METHOD FOR PROCESSING A MEASUREMENT SIGNAL

Abstract

A method is proposed for processing a measurement signal, in particular for a steering system. The method comprises the following steps: A measured variable is acquired based on the measurement signal, wherein the measured variable comprises items of information about a physical variable, and wherein the measured variable is a superposition of the actual value of the physical variable and the measurement noise. Filter parameters of a filter are ascertained based on the measured variable and a mathematical model of the measurement noise. The measurement signal is filtered by means of the filter, whereby an estimated value of the physical variable is obtained, wherein the filter has the ascertained filter parameters. The filter parameters are ascertained in such a way that a deviation between the estimated value of the physical variable and the actual value of the physical variable is approximated and minimized. Furthermore, a control unit for a steering system, a steering system, a computer program, and a computer-readable data carrier are disclosed.

Claims

1. A method for processing a measurement signal, in particular for a steering system (10), having the following steps: acquiring a measured variable based on the measurement signal, wherein the measured variable comprises items of information about a physical variable, and wherein the measured variable is a superposition of the actual value of the physical variable and the measurement noise; ascertaining filter parameters of a filter based on the measured variable and a mathematical model of the measurement noise; and filtering the measurement signal by means of the filter, whereby an estimated value of the physical variable is obtained, wherein the filter has the ascertained filter parameters; wherein the filter parameters are ascertained in such a way that a deviation between the estimated value of the physical variable and the actual value of the physical variable is approximated and minimized.

2. The method as claimed in claim 1, wherein the filter is a filter having finite impulse response.

3. The method as claimed in claim 1, wherein the physical variable is a steering column torque and/or the measured variable is the measured value of a torque sensor (22).

4. The method as claimed in claim 3, wherein at least one actuator (24) of the steering system (10) is controlled based on the estimated value of the physical variable.

5. The method as claimed in claim 1, wherein the measurement noise is modeled in the mathematical model as a Gaussian process.

6. The method as claimed in claim 1, wherein the measurement noise and the physical variable are uncorrelated in the mathematical model.

7. The method as claimed in claim 1, wherein the measurement noise is modeled in the mathematical model as time-independent or having a known chronological dependence.

8. The method as claimed in claim 1, wherein the filter parameters are recursively determined, in particular by means of a recursive method of least squares.

9. The method as claimed in claim 1, wherein the measured variable is only taken into consideration in a predefined time window to ascertain the filter parameters.

10. A control unit for a steering system (10) of a motor vehicle, wherein the control unit (28) is designed to carry out a method as claimed in claim 1.

11. A steering system of a motor vehicle, having a sensor (22), which is designed to acquire a measured variable, which comprises items of information about a physical variable, and a control unit (28) as claimed in claim 10.

12. A computer program having program code means in order to carry out the steps of a method as claimed in claim 1 when the computer program is executed on a computer or a corresponding processing unit, in particular a processing unit (30) of a control unit for a steering system (10) of a motor vehicle, wherein the control unit (28) is designed to carry out a method as claimed in claim 1.

13. A computer-readable data carrier, on which a computer program as claimed in claim 12 is stored.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Further advantages and properties of the invention result from the following description and the appended drawings, to which reference is made. In the figures:

[0038] FIG. 1 schematically shows a steering system according to the invention;

[0039] FIG. 2 schematically shows a flow chart of a method according to the invention for processing a measurement signal; and

[0040] FIG. 3 shows an illustration of the functionality of a computer program according to the invention.

DESCRIPTION

[0041] FIG. 1 schematically shows a steering system 10 for a motor vehicle, wherein the steering system 10 is embodied as an electromechanically assisted steering system having steering column assistance (English: “column drive EPS”).

[0042] The steering system 10 includes a steering wheel 12, which is connected via an upper part of a steering column 14 and via a steering intermediate shaft 16 to a pinion 18. The pinion 18 meshes with a toothed rack 20, so that a torque is applied to it when the driver rotates the steering wheel 12.

[0043] A torque and/or steering angle sensor 22 is arranged on the steering column 14, which is designed to measure steering torques and/or a steering angle. In particular, it is thus a steering torque and steering angle sensor, which is also referred to in English as a “torque and angle sensor (TAS)” and can provide a steering angle in addition to the steering torque.

[0044] Furthermore, an electric motor 24 is provided, which is connected via a gear 26 in a torque-transmitting manner to the steering intermediate shaft 16.

[0045] The gear 26 is designed in FIG. 1 as a worm gear. Alternatively, however, the gear 26 can be designed as a spur gear, as a bevel gear, or as any other suitable type of gear.

[0046] In any case, at least a torque which is provided by the electric motor 24 is transmitted via the gear 26 to the steering intermediate shaft 16 to execute a steering assistance.

[0047] The electric motor 24 and the torque and/or steering angle sensor 22 are each connected in a signal-transmitting manner, which is only schematically indicated in FIG. 1, to a control unit 28 of the steering system 10.

[0048] In general terms, the control unit 28 is designed, based on measurement data from the steering system 10, in particular based on measurement data of the torque and/or steering angle sensor 22, to ascertain a torque to be applied and to transmit corresponding control commands to the electric motor 24, so that the electric motor 24 provides the torque to be applied.

[0049] It is to be noted that the above-described steering system 10 having steering column assistance is only an example used for illustration. The following explanations also apply, possibly with minor modifications, to any other type of steering system, in particular to steering systems having a pinion drive (English: “single pinion drive EPS”), steering systems having dual pinion (English: “dual pinion drive EPS”), steering systems having concentric toothed rack drive via a recirculating ball nut, steering systems having a belt drive, and so-called steer-by-wire steering systems, in which there is no mechanical operational connection between the steering wheel 12 and the wheels of the motor vehicle.

[0050] In general terms, all various types of steering systems having electromechanical steering assistance share the feature that the control unit 28 detects a torque applied by the driver to the steering wheel 12 and controls the electric motor 24 based thereon in order to generate a specific assistance torque.

[0051] In this case, specific frequency components in the measurement signal are typically amplified in order to generate a specific steering feeling. However, measurement noise present in the measurement signal is thus also amplified simultaneously, which can result, for example, in undesired noise formation in the steering system 10.

[0052] The measurement noise is frequency dependent and can be characterized by means of the so-called signal-to-noise ratio (SNR). More precisely, the signal-to-noise ratio at a specific frequency co may be characterized by a ratio of the spectral power densities of the actual physical variable P.sub.phyVal(ω) and the spectral power density of the interference P.sub.noise(ω), i.e., by

[00001]$SNR\left(\omega \right)=\frac{P_{\mathrm{phyVa}l}\left(\omega \right)}{P_{\mathrm{noise}}\left(\omega \right)}.$

[0053] In order to reduce the measurement noise reliably over the entire relevant frequency range, the control unit 28 is designed to carry out the method steps explained in the following on the basis of FIGS. 2 and 3.

[0054] More precisely, the control unit 28 comprises a processing unit 30 and a data carrier 32, wherein a computer program is stored on the data carrier 32, which is executed on the processing unit 30 and comprises program code means in order to cause the steering system 10 to carry out the steps of the method explained hereinafter.

[0055] Firstly, a measurement signal is generated by means of the torque and/or steering angle sensor 22 (step S1). Depending on the embodiment of the torque and/or steering angle sensor 22, the measurement signal contains items of information about a torque which acts in the steering column 14, and/or items of information about an angle of rotation of the steering column 14.

[0056] The measurement signal is relayed to the control unit 28 and is further processed thereby. A measured variable is acquired by the control unit based on the measurement signal (step S2).

[0057] In the following, the case is explained as an example that the measured variable is a measured torque T.sub.column,meas acting in the steering column 14. Accordingly, the underlying physical variable is the actual torque T.sub.column acting in the steering column 14. The measured torque is in this case a superposition of the actual torque T.sub.column and the measurement noise ν.sub.meas, and the following thus applies

T.sub.column,meas=T.sub.column+ν.sub.meas.

[0058] The measured variable T.sub.column,meas can be measured directly by the torque sensor 22. For example, the torque T.sub.column,meas acting in the steering column 14 is determined from a torsion angle of a torsion bar of the steering column 14.

[0059] Filter parameters p of a filter are now ascertained based on the measured torque T.sub.column,meas and a mathematical model of the measurement noise ν.sub.meas (step S3).

[0060] The filter in the exemplary embodiment described hereinafter is a filter having finite impulse response.

[0061] However, it would also be conceivable that a filter having infinite impulse response is used.

[0062] Here and in the following, vector-valued variables are identified by an underline.

[0063] In general terms, upon the filtering of the measured torque T.sub.column,meas by means of the filter, an estimated value {tilde over (T)}.sub.column of the physical variable T.sub.column is obtained.

[0064] Accordingly, in step S3, the filter parameters p are determined in such a way that a deviation between the estimated value {tilde over (T)}.sub.column and the physical variable T.sub.column is as small as possible, as explained in greater detail hereinafter.

[0065] At least one, in particular all of the following assumptions underlie the mathematical model:

[0066] The measurement noise can be described sufficiently accurately by a Gaussian process. That is to say, the knowledge about the process mean value and the auto-correlation function of the process are sufficient to characterize the measurement noise.

[0067] The actual value of the physical variable T.sub.column and the measurement noise ν.sub.meas are uncorrelated, i.e., the cross-correlation function E{T.sub.columnν.sub.meas} is equal to zero.

[0068] Characteristic statistical variables of the measurement noise are chronologically constant or their chronological dependence is known. In particular, the mean value and/or the auto-correlation function of the measurement noise is chronologically constant or its chronological dependence is known.

[0069] The actual value of the physical variable T.sub.column is time variant and can thus change with the time. In particular, the spectral power density of the physical variable T.sub.column is also time variant.

[0070] The functionality of the computer program is illustrated in greater detail in FIG. 3, more precisely the ascertainment of the filter parameters. As can be seen therein, the computer program includes a filter module 34 and an update module 36.

[0071] The update module 36 receives the measured variable T.sub.column,meas and ascertains, based on the mathematical model of the measurement noise v.sub.meas, updated filter parameters p for the filter module 34.

[0072] The filter module 34 consists of the above-described filter having the impulse response h.sub.filt(p) dependent on the n parameters in the filter parameter vector p, wherein n is a natural number greater than zero. The time response y(t) at the point in time t≥0 of the filter to the excitation with the measured variable T.sub.column,meas is given via convolution

y(t)=h.sub.filt(p)*T.sub.column,meas

[0073] This describes the filtering of the measured variable via the filter module 34.

[0074] By way of the filtering of the measured variable T.sub.column,meas using the filter module 34, more precisely using the filter, the estimated value {tilde over (T)}.sub.column of the physical variable T.sub.column is obtained if the adjusted filter parameters p from the update module 36 are used in the filter module 34. This means that for this purpose the general time response of the filter y(t) corresponds to the estimated value of the physical variable {tilde over (T)}.sub.column.

[0075] The filter parameters p.sup.T=[p.sub.0, p.sub.1, . . . , p.sub.n-1] of the filter are adjusted by the update module 36 in each data acquisition step.

[0076] The adjustment of the filter parameters p, takes place in that a set of filter parameters p is ascertained which minimizes the quality function

J=E{({tilde over (T)}.sub.column−T.sub.column).sup.2}=E{(h.sub.filt(p)*T.sub.column,meas−T.sub.column).sup.2}

[0077] which is dependent on a deviation between the estimated value {tilde over (T)}.sub.column and the physical variable T.sub.column.

[0078] The variance of the error between the estimated value {tilde over (T)}.sub.column and the physical variable T.sub.column is used as the quality function J.

[0079] The quality function J can be understood as a function of the filter parameters p, it is thus J=J(p.sub.0, p.sub.1, . . . , p.sub.n). The filter parameters p are determined so that the quality function J assumes a minimum.

[0080] For this purpose, the equation

[00002]$\frac{d}{d\underset{\_}{p}}J=\underset{\_}{0}$

is solved and generally supplies as a condition for the minimum of the quality function J

[00003]$\frac{\mathrm{dJ}}{d\underset{\_}{p}}=\frac{d}{d\underset{\_}{p}}E\left\{{\left(h_{\mathrm{filt}}\left(\underset{\_}{p}\right)*T_{\mathrm{column},\mathrm{meas}}-T_{\mathrm{column}}\right)}^2\right\}=E\left\{2\left(h_{\mathrm{filt}}\left(\underset{\_}{p}\right)*T_{\mathrm{column},\mathrm{meas}}-T_{\mathrm{column}}\right).Math.\frac{d}{d\underset{\_}{p}}\left(h_{\mathrm{filt}}\left(\underset{\_}{p}\right)*T_{\mathrm{column},\mathrm{meas}}\right)\right\}.$

[0081] One possibility for implementation is represented by the use of a filter having finite impulse response. The time response of the filter at an arbitrary sampling point in time t.sub.k of the processing unit 30 may then be specified by

{tilde over (T)}.sub.column(t.sub.k)=h.sub.filt(p)*T.sub.column,meas=p.sup.T.Math.T.sub.column,meas(t.sub.k)

[0082] The vector T.sub.column,meas is given here by

[00004]${\underset{\_}{T}}_{\mathrm{column},\mathrm{meas}}\left(t_k\right)=\left[\begin{array}{c}{T}_{\mathrm{column},\mathrm{meas}}\left(t_k\right)\\ T_{\mathrm{column},\mathrm{meas}}\left(t_{k-1}\right)\\ T_{\mathrm{column},\mathrm{meas}}\left(t_{k-2}\right)\\ \mathrm{.Math.}\\ T_{\mathrm{column},\mathrm{meas}}\left(t_{k-n+1}\right)\end{array}\right]$

[0083] where t.sub.k-m−t.sub.k-m-1=t.sub.s and t.sub.k-m:=t.sub.k−m.Math.t.sub.s, wherein t.sub.s represents the sampling time of the processing unit 30 and m is a natural number greater than or equal to zero.

[0084] The solution to the equation

[00005]$\frac{d}{d\underset{\_}{p}}J=\underset{\_}{0}$

supplies, upon use of a titter having finite impulse response

−2E{T.sub.column,measT.sub.column}.sup.−1.Math.E{T.sub.column,measT.sub.column,meas.sup.T}p=0

The set of filter parameters p, which fulfills this condition is give by

p=E{T.sub.column,measT.sup.T.sub.column,meas}.sup.−1E{T.sub.column,measT.sub.column}.

[0085] The filter parameters p are also dependent on the actual value T.sub.column, which is not known by way of measurements. If it is presumed that the physical variable T.sub.column and the measurement noise ν.sub.meas are uncorrelated, it then follows for the cross-correlation function

E{T.sub.column,measT.sub.column}=E{T.sub.column,measT.sub.column,meas}−E{vv}.

[0086] In contrast, the auto-correlation function E{vv} of the measurement noise is known on the basis of the assumptions i), ii), and iii), so that the set of filter parameters p, which minimizes the quality function J, can be determined with

p=E{T.sub.column,measT.sup.T.sub.column,meas}.sup.−1(E{T.sub.column,measT.sub.column,meas}−E{vv}).

[0087] Alternatively, this can be expressed by the auto-correlation matrix R.sub.TT of the measured variable T.sub.column,meas and the auto-correlation r.sub.TT and r.sub.νν of the measured variable T.sub.column,meas or of the noise ν, respectively, specifically as

[00006]$\underset{\_}{p}={\underset{\_}{R}}_{\mathrm{TT}}^{-1}\left({\underset{\_}{r}}_{\mathrm{TT}}-{\underset{\_}{r}}_{\mathrm{vv}}\right)={\underset{\_}{R}}_{\mathrm{TT}}^{-1}{\underset{\_}{r}}_{\mathrm{TT}}-\underset{\mathrm{adjustment}}{{\underset{\_}{R}}_{\mathrm{TT}}^{-1}{\underset{\_}{r}}_{\mathrm{vv}}}.$

[0088] The first term R.sub.TT.sup.−1r.sub.TT corresponds to a solution in the absence of noise, while the second term R.sub.TT.sup.−1r.sub.vv represents an adjustment of the solution due to additive noise.

[0089] Preferably, the filter parameters p are not determined again using the above equation in each time step from previously stored measurement signal data. Rather, the filter parameters are calculated recursively according to the above equation, in particular by means of a recursive method of least squares. This means that a new set of filter parameters is determined from the filter parameters ascertained in the previous time step and a current value of the measured variable.

[0090] Optionally, in the above-described method, the measured variable can only be taken into consideration within a predefined time window, so that data points of the measured variable which are too far in the past are no longer taken into consideration. The filter is thus adaptively adjusted, whereby the time variance of the physical variable is taken into consideration.

[0091] After the filter parameters have been adjusted by means of the above-described method, the measurement signal is filtered by means of the filter, whereby the estimated value {tilde over (T)}.sub.column having the minimal deviation from the actual value T.sub.column is obtained (step S4).

[0092] Based on the estimated value {tilde over (T)}.sub.column, at least one actuator of the steering system 10 is controlled (step S5). In particular, the electric motor 24 is controlled based on the estimated value {tilde over (T)}.sub.column.

[0093] By means of the above-described method, the measurement noise is reliably reduced over the entire relevant frequency range, without negatively affecting the steering feeling or the robustness and stability properties of a control loop.