DE102014211273A1 - Vehicle mass estimation method - Google Patents

Abstract

In a vehicle mass estimation method for estimating the mass of a vehicle, a plurality of mass estimation models are prepared for the vehicle, each of the models is estimated to have at least one parameter measurable on the vehicle, the measurable parameter is measured on the vehicle and the models are evaluated using a model Comparison of the estimated and measured parameters.

Description

The invention relates to a vehicle mass estimation method, by means of which it is possible to estimate the mass of a vehicle, in particular a passenger car. Furthermore, with the vehicle mass estimation method relevant here, it is also possible to estimate the mass of a team, ie a vehicle with a trailer, as well as a trailer alone.

Such an estimation of the vehicle mass and in particular also the trailer mass serves as the basis for a dynamic driving stabilization of the vehicle and possibly its trailer.

Various methods are known for estimating the vehicle mass from a vertical dynamics of the respective vehicle, which operate in particular with low-pass filters of the forces of the associated wheel suspension and with non-linear state observers (EKF).

That's how it is EP 1863659 B1 a method for determining the mass of a vehicle by means of a vertical movement between a vehicle body and the vehicle wheels, wherein the mass estimation is performed by means of a vehicle based on the vertical dynamics of the vehicle, non-linear state observer.

In EP 1430276 B1 is a method for determining the mass of a motor vehicle, taking into account different driving situations with evaluation of the respective vehicle acceleration known. In this case, in addition to the driving force of a drive unit, the respective resistance forces, resulting from rotational forces, from the air resistance, from the rolling resistance and from the slope force, and in addition braking forces are taken into account at associated friction brakes. However, such mass estimates with the vehicle acceleration, ie based on the longitudinal dynamics of the vehicle, always include the trailer mass. There are then comparatively many necessary model assumptions that can not be determined with high certainty.

Out DE 102006045305 B3 For example, a method for determining the mass of a motor vehicle by measuring the height of the suspension under calibration by a vehicle mass estimate from the longitudinal dynamics is known. When using a level control in the vehicle, however, a mass estimation on the evaluation of measurements of the ride height is not possible because just exactly the expected change in the ride height is compensated by the regulation.

The DE 19744066 B4 describes a method for detecting a trailer operation in a motor vehicle, wherein a pressure signal that detects the pressure controlled to the trailer, is evaluated. For the detection of the trailer so the trailer must be mechanically connected via a trailer hitch and in addition by means of other connections to the towing vehicle. In addition, an additional sensor for pressure measurement is necessary. A statement about the mass of the trailer can not be made.

According to US 2009/0306861 it is possible to detect a trailer by comparing the expected dynamics of the vehicle with measured sensor sizes. However, if the center of gravity of the trailer is close to a wheel axle, the influence of the trailer on the dynamics of the vehicle exists only with strong yaw rate changes. In general, the influence of the trailer on the dynamics of the vehicle is low in this configuration.

According to the invention, a vehicle mass estimation method is provided for estimating the mass of a vehicle in which a plurality of models for estimating the mass are created for the vehicle, the models are each estimated to have at least one parameter measurable on the vehicle, the measurable parameter is measured on the vehicle and an evaluation of the Models by means of a comparison of the estimated and the measured parameter takes place. In particular, in the parameterization of the models according to the invention between the various models alone, the parameter of the mass of the vehicle is set as different or respectively different.

According to the invention, the plurality of models are preferably created or executed cyclically and, in the process, a model is selected for a new estimation of the mass of the vehicle which has previously received a good rating compared to the other models. In particular, several models (all with different masses) are initialized. These are then executed cyclically (preferably according to the Kalman filter). In each step all models are evaluated. This then provides for each model at any time a rating or a probability. It can be determined by a cyclic procedure, which model has received a good rating several times, and then this model is recognized as the model of the vehicle's mass that is most likely correct. Preferably, after a certain time, a "mass region" of models is more likely to be rated than others. Then you can re-initialize, so to speak, and just in the previously found region with higher density of models. For re-estimating the mass of the vehicle, it is also preferred to delete models with the least probability from the previous models α and to create new models. As a result, a discretization in the probable range can be refined. In addition, advantageously for re-estimating the mass of the vehicle, the models are optimized by means of the previously measured parameters. The several models for estimating the mass and their evaluation then run in parallel or at the same time. Furthermore, one can advantageously for the evaluation of the models, the weighted average over at least two, preferably created over all models. The mass of the respective model and its probability or quality are taken into account for the weighted average.

In other words, according to the invention, a multi-model approach is used for mass estimation. For mass estimation, a comparison of measured and estimated parameters is performed. Thus, the quality of the models is evaluated, which are parameterized under the assumption of different masses. In particular, the mass is selected according to the model with the highest quality.

The plurality of models are preferably created according to the invention, each with a linear equation of state.

In the evaluation, advantageously a previously defined probability for the model is used as the evaluation criterion.

As a measurable parameter is particularly preferably a parameter of the vertical dynamics of the vehicle, in particular a ride height of a suspension, a spring stiffness of a suspension, a damping value of a suspension and / or a wheel acceleration of a vehicle wheel used. Alternatively or additionally, a parameter of the longitudinal dynamics or the transverse dynamics or the steering angle of the vehicle, in particular a total acceleration, is used as the measurable parameter.

With the models provided according to the invention, a partial mass of the vehicle is also preferably estimated, with preferably four multi-model approaches being used for each quarter vehicle mass. The total mass of the vehicle is then determined as a sum of the sub-masses.

In order to further improve the vehicle mass estimation method according to the invention, it is further preferably provided to take into account braking and / or acceleration processes as well as cornering and steering operations of the vehicle in the models according to the invention.

The invention is based on the finding that methods for determining chassis parameters, in particular also the vehicle mass, from the vertical dynamics are well known. However, the known methods have the particular disadvantage that nonlinear state observers are relatively complex and require comparatively much computing time. In addition, unstable or divergent state estimates may result due to insufficient quality of the parameterization.

By contrast, with the solution according to the invention, it is possible to reduce the computing time per model to approximately 1/10 in comparison with the use of non-linear state observers. The solution according to the invention is further based on the fact that many variables which are required for state estimation are precomputed by means of associated equations. Thus, these quantities do not have to be recalculated as usual in every time step or each estimation run. Furthermore, a more robust modeling and parameterization is created. The functionality of the invention also results in a built-in vehicle level control. In particular, according to the invention, no reference level, ie the height of the vehicle body with a relaxed spring of the associated wheel suspension, is required.

In an advantageous embodiment of the solution according to the invention, the weighted mean value of all model masses is weighted with the respective model quality. Well-known methods evaluate the model errors based on the variance of a normal distribution - this is not done here. According to the invention, the same variance is used for the evaluation for all models. The variance corresponds to a design parameter: the greater the variance is assumed, the slower the mass estimation converges. Well-known methods also used a normal distribution for the model errors (Kalmann filter) - this restriction is resolved by the proposed method.

Combining the mass estimate according to the invention with a well-known mass estimate from the vehicle longitudinal dynamics, the distribution of the masses (vehicle mass, trailer mass) can be determined in a vehicle-trailer combination. In such a known mass estimation method, as above, for example EP 1430276 B1 described, the mass of a team determined by the total during a braking operation or an acceleration detected inertial mass.

An exemplary embodiment of the solution according to the invention will be explained in more detail below with reference to the attached schematic drawings. It shows:

1 the operating diagram of a quarter vehicle used in accordance with the invention,

2 a sequence of the method according to the invention as a whole,

3 a sequence of the method according to the invention at one of its time steps,

4 a first graph for the inventive evaluation of the models and

5 a second graph for the evaluation of the models according to the invention.

1 illustrates the procedure according to the invention, according to which a vehicle in four parts, so-called quarter vehicles 10 , is split. To every quarter vehicle 10 includes a suspension with associated vehicle wheel occupying a height z _R. The associated vehicle body has (simplified in accordance with the invention) a partial mass m _A and a height Z _A. Between the vehicle body and the vehicle wheel is the spring and damping device of the suspension with a spring constant F _c and an attenuation value F _d .

For the dynamics of the thus reduced quarter vehicle, several linear state observers of the form become x = (z _A - z _R ; ż _A - ż _R ; z _R ; ż _R ; z ·· _R ) with the measuring vector z = (z _A - z _R ; z ·· _R ) used.

For parameterization, the parameters used are therefore, in particular the spring stiffness F _c and the damping value or damping parameter F _{d of} the vehicle suspension. In a vehicle with air springs, which can be installed on one or two axes, z. B. measured via an air pressure sensor, the internal pressure and the spring stiffness F _{c are} determined from a lookup table.

In 2 the essential steps of the procedure are illustrated. First, the vehicle is as explained in the four-quarter vehicles 10 divided. For each of the quarter vehicles 10 becomes a simulation or estimation using a multi-model approach with multiple models 12 carried out. Results 14 the estimate 12 be by means of a comparison with measurement data of the vehicle dynamics of a rating 16 subjected. Based on a probability 18 This results in the best model model and by this in an overall estimate 20 a total mass 22 of the vehicle. On the basis of the measurement data of the vehicle dynamics also carried out a correction or optimization 24 the models.

As in 3 illustrates, in this way at a time step of the method according to the invention several models 26 with the running number i, i + 1, ... for the mass mass m _A , set up. For every model 26 is based on a previous state 28 by the estimate 12 a predicted state 30 determined. This goes by means of said optimization 24 in a corrected state 32 and by rating 16 in a quality value 34 one. This becomes a model probability 36 determined by which the best model 26 is selected for mass estimation. For optimizing 24 and the rating 16 As explained, a measurement is also carried out in each time step 38 of measurable parameters of the respective model 26 considered. The measurement therefore exists once in each time step, namely by sensors on the real vehicle. This measurement is then compared to corresponding estimated states of the respective models.

Correcting or optimizing 24 the models 26 is similar to a Kalman filter: x ^ a _{priori, k} = Fx ^ _{posteriori, k-1}

z_k

ist die Messung in Zeitschritt k und H die Beobachtungsmatrix.

L_k

entspricht dem stationären Wert des Kaiman-Gains eines Viertelfahrzeuges, welches eine in etwa zu erwartende Viertelfahrzeugmasse besitzt.

Here F is the system matrix and x ^ _{posteriori, k-1} the state estimate from the last time step. x ^ _{posteriori, k-1} = x ^ _{priori, k} + L _k (z _k - Hx ^ _{priori, k)}

z _k

the measurement in time step k and H is the observation matrix.

L _k

corresponds to the stationary value of the Cayman Gain of a quarter vehicle, which possesses an approximately expected quarter vehicle mass.

The stationary value used is the "settled" Kalman gain.

• (1) P_priori,k = FP_{posteriori,k-1}F' + Q
• (2) S_k = HP_priori,kH' + R
• (3) L_k = P_priori,kHS –1 / k
• (4) P_posteriori,k = (I – L_kH)P_priori,k

errechnet werden.The stationary values can be evaluated by multiple evaluations of

• (1) P _{priori, k} = FP _{posteriori, k-1} F '+ Q
• (2) S _k = HP a _{priori, k} H '+ R
• (3) L _k = P _{priori, k} HS -1 / k
• (4) P _{posteriori, k} = (I - L _k H) P _{priori, k}

be calculated.

The Kalman filter is a state observer that, in normal use, computes (with all equations) a time-synchronized estimation of the states. Present are only the equations x ^ a _{priori, k} = Fx ^ _{posteriori, k-1} and x ^ _{posteriori, k} = x ^ _{priori, k} + L _k (z _k - Hx ^ _{priori, k)} used synchronously. The equations (1) and (4), however, are calculated only once and separately. The one-time computation is cyclically calculated until the Kalman gain is "settled".

The Kalman gain L is set constant in this procedure and can therefore be pre-calculated separately, depending on the selected process noise matrix Q and the measurement noise matrix R, and depending on the model which specifies F. Specifically, this means that a model of mass 400 kg has a slightly different F than a model of mass 500 kg. Practically, one should use a model of the mass to which suitable measurements were made. For this purpose, the equations (1) to (4), the so-called Kalman equations, are calculated in a loop until they have settled and the Kalman gain reaches a steady state value. This is then saved and can be used at any time.

A solution using the so-called Riccati equation P = F (P - PH '(HPH' + R) ^-1 HP) F '+ Q on the other hand, it is not recommended as an alternative, since due to lack of observability, the variances of z _R and ż _R diverge.

The mentioned procedure can be carried out before use in the vehicle, since the gains are independent of the measurements z _k for the Kalman filter.

Where P is _{posteriori, k-1 is} the estimated covariance matrix of the last state vector, Q is the discretized covariance matrix of the system noise, R is the measurement noise covariance matrix and / or the unit matrix.

The Kalman Gain L is the same for all models at every time step. The one-time parameter selection of Q and R does not make sure that the filter is consistent. This would be possible for a single mass, but in the case of several masses, the parameters Q and, if appropriate, R would also have to be changed for each case. Practically, however, not all the necessary series of measurements are available. All this means that the covariances computed by the Kalman filter are not for evaluation 16 the models 26 to be used.

As already too 3 In this case, a multi-model approach is carried out in which for the movement of each wheel a total of N models 26 be initialized with different masses m _i . So shows 3 a model 26 for a mass estimate i and another model 26 for a mass estimate i + 1.

mit der Messung h_k = z_k(1) und Schätzung h_i,k = x_i,k(1) des Höhenstandes z_A aus dem Modell i, gewichtet mit dem Exponent δ. Sofern beispielsweise δ = 2 angesetzt wird, entspricht dies dem Optimierungskriterium einer Least-Squares-Optimierung und die Bewertung 16 wird ähnlich zu einer Likelihood-Bewertung bzw. Wahrscheinlichkeitsbewertung (unter einer Normalverteilungsannahme). By K successive measurements z of the vertical dynamics, in this case the height z _A and preferably also a wheel acceleration, the evaluation 16 of the individual models 26 performed with the different compositional masses or sub-masses m _A. For each model i with the mass m _{i is} preferably a specially defined quality measure 34 determined for an error E _i

with the measurement h _k = z _k (1) and estimate h _{i, k} = x _{i, k} (1) of the height z _A from the model i, weighted by the exponent δ. If, for example, δ = 2 is assumed, this corresponds to the optimization criterion of a least-squares optimization and the evaluation 16 becomes similar to a likelihood score (under a normal distribution assumption).

For each model i after K measuring steps, the following preferred model probability can be used 36 define:

m ^_k = m(i_K) The vehicle mass of the quarter vehicle 10 then becomes the mass m ^ _k which is used in the model with the smallest quality measure E _{(i | K)} , or the greatest probability P _{(i | K)} , respectively:

m ^ _k = m (i _K )

To estimate the quality over the mass estimation and possible termination criteria when a correspondingly reliable estimation is achieved, the shape of the curve, as used in 4 is shown. The vertical diagram axis shows the model probability 36 above the respectively associated vehicle mass of the quarter vehicle 10 (in kg) on the horizontal diagram axis.

5 illustrates how after half the computation time the model probabilities 36 lead to a "wider" curve than at the end of the calculation (see 4 ). As time goes by, the probabilities focus on a particular area.

After a certain cycle time will be unlikely models 26 not pursued further in order to save computing capacity. In the range of high model probabilities 36 become new models 26 initialized to improve the resolution of the mass estimate.

From the N models 26 become, for example, α models (α <N-2) with the lowest probabilities 36 deleted and at the same time α new models 26 initialized in the following way:
Most probable model is on the edge (i = 1 or i = N):
Δm _new > Δm _old (width of the discretization of the masses is increased)
α new models continued the simulation on the side of the model with P _max .

Most likely model is not on the edge (1 <i <N):
Δm _new <Δm _old (width of the discretization of the masses is reduced)
α new models around the model with P _max divide the most likely intervals.

Finally, it should be noted that for the calculation of the Kalman gain - in addition to the quarter vehicle mass - the variance entries for the system noise Q and the measurement noise R can be specified. To determine the process noise parameters, the parameters of the measurement noise and the exponent δ, an optimization algorithm ("differential evolution") can be used. This determines the optimal parameters based on training data, respectively measuring runs of vehicles of known mass, based on the evaluation of how well the mass (and, among other dynamic variables) have been estimated.

The mass m _{G of} the total vehicle finally results from the sum of the four individual results m _A from the four-vehicle models:

To estimate the mass of a trailer, mass estimates from longitudinal dynamics are well known. In the known methods, if a trailer is present, the total mass of the team is estimated. By contrast, in the case of the method of mass estimation from vertical dynamics presented here, the pure vehicle mass m _G (plus the relatively small static trailer load on the towing hitch of the towing vehicle) is determined.

With m ^ _trailer = m ^ _trailer - m ^ _vehicle Thus, the mass of the trailer can be determined.

mit der Erdbeschleunigung g = 9,81 m / s² .In an area in which an existing trailer is safely assumed, the vehicle mass - and thus also the trailer mass - can still be corrected by the expected static coupling forces that have been added to the vehicle in the mass estimation by the vertical dynamics:

with the acceleration of gravity g = 9.81 m / s² ,

Systematic errors in the estimation of vehicle mass due to characteristics of the trailer (mass, moment of inertia, ...) can be considered in a second correction: m ^ _{corr, 2} = f (m ^ _{follower, corr} , ...)

LIST OF REFERENCE NUMBERS

10

district vehicle

12

estimate

14

Result

16

rating

18

probability

20

overall estimate

22

total mass

24

Optimize

26

model

28

Status

30

Status

32

Status

34

quality value

36

model probability

38

Measurement

QUOTES INCLUDE IN THE DESCRIPTION

This list of the documents listed by the applicant has been generated automatically and is included solely for the better information of the reader. The list is not part of the German patent or utility model application. The DPMA assumes no liability for any errors or omissions.

Cited patent literature

• EP 1863659 B1 [0004]
• EP 1430276 B1 [0005, 0020]
• DE 102006045305 B3 [0006]
• DE 19744066 B4 [0007]
• US 2009/0306861 [0008]

Claims (10)

Vehicle mass estimation method for estimating the mass (m _G ) of a vehicle in which a plurality of models ( 26 ) to estimate the mass, with the models ( 26 ) at least one parameter measurable on the vehicle (z _A ) is estimated, the measurable parameter (z _A ) is measured on the vehicle and a rating ( 16 ) of the models ( 26 ) by means of a comparison of the estimated and the measured parameter (z _A ). A vehicle mass estimation method according to claim 1, wherein said plurality of models ( 26 ) are generated cyclically, and for re-estimating the mass (m _G ) of the vehicle that model ( 26 ), the previous one compared to the other models ( 26 ) nice rating ( 16 ) had received. A vehicle mass estimation method according to claim 2, wherein for re-estimating the mass (m _G ) of the vehicle from the previous models ( 26 ) α models ( 26 ) with the lowest probability ( 18 ) and new models ( 26 ) to be created. Vehicle mass estimation method according to claim 2 or 3, wherein for re-estimating the mass (m _G ) of the vehicle, the models ( 26 ) are optimized by means of the previously measured parameters. A vehicle mass estimation method according to any one of claims 1 to 4, wherein the plurality of models ( 26 ) are each created with a linear equation of state. Vehicle mass estimation method according to any one of claims 1 to 5, wherein in the assessment ( 16 ) as the evaluation criterion a previously defined probability ( 18 ) is used for the model. Vehicle mass estimation method according to one of Claims 1 to 6, in which a parameter (z _A ) of the vertical dynamics of the vehicle, in particular a wheelbase (z _A ), a spring stiffness (F _c ) of a wheel suspension, an attenuation value (F _d ), is used as the measurable parameter. a wheel suspension and / or a wheel acceleration of a vehicle wheel is used. Vehicle mass estimation method according to one of claims 1 to 7, in which a parameter of the longitudinal dynamics or the lateral dynamics of the vehicle, in particular a total acceleration, is used as a measurable parameter. Vehicle mass estimation method according to one of claims 1 to 8, wherein with the models ( 26 ) a partial mass (m _A ) of the vehicle is estimated, preferably using four multi-model approaches for each quarter vehicle mass. A vehicle mass estimation method according to any one of claims 1 to 9, wherein in the models ( 26 ) Braking and / or acceleration processes and / or turning or steering operations of the vehicle are taken into account.

Images