclination angle rate φ and inclination angle acceleration φ and again the vehicle transverse speed v _y with a correspondingly shown frictional force in the opposite direction, which acts on the vehicle wheels R depending on the road friction coefficient μ. The vehicle F is aligned on the roadway B according to FIG. 5a in the horizontal direction, wherein the roadway B can also have a roadway bank angle qu.

The measurement equations of the vehicle model or Kalman filter responsible for the roll movement result from the application of the pulse set and the swirl set and are accordingly:

ar ° ^r = (F _Uv + F _uh ) / m-gθ ₊ v _aχ (8) ψ— = {l _vFsv - l _h F _{Sh +} M _B ) / J _{zz +} v

where v _a , v _a and _ψ corresponds to a measurement noise of the variables a _y ^nsor , a _x ^{s nsor} and Ψ ^w measured with a sensor. The lateral forces F _Sv and F _{sh of} the tires in the transverse direction, i.e. in the y direction, correspond to a circumferential by force Fu and Fr ^ of the tires in the vehicle's longitudinal direction, that is, in the x-direction. The lateral forces F _Sv and F _S h, each multiplied by the distance l _v and l _h between the center of gravity S and the front vehicle axis A _v and the rear vehicle axis A _h according to FIG. 3 flow into the yaw acceleration ψ ^ienw . The torque M _B corresponds to a torque acting on the circumferential forces Fu _v , i _x with a radius to the center of gravity S. Jz illustrates an moment of inertia in the z direction, that is, about the vertical axis of the vehicle F. The yaw acceleration ψ ^{se or} can be determined from the yaw rate Ψ, for example by means of a DTι_filter.

If the vehicle F changes from the rolling movement into the tilting movement according to FIG. 5b, the estimate of the states according to FIGS. 1 and 2 is transferred to the second vehicle model, in particular the second Kalman filter. To shorten the settling phase of this second filter, it is initialized with the previously estimated states of the first filter responsible for the roll motion. The transition from the estimates of the first filter responsible for the rolling motion to the estimates of the second filter responsible for the tilting motion is carried out by means of a weighted filter changeover according to FIG. 2. Within this changeover, the states estimated by both vehicle models and Kaiman filters are dependent on the Rolling or tilting angle \ φ \ weighted and then added in the addition device Σ according to FIG. 1. The weight function according to FIG. 2 is: x _g B = a, Λ - ε) + ^X ki _PP - ε O)

The two angles φ _x , φ ₂ define the range in which the weighted switchover is carried out (see FIG. 2). φ is the angle of the vehicle F at which the first wheel R of the unloaded lane lifts off, the angle φ ₂ denotes the angle at which the second wheel R of this lane also loses contact with the ground. Within this range between φ and φ _{2 there} is no clear assignment, whereas outside of this range there is a clear assignment to one of the two vehicle models, preferably Kalman filters. This uniform blending of the states from one vehicle model to the other or a filter enables a smooth transition of the state estimation to be achieved.

The basis for the system equation of the vehicle model responsible for the tilting movement, preferably the Kalman filter, is also formed by the momentum and the swirl theorem. It is noteworthy that the system equation, in contrast to the vehicle model or filter responsible for the roll motion, differs for the tilting motion over the left and right sides of the vehicle F. Even within the system equation of the second vehicle model or filter responsible for the tilting movement, highly non-linear tire forces are replaced by values from acceleration sensors. Generally speaking, the system equations of this second Kalman filter result in:

where the terms _w , w _vx and w _{φ represent} a noise component of the corresponding states and ξ, S, λ represent actual quantities. The system equations of the individual disturbance variables _w , w _vx , w _φ correspond to those of the vehicle model or Kalman filter responsible for the roll movement. However, the roadway bank slope qu and roadway bank rate Φ cannot be estimated with this filter since there is no difference between the effects of the road bank slope and the tilt angle when the vehicle T is tilting. These two disturbances are therefore not observable. Within this filter, too, the non-linearities derived from the tire characteristics are transferred to the measurement equation. The generalized measurement equations of the second filter responsible for the tilting movement result from the momentum and swirl theorem:

Ψsensor = Ψ + Vφ where θ _{0 is} a static pitch angle component and the term

t σ a gravitational acceleration component and M _{Rx i} a reverse)

Representing torque. All quantities are converted to a horizontal coordinate system, from which the sm <p, cos ^

Inferences. Instead of yaw acceleration Ψ _{se /} „ _or als

Using the measured variable, the yaw rate Ψ can be defined both as a state variable and as a measured variable. As a result, the filter equations of the rolling observer, that is to say of the first vehicle model or Kalman filter, do not become linear, but the sensor property, in particular the measurement noise, can thus be taken into account more precisely in the filter.

The circumferential forces Fu _h , the individual wheels R of the vehicle F can be estimated using the brake pressures per wheel provided by a preferably available ESP system (electronic stability program) and the knowledge of the rotational speeds of the individual wheels R. This is preferably done using a deterministic Luenberger observer. Its estimated circumferential forces Fu can in principle be used within the two vehicle models or Kalman filters to replace the longitudinal acceleration sensor for measuring the acceleration in the x direction, that is to say a _x ^nsor . In addition, with the help of the estimated circumferential forces F _υ, four additional measurement equations can be introduced within the Kalman filter. In addition, the normal forces of the individual wheels R of the vehicle F are calculated using a static model or using a dynamic model. These calculated normal forces are required for the tire model used within the two Kalman filters. With the help of the present invention, a movement state, in particular a wobbling or tipping of a vehicle based on acceleration information, an acceleration in the y-direction a _y, a yaw acceleration ψ and optionally an acceleration value in the x-direction a _x on the vehicle state, in particular the roll, can thus be determined - or tilt angle φ are closed. When modeling a truck in which the load causes a significant shift in the center of gravity, the roll rate φ is also required to simulate the vehicle conditions.

Although the present invention has been described above on the basis of preferred exemplary embodiments, it is not restricted thereto, but rather can be modified in many ways. In principle, a weighting other than the linear weighting of the corresponding vehicle models shown with reference to FIG. 2 is also conceivable during the transition. In theory, vehicle modeling can also be provided using a single Kalman filter, the parameters of which are adapted accordingly to the vehicle modeling.

In conclusion, the following should be pointed out: The in the. The terms "vehicle condition", "condition of a vehicle", "vehicle movement condition" and "movement condition" used in the present statements are all used interchangeably. For example, when determining a vehicle condition, the determination of a roll angle or tipping angle is considered in accordance with the present exemplary embodiments Vehicle movement size meant.

Claims

claims

1. Procedure for determining a vehicle condition with the following steps:

Estimating a first state of a vehicle (F) using a first vehicle model based on predetermined ones

Parameters (Ψ, Ψ, α .α _^ );

Estimating a second state of the vehicle (F) by means of a second vehicle model based on the predetermined parameters (Ψ, Ψ, a _y , a _x );

weighted switching from the first vehicle model to the second vehicle model during the transition of the vehicle (F) from the first state to the second state depending on at least one estimated parameter (φ)

2. The method according to claim 1, characterized in that the first vehicle model by means of a first Kalman filter and the second vehicle model by means of a second Kalman filter simulates movement states of the vehicle (F).

3. The method according to claim 1 or 2, characterized in that the first state of the vehicle for a rolling movement of the vehicle (F) and the second state of the vehicle for a tilting movement of the vehicle (F), wherein a rolling movement is a rotational movement about a vehicle longitudinal axis describes with all wheels contacting the ground and wherein a tilting movement corresponds to a rotary movement following the rolling movement with loss of ground contact at least of the wheels (R) of a track.

4. The method according to any one of the preceding claims, characterized in that the weighted switching from the first vehicle model to the second vehicle model, the second vehicle model with parameters (Ψ, Ψ, a _y , a _x ) of the state of the first vehicle model is initialized.

5. The method according to any one of the preceding claims, characterized in that the weighting during the weighted switchover takes place as a function of an estimated angle (φ), preferably a roll or tilt angle of the vehicle (F), in particular with an increasing value for the estimated angle (φ) linear increase in the weighting of the second vehicle model with a simultaneous linear decrease in the weighting of the first vehicle model.

6. The method according to claim 5, characterized in that the weighted switchover takes place when the angle (φ) lies between a first predetermined angle value (φ _x ) and a second predetermined angle value (φ ₂ ), preferably the first predetermined angle value ( φ) describes a vehicle angle at which a first unloaded wheel (R) lifts off a lane, and the second predetermined angle value (φ ₂ ) describes a vehicle angle at which a second unloaded wheel (R) of the same lane loses contact with the ground.

7. The method according to any one of the preceding claims, characterized in that, when estimating the first state as disturbance variables, a longitudinal inclination (Θ), a road inclination (Φ), a road inclination rate (Φ) and / or a road friction coefficient (μ) are simulated and taken into account ,

8. The method according to claim 7, characterized in that the longitudinal road gradient (bahn) and the road gradient rate (Φ) are simulated by a Markov process and the road friction (μ) are modeled as a quasi-constant variable.

9. The method according to any one of the preceding claims, characterized in that upon detection of a tipping of the vehicle (F) as a state of motion individual wheel brakes of the vehicle (F) be selectively activated to stabilize the vehicle (F).

10. The method according to any one of the preceding claims, characterized in that in the vehicle modeling, the vehicle mass (m), the position of the vehicle's center of gravity (S), the wheelbase, the track width and / or the roll characteristic, in particular the roll stiffness, and / or the vehicle damping is taken into account.

11. The method according to any one of the preceding claims, characterized in that circumferential forces of individual wheels (R), preferably by means of a deterministic Luenberger observer system, are estimated by means of the brake pressures per wheel (R) provided by the vehicle (F) and by means of the wheel revolutions provided, from which a Longitudinal vehicle acceleration (a _x ) is estimated.

12. Device for determining a vehicle state, in particular for operating a method according to one of the preceding claims, with:

a first estimation device for estimating a first state of a vehicle (F) by means of a first vehicle model on the basis of predetermined parameters

(Ψ, Ψ, a _y , a _x );

a second estimating device for estimating a second state of the vehicle (F) by means of a second Vehicle model based on the predetermined parameters (ψ, Ψ, a _y , a _x );

a switching device for weighted switching from the first vehicle model to the second vehicle model when the vehicle (F) changes from the first state to the second state as a function of at least one estimated parameter (φ).

13. The apparatus according to claim 12, characterized in that a yaw acceleration measuring device, a transverse acceleration measuring device and preferably a longitudinal acceleration measuring device and / or a roll rate measuring device is provided for providing the predetermined parameters.