DE10205971A1 - Method and device for determining offset values by a histogram method - Google Patents

Abstract

The invention relates to a method for determining the offset value of the output signal of a vehicle sensor, DOLLAR A - in which the values of the output signal of the vehicle sensor are determined, DOLLAR A - in which the value range of the output signal of the vehicle sensor is divided into at least two partial areas (intervals, classes) , DOLLAR A - in which the values of the output signal are assigned to the partial areas and DOLLAR A - in which the number variables representing the number of values of the output signal assigned to the partial areas are determined. DOLLAR A The essence of the invention is that the number sizes are determined by at least one low pass.

Description

State of the art

The invention relates to a device or a method ren method for determining a corrected offset value out.

With the adaptive cruise control ACC (ACC = Adap tive cruise control) is compliance for a vehicle a distance to the front that depends on the speed outgoing vehicle regulated. By DE 197 22 947 C1 a method is known from this area, in which under including the future course of a vehicle the ACC system is included in the ACC control. For this, the future course area will at least one before outgoing vehicle determined and then to all detec a lateral transverse offset is determined for vehicles. In the case of stationary curvature conditions of the road, i.e. H. in a straight line or in the area of constant curvature a curve, can also be easily done with the known method with the help of a yaw or yaw rate signal the future Determine the driving corridor.

DE 196 36 443 A1 describes a system for monitoring Known sensors in a vehicle. This system contains Means with which at least for two sensors, starting out of at least the signals you generated, for which Sensors who have the same defined comparison parameters are determined  the. The system also contains other means with which in Dependence on the at least determined comparison variable a reference quantity is determined. Starting from little At least the ascertained reference quantity is monitored in means at least for one sensor leads. In addition to the monitoring means, the system includes additional means with which at least one sensor Correction of the signal generated by him, at least in Ab dependency on the reference quantity.

DE 44 19 364 A1 describes a method for real-time averaging of the offset component of a measurement signal is known. The Measurement signal is in the form of digita at fixed intervals Measured values determined. The measured values are fixed using a stored duration stored in a histogram. An in the most frequently measured measured value The range is determined as the offset component of the measurement signal.

The features of the preambles of the independent claims ge hen from DE 44 19 364 A1.

Advantages of the invention

The invention is based on a method for determining the offset value of the output signal of a vehicle sensor,

• - in which the values of the output signal of the vehicle sensor be determined,
• - in which the value range of the output signal of the drive tool sensor in at least two sub-areas (intervals, Classes) is divided,
• - in which the values of the output signal the sub-areas be assigned,
• - where the number of assigned to the subareas Number quantities representing values of the output signal be determined.

The advantageous core idea of the invention is that the number sizes are determined by at least one low pass be communicated. The determination by means of a low pass offers itself especially because low-pass functions without exaggeration effort in control unit software, for example can be mented.

An advantageous embodiment is characterized in that

• - the at least two sub-areas (intervals, classes) a separate low pass is assigned,
• - With each of the low passes assigned to its sub-area nete number size determined.

This is advantageously a low pass or low-pass filters with exponential characteristics (this depends together with that low-pass filters with exponential characteristics are particularly suitable as counters).

A particularly advantageous embodiment is characterized in that

• - The low pass in its assigned sub-area the value of the off determined at a sampling time gating signal falls, with a first input signal is controlled and
• - those low passes, in their assigned value ranges the value of the Output signal does not fall, with a second input signal can be controlled,
• - The second input signal being different from the first input signal differs.

After it was determined in which part of the he average sensor value falls, the various low pass se in a simple way with different input signals can be controlled. Due to the different input signals (e.g. 0 or 1) result in different out  low pass signals. This is also desirable because the filling of the individual during the counting process Histogram classes (= partial areas) changes.

An advantageous embodiment of the invention is that

• - as a result of controlling the low pass with a first Input signal (k = 1) an increase in through this low pass number determined and
• - as a result of the activation of the low passes with a second Input signal (k = 0) a decrease in through this Number passes determined low passes.

The decrease in the number sizes in the from the filling process Low passports not currently affected ensures that not all low passes over time extremely high counters reach stands.

A further embodiment of the invention consists in

• - That determined from at least two successive Values of the output signal of the vehicle sensor a first Average is determined in a first way and
• - that instead of the values of the output signal of the driving the first type of the sub-areas be assigned.

This means that the output signals of the vehicle no longer go sors themselves, but mean values of these output signals in the histogram. This allows a filtering out of any high-frequency interference that occurs has a positive effect on the stability of the process.

For example, the vehicle sensor can be a Turn rate sensor act, which the yaw movement of the driver stuff recorded. The determined offset value can, for example for automatic distance control and / or control (ACC) be used in the motor vehicle.  

An advantageous embodiment is characterized

• - that the offset value by forming a weighted Average is determined and
• - that in the formation of the weighted mean one Subarea (interval, class) with the largest number size and at least one neighboring interval (class se) are taken into account.

In this way, the offset value can be safe and robust are determined as if only the partial area with the largest number size is considered. That has to do with it that variance effects (i.e. the spread of the Data) can also be viewed.

An advantageous embodiment is characterized that into the weighting factors of the weighted average the linear or quadratic values of the number sizes go. This procedure is very easy to implement.

This method according to the invention can be run in parallel with others Methods for determining the offset value are used. Through a combination or parallel application different ner method for determining the offset value can be for example, a more accurate corrected or averaged Offset value can be determined. This can be done, for example averaging takes place.

The vehicle sensor is a sensor that detects at least one movement of a vehicle. In the Er averaging the offset value involves at least two of the at least two different times occurred evaluations of the output signal.  

The determination of the offset value is based on a sorting pre at least two of these evaluations of the output signal.

Advantageously, in the present invention by evaluating the output signal of the vehicle sensor movement values representing the movement of the vehicle determined. The sorting process can be designed in this way be that the determined movement values according to size be sorted.

It is advantageous if at least a subset of possible movement values obtainable by evaluating the output signal (ω) of the vehicle sensor ( 21 b) is divided into at least two intervals. The determination of the offset value can then advantageously be based on the assignment of the movement values obtained by evaluating the output signal to the intervals, it being possible in particular to determine the offset value based on the number of movement values assigned to the intervals.

It is also advantageous if the evaluation of the respective number of the at least two intervals arranged movement values the interval with the largest Number of assigned movement values is determined and depending on this determined interval of the offset value is averaged.

The offset value can advantageously be formed by the formation a weighted average. In this Forming the weighted average can, for example the interval with the largest number of assigned Movement values and at least one neighboring interval be taken into account.  

In the weighting factors of the weighted average advantageously the linear or quadratic values the number of assigned movement values.

The number of assigned values can advantageously be low-pass filtering to a specific person range can be limited. It can be in a front partial design around a low pass filtering with Expo act nential characteristic.

Advantageously, at least a subset continues to be used possible by evaluating the driving output signal motion sensor obtainable movement values in at least two Intervals is divided. The determination of the offset value is based on the assignment of the evaluation of the off motion signals obtained at the intervals as well as low pass filtering.

This low-pass filter advantageously uses a Low pass with exponential characteristics used.

drawing

An embodiment of the invention is as follows Drawing shown and in the description below explained in more detail.

The drawing consists of FIGS. 1 to 7.

Fig. 1 schematically shows a vehicle control system and its input and output channels. This shows the functions relevant to the present invention.

Fig. 2 shows a special variant of the present invention for the case

• - that the vehicle sensor is a rotation rate sensor,  
• - That the determination of the offset value of the output signal of the rotation rate sensor depending on the output signal of the Yaw rate sensor itself and from the output signals other sensors is, which the steering wheel angle and the Record wheel speeds and
• - That the offset value of the output signal of the rotation rate sensor is determined by the following four methods:
  Standstill adjustment procedure, steering angle procedure, histogram procedure and regression procedure.

Fig. 3 shows in a qualitative manner the time course of a determined yaw rate offset and the associated error band.

Fig. 4 shows in the form of a flow chart, the determination of a performance interval. This is required for calculating the yaw rate offset value using the standstill adjustment method.

Fig. 5 shows a histogram in a qualitative manner, as methods of determining the offset value by the histogram is based.

Fig. 6 shows an xy plot in qualitative manner, as it lies in determining the offset value by the regression method is based.

Fig. 7 shows in a qualitative manner in the event that, in addition to the offset value, an error band is additionally determined in the various methods, as the corrected offset value is determined by averaging the minimum of all maximum values and the maximum of all minimum values.

embodiment

The invention will be described with reference to FIGS. 1 to 7. The special form of the selected exemplary embodiment - the use of the device according to the invention or the method according to the invention in a system for automatic distance control in a motor vehicle - is not intended to represent a restriction of the idea according to the invention.

In Fig. 1, a vehicle control system 16 and the input and output channels essential to the invention are shown in a schematic manner. The input signals of the vehicle control system 16 are the output signals of a vehicle sensor 10 as well as further signals from input data sources 11 not specified in greater detail on the input channels.

The vehicle control system is in place

• - from blocks 13 a,. , ., 13 n, which for determining the offset values ω _{off1 representing} the offset of the output signal of the vehicle sensor 10 . , ., ω _offn serve,
• - From block 14 , in which from the in blocks 13 a,. , ., 13 n determined offset values, ω _off1,. , ., ω _offn a corrected offset value ω _{offcorr is} determined and
• from block 12 , which includes all other functions of vehicle control system 16 .

The output signals of the vehicle control system 16 go to the m further control units 15 a,. , , 15 m. These m further control units can include, for example, in a special version the engine control unit, the ESP control unit (ESP = electronic stability program) or the transmission control. It is also conceivable that the output signals of the vehicle control system 16 are forwarded to a driver information system.

In Fig. 2 is a special form of the present invention, in the case of a system for adaptive Geschwin digkeitsregelung (ACC), respectively. Again, only the sensors essential for the invention and input and output channels are shown. The radar sensors are not shown, for example.

In this special form, the vehicle sensor 10 is designed as a rotation rate sensor 21 b. The other input sources 11 are designed as a steering wheel angle sensor 21 a and as a wheel speed sensors 21 c.

The rotation rate sensor 21 b provides an output signal ω, which represents the yaw rate, the steering wheel angle sensor 21 a provides an output signal δ, which presents the steering wheel angle re, and the wheel speed sensors 21 c provide output signals n _i , which represent the wheel speeds of the individual wheels.

These output signals ω, δ and n _i and the signal v representing the vehicle's longitudinal speed are input to the vehicle control system 16 . The longitudinal vehicle speed is generally not detected directly by a sensor, but is determined from the signals n _i supplied by the wheel speed sensors 21 c, which represent the wheel speeds of the individual wheels. This determination can take place, for example, in the control unit of a vehicle dynamics control system (ESP).

In the vehicle control system 16 to determine the offset value of the output signal is ω of the yaw rate sensor 21 b to the following four types:

• - by a standstill adjustment procedure 23 a,
• by a steering angle method 23 b,
• - by a histogram method 23 c and
• - by a regression method 23 d.

The standstill adjustment method 23 a requires as input signals the signal v representing the longitudinal vehicle speed de and the signal ω representing the yaw rate. In addition to these two variables, the steering angle method 23 b also requires the signal δ representing the steering wheel angle. The signal ω representing the yaw rate and the signal v representing the longitudinal vehicle speed are required as the input signal for the histogram method 23 c. The regression method 23 d requires as input signals the signals n _i representing the wheel speeds, the signal ω representing the yaw rate and the signal v representing the longitudinal vehicle speed.

As output signals the Stillstandsabgleichsverfah provides ren 23 a a signal ω _off1, which represents an approximate value for the yaw rate offset of the rotation rate sensor 21 b and a further signal Δω _off1, representing half the width of the associated error band. The steering angle method 23 b provides a signal ω _off2 , which represents an approximation value for the yaw rate offset of the yaw rate sensor 21 b, and a further signal Δω _off2 , which represents half the width of the associated error _band . The histogram method 23 c supplies a signal ω _off3 , which represents the yaw rate _{offset of} the yaw rate sensor 21 b, and a further signal Δω _off3 , which represents half the width of the associated error _band . The regression method 23 provides a signal d ω _off4 which tenoffset the Giarre the rotation rate sensor 21 b and represents a further signal Δω _off4 which represents the half width of the supplied hearing error band.

So far, a distinction has been made between signals and the physical quantities represented by the signals. For the sake of clarity, this distinction is no longer to be strictly maintained in the following. In particular, in the following the quantities ω, δ, n _i , v, ω _off1 , ω _off2 , ω _off3 , ω _off4 , Δω _off1 , Δω _off2 , Δω _off3 and Δω _{off4 are} no longer only the signals, but also those of the signals denote represented sizes.

The rule that the corresponding one serves to differentiate Size is preceded by the word "signal" if that means the Signal and not the physical quantity is meant.

The error band is a measure of the maximum absolute error with which the quantity representing the offset of the output signal of the rotary sensor 21 b is afflicted.

Using the standstill adjustment method as an example, this means that the true offset value of the yaw rate sensor is highly likely to _lie between the values (ω _off1 - Δω _off1 ) and (ω _off1 + Δω _off1 ). The term "half width Δω _off1 " thus also becomes clear. The same applies to the error bands in the other three processes.

From these approximate values ω _off1,. , ., ω _off4 and the associated widths of the error bands, a corrected offset value ω _{offcorr of} the yaw rate sensor is calculated in the means 24 , which represents a special configuration of the means 14 , which is compared to the quantities ω _off1,. , ., ω _{off4 is} characterized by increased accuracy.

It should be noted that the four methods used namely the standstill adjustment method, the steering angle method, the histogram method and the regression method drive different areas of validity, depending gig of driving condition. That means depending on Driving state, possibly not all four used ver  drive the sizes representing the offset value at the same time determine. For those processes that currently do not determine the values representing the offset value, the last valid, determined with this procedure, the Size representing offset value taken. At the same time the error band determined for this size was shown in time tet. That means that the associated error band all the more the longer the last valid determination of this becomes wider size representing the offset value. This Recourse to the last determined offset value is required for each procedure a storage of at least that too last determined offset value and the last determined Margin of error.

This widening of the error band is shown in Fig. 3 in a qualitative manner. In this xy diagram, the time t is plotted along the abscissa, the values ω _offi , ω _offmini and ω _{offmaxi are plotted} on the ordinate direction. In this case, ω _offi denotes the yaw rate sensor offset value determined by the i-th method. ω _offmini is the minimum value defined by the error _band , ie ω _offmini = ω _offi -Δω _offi and ω _offmaxi is the maximum value defined by the error _band , ie ω _offmaxi = ω _offi + Δω _offi . It should also be pointed out that it is entirely possible to use different Δω _offi for the determination of ω _offmini and ω _offmaxi from ω _offi , ie ω _offi is no longer exactly in the middle between ω _offmini and ω _offmaxi . This can also be achieved by introducing weighting factors, ie ω _offmini = ω _offi -c ₁ .Δω _offi and ω _offmaxi = ω _offi + c ₂ .Δω _offi . C ₁ and c _{2 are} different weighting factors, which can be constant, but can also depend on parameters such as time.

The validity range for the i-th method extends from time t = 0 to time t = t ₁ . For times t <t ₁ , the validity requirements are no longer met. For this reason, no further offset values are determined for t <t ₁ , but the offset value determined at time t ₁ is used. The error _band , given by the values ω _offmini and ω _offmaxi , is then continuously expanded for t <t ₁ .

The determined corrected offset value ω _offkorr is fed to a further block 12 , in which further functions of the vehicle control system are implemented. In addition to other variables not shown in FIG. 2, block 12 also has the variable ω _offcorr as an input signal.

In the following, blocks 23 a, 23 b, 23 c and 23 d are discussed in more detail, which represent the variables ω _off1 , _{... Representing} the offset of the output signal of the rotation rate sensor 21 b. , ., deliver ω _off4 .

The standstill compensation process 23 a is independent of the instantaneous movement state of the vehicle, thus in particular in the case of a non-stationary vehicle, ω the offset of the output signal of the yaw rate sensor 21 b repre sentierende _off1 sizes available.

The standstill is determined when the following three conditions a), b) and c) are simultaneously fulfilled:

• - The longitudinal speed of the vehicle is less than a first predetermined maximum value, which is referred to as limit speed v _G.
• - The yaw rate ω is smaller than a second predetermined one Maximum value.  
• - The acceleration of yaw, that is the time derivative the yaw rate, is less than a third predetermined Maximum value.

At a very low longitudinal vehicle speed ω, the very small wheel speeds n _i then present can no longer be detected by the wheel speed sensors and are incorrectly recognized as zero. As a result, in a device located in the vehicle for determining the longitudinal velocity of the vehicle, this is erroneously determined as disappearing ver. The limit speed v _G is the longitudinal speed of the vehicle which is just recognized as not vanishing. For each longitudinal speed lower than the limit speed v _G , the longitudinal speed zero is incorrectly determined.

Points b) and c) ensure that a Vehicle with a vanishing longitudinal speed stands on a rotating platter, not standing still is taken.

For the standstill adjustment, the output signal ω of the rotation rate sensor 21 b is read in at regular discrete sampling times t _i with constant time intervals Δt _a in block 23 a. At the same time, at these sampling times t _i ge it is checked whether the conditions a), b) and c) defining the standstill are fulfilled. In addition, a time interval T _n = n.Δt _{a is} determined, where n is an integer greater than 2. If the conditions a), b) and c) defining the standstill are met during a time interval T _n at all sampling times t _i falling in this interval, the standstill case is considered to be given. Such a time interval T _n , during which the conditions a), b) and c) defining the standstill are all fulfilled at the same time for each of the sampling times falling within this interval, is referred to below as the filling interval.

The process sequence for the detection of a compliance interval is shown in FIG. 4. In block 71 , the integer variable i = 0 is set. The output signal ω of the yaw rate sensor is then read in in block 72 . In block 73 it is checked whether all three standstill conditions a), b) and c) are fulfilled at the same time. If this is the case, then the number i is incremented by 1 in block 75 . If this is not the case, then i = 0 is set in block 74 and, at the same time, block 71 is started again at _a later point in time t _a . The variable i indicates the number of uninterrupted fulfillments of all three conditions a), b) and c) defining the standstill. In block 76 it is checked whether this number i of the uninterrupted fulfillments has reached the value n. If this is the case, the detection of a fulfillment interval is documented in block 77 by "fulfillment interval = true". At the same time, the variable "value" is assigned the most current value ω of the output signal of the yaw rate sensor. In the exemplary embodiment, this is a filtered value. Has in block 76 the number i to the value n is not yet reached, then (block 78) is added to ei nem to the time interval t _a later time he neut in block 72 the output of the yaw rate sensor is read ω.

As soon as a fulfillment interval has been detected, it is advantageous to shorten the length of the following time intervals. The time interval T _{n is} now replaced by the shorter time interval T _m with the length of the time interval T _m = m.Δt _a where m is an integer less than n. Only when the case of non-standstill is detected again does one go back to the longer time interval T _n = n.Δt _a and maintain this length of the time interval again until a fulfillment interval is detected again.

For the determination of the offset value ω _off1 in block 23 a, the last sampled signal ω of the penultimate filling _interval is used in each case. An exception here is only the first compliance interval, which is also characterized by the larger length T _a = n.Δt _a . Since no directly previous performance interval is present, here for the determination of the offset value ω _off1 in a block 23 the last sampled signal ω in this fulfillment Inter vall used. Instead of the last sampled signal, a filtered or averaged or both filtered and averaged signal can also be used.

In order to explain the mode of operation of the standstill adjustment in block 23 a clearly, the following straight-line vehicle movement with disappearing yaw rate and decreasing yaw acceleration is considered:
Phase 1: The vehicle travels at a constant longitudinal speed
Phase 2: The vehicle begins to brake. The longitudinal vehicle speed is even greater than the limit speed v _G defined in a).
Phase 3: The braking process ends. The longitudinal vehicle speed is already lower than the limit speed v _G.
Phase 4: The vehicle is exactly at a standstill.
Phase 5: The vehicle begins to accelerate. The vehicle's longitudinal speed is even lower than the limit speed V _G.
Phase 6: The acceleration process has progressed. The longitudinal vehicle speed is greater than the limit speed v _G.

In phase 1 and phase 2, condition a) is not met. The vehicle is not at a standstill.

In phase 3, all three conditions are met, so the case of a standstill is assumed, although the vehicle was still coasting. In order to avoid further processing of the first signals ω determined there as signals representing vehicle standstill, the time interval T _{n was} introduced.

The length of the time interval T _n is determined by the necessary consideration of the coasting during braking processes below the limit speed v _G defined in a) and the necessary consideration of filter run times of the signal ω in block 23 a. Ideally, the length of the time interval T _{n is} longer than an assumed realistic duration of phase 3 and the filter runtimes. In phase 4, all 3 conditions a), b) and c) are met at all sampling times t _i .

Therefore, now, in each case over shorter time intervals T _m, the output of the yaw rate sensor 21 b in a block 23 is read therefrom and the yaw rate ω _off1 offset from the determined before last maintenance interval.

In phase 5, all 3 conditions a), b) and c) are met at all sampling times, even though the vehicle is no longer at a standstill. However, the yaw rate _offset ω _off1 is determined from the penultimate maintenance interval, which ideally still falls into phase 4, in which there was still a real stationary vehicle.

In phase 6, the 3 conditions a), b) and c) are no longer fulfilled at the same time, which is why the vehicle is assumed to be no longer at a standstill and no further offset values ω _{off1 are} determined. Since the vehicle is now in motion, no further offset values ω _{off1 can} be determined until the vehicle _comes to a standstill again, characterized by the fulfillment of all three conditions a), b) and c).

The standstill adjustment method also works when the vehicle is not detected as standing still. In the case of the vehicle not detected as standing still, the last valid offset value ω _off1 determined is _read from a memory and used further.

However, the error band with half the width Δω _{off1 is} now widened in time. This clearly means that the wider the error band, the more time has passed since the last valid determination of the offset value ω _off1 .

In contrast to the standstill adjustment method, the steering angle method only works when the vehicle is not stationary. The relationship is the basis for calculating the yaw rate offset in the steering angle method

which is known from the specialist literature (see, for example, "Bosch - Automotive Pocket Book", 23rd edition, page 707). In addition to the longitudinal vehicle speed v and the steering wheel angle δ, the steering ratio i _L , a characteristic vehicle speed v _ch and the wheelbase l also go into this relationship. The steering ratio is the mechanical ratio between the steering wheel angle δ and the steering angle of the front wheels.

The offset ω _off2 is the difference between the yaw rate w _LWS calculated from the above relationship (1) and the yaw rate ω measured by the yaw rate sensor.

In addition to the calculation of an offset value ω _off2 , relation (1) also allows the width of an error _band to be calculated using the error propagation law.

The histogram method illustrated in FIG. 5 is based on a statistical evaluation of the ω b by the rotational speed sensor 21 delivered yaw rate signals. Only discreet yaw rate values are used. , ., ω _k-1 , ω _k , ω _{k + 1} ,. , , considered with constant discrete distances Δω. So it is ω _k-1 = ω _κ - Δω and ω _{k + 1} = ω _k + Δω. An interval of the width Δω, in the center of which the discrete yaw rate value lies, is assigned to each of these discrete yaw rate values. The yaw rate value ω _k therefore includes the interval with the lower limit ω _k - Δω / 2 and the upper limit ω _k + Δω / 2. These intervals are referred to below as classes.

All the yaw rate values measured by the yaw rate sensor 21 b are now assigned to the associated class. This gives the histogram shown in FIG. 5, which indicates the classes on the x-axis and the number of yaw rate values falling in each class on the y-axis. The yaw rate _{offset value} ω _off3 can be determined from this histogram in different ways:

• a) the average of the class with the highest count we passed on as yaw rate offset or
• b) The yaw rate offset is formed by forming a weighted average from the class with the highest count and the n neighboring classes on the left and on the right. This happens through the relationship Here, k is the index of the class with the highest count, ω _{i is} the yaw rate mean of the class identified by index i and z _{i is} the count of the class identified by index i.
• c) The yaw rate offset is formed by forming a weighted average over all classes. This can be done by taking the counters into account linearly
  or by taking the counters into account with a square
  respectively. The quadratic consideration of the counter readings is particularly suitable for suppressing occurring secondary maxima of the yaw rate in the histogram.

The error band can be, for example, the narrower be true, the more pronounced the maximum of the histogram is.

With the histogram method it is recommended for the viewer only yaw rate values up to a predetermined value ximal amount to consider. Furthermore, from the Histogram method only then determines offset values if the longitudinal vehicle speed is a preference size exceeds. Shouldn't these conditions then the histogram method is used used with the last valid offset value, where the assigned error band with increasing temporal Distance widens. Since the histogram method is a statisti is a sufficiently large number of measured values.

Therefore, when the vehicle starts, the histo values of the last ignition cycle as initialization read values.

The individual classes (also referred to as sub-areas or intervals) in the histogram can be filled, for example, by simply adding up the number of yaw rate values to be assigned to the individual classes. In the exemplary embodiment, however, the individual classes are not filled in the histogram by simply adding up the number of yaw rate values to be assigned to the individual classes, but by low-pass filtering with exponential characteristics. The values z _i thereby move always between 0 and the filter input value z _FE, which may be in play as z = _FE. 1 These variables z _i are also referred to as number variables (since they represent the number of values of the output signal assigned to the partial areas). z _i = 0 means an empty class (= partial area). If z _i assumes the filter input value z _FE , e.g. z _FE = 1, this means an infinite number of entries in this class. The counter reading z _i = z _FE in one class, i.e. z. B. zFE = 1, can never be achieved, it can le diglich values arbitrarily close to the filter input value, so z. B. 1 can be achieved. This prevents the classes from overflowing if there are too many entries.

A separate low pass is assigned to each class. The discrete equation (= difference equation) describing the i-th low pass is

z _i (tj) = (1 - dt / tauHist) .z ₁ (tj-1) + (dt / tauHist) .k.

Here, z _i (tj) is the counter reading of the i-th class at the sampling time tj. The sampling instant immediately preceding the sampling instant tj is tj-1. Thus z _i (tj-1) is the count of the i-th class at the sampling time tj-1. dt is the time interval between the sampling times and tauHist is the filter time constant.

k represents the input signal of the filter.

The factor k takes the value zero when the measured Yaw rate value does not fall into the corresponding class i and takes the value one when the measured yaw rate value in the corresponding class i falls. That means the Low pass is driven with an input signal which can take two values. These two possible input signals can also be used as a "first input signal" (e.g. k = 1) and "second input signal" (e.g. k = 0).

The filter time constant results as
max (10000 / vxvRef.FakTau, 200 s).

The mathematical symbol max denotes the maximum value of the two in the following brackets Expressions. vxvRef is the current one vehicle speed (measured in m / s), the size FakTau is explained in more detail below. 200 s means 200 Seconds.

The size FakTau takes the value 5.zimax when a movement supply of the vehicle. When the vehicle is at a standstill Size FakTau the value 150 s (= 150 seconds), zimax is the entry value of the class with the highest counter was standing. As a result, FakTau has a small value takes when there are only a few entries in the histogram. This means that tauHist also takes on a small value, i.e. H. the Filter time constant is small. However, there are a lot of hi make program entries, then FakTau and thus TauHist a great value. This will make the filter time constant bigger, the filter becomes less dynamic. That can be done also express that the filter time from the maximum Histogram height is dependent, d. H. the more pronounced it is Forms histogram, the less it can change.

If the measured yaw rate value does not fall into the i-th class, ie k = 0, then the disappearance of the last summand in the difference equation describing the filter leads to the value in the i-th class becoming somewhat smaller (reduction in the number size) , That is, the filter also has the useful property that the entries in the classes slowly fade away. This prevents all classes from filling up over time and the values z _i for all classes being very close to one. If k = 1, the number size increases.

It has proven to be suitable, not individually Enter yaw rate values in the histogram, but always Enter mean values from 10 yaw rate values. This will The arithmetic values obtained from 10 yaw rate values Determined mean and this arithmetic mean is in the histogram (above that of the corresponding class assigned low pass). Is the controller cycle at 100 milliseconds (i.e. every 100 milliseconds a Yaw rate sensor offset value determined), then takes place de second a histogram entry.

It should be emphasized here that all of the previous Ab cut the numerical values mentioned only as example values (which, however, turned out to be suitable in the implementation have to be seen). The generality of the invention should not be restricted in any way.

In the weighted mean value formations for the variable ω _off3 described on the previous pages and represented by mathematical formulas, the variable z _i appears again and again. When discussing these formulas, z _{i had} the descriptive meaning of counter readings, ie z _i is equal to the number of entries in the corresponding sub-area (= class, interval). The introduction of the low-pass filter gave z _i the meaning of a number size, ie z _i is a measure of the number of entries, but no longer necessarily the number of entries themselves. When implementing the specified calculation formulas for size ω _off3 , z _i of course, the number quantities supplied as the output signal of the low-pass filters can also be used. Even then, it is ensured that intervals with a higher number of entries (example: 100 entries, z _i e.g. 0.95) are weighted more than intervals with a lower number of entries (example: 17 entries, z _i z z B. 0.43).

Fig. 6 shows qualitatively an xy diagram obtained with the regression method. The vehicle longitudinal velocity v is plotted on the x-axis and a variable wdiff is plotted on the y-axis. This quantity wdiff results from ω _diff = ω -ω _comp . Here ω is the yaw rate measured by the rotation rate sensor and wcomp is a yaw rate determined from the wheel speeds, given by

Here v ₁ and v _{r are} the actual wheel speeds on the left and right wheels of the non-driven axle, r is the nominal value of the wheel radius, r + Δr ₁ is the actual radius of the left wheel of the non-driven axle, r + Δr _r is the actual radius of the right wheel of the non-driven axle, b is the track width.

This determination of ω _comp is based on the fact that different wheel speeds of the inner and outer wheels are present during cornering.

Analytical considerations show that there is an approximately linear relationship ω _diff ω _off + mv between ω _diff and v. Here ω _off denotes the y-axis penetration point and m the slope, which also depends on the wheel radii and the track width. The formula approximately applies to the slope m

At discrete times, a new point is entered into the xy diagram by measuring v and determining ω _diff . A regression line is now laid through these points. The y-axis intersection point and the slope m are determined from this regression line. The y-axis penetration point represents the yaw rate _offset ω _off4 and the slope m can be used to determine differences in the wheel radii of the non-driven wheels using equation (2). These differences can be due to tire tolerances. The determination of these differences can lead to the saving of a tire tolerance comparison.

If the slope m has the value zero, then they have not started driven wheels the same wheel radii.

Since the calculation of ω _comp only applies to slip-free wheels, the wheels of the non-driven axle should be considered in the regression method.

The following three conditions are required for the regression procedure to be valid:

• - The longitudinal vehicle speed is greater than one before given limit speed
• - The yaw rate ω is less than a predetermined limit yaw rate
• - The vehicle lateral acceleration a _y is less than a predetermined limit lateral acceleration

Point 1) ensures that the regression ver do not drive when the vehicle is stationary.

Points 2) and 3) ensure that the vehicle is not in a dynamic range moved.  

The regression method requires a speed va riation. If this speed variation is too small, i. H. the points in the xy diagram lie in the x-axis direction then the result will be inaccurate.

The error band takes into account, for example, the variance and the mean value of the values of the longitudinal vehicle speed.

From the determined offset values ω _off1 . , ., ω _off4 and the associated error bands with half the widths ω _off1,. , ., Δω _off4 , a corrected offset value ω _{offcorr is} determined in block 24 .

The corrected offset value can be determined in various ways. For example, as shown in FIG. 7, the corrected offset value can be determined as the average of the minimum of the maximum values and the maximum of the minimum values. For this purpose, from which, by the i-th ren procedural determined offset value ω _offi and the associated Def lerband with the half-width Δω _offi a minimum value ω _offi - Δω _offi calculated and a maximum value Δω _offi + Δω _offi. Now the minimum and maximum values of all 4 methods are plotted along the ω axis. The largest minimum value (identified in FIG. 7, max (ω _offi - Δω _offi )) is determined from the 4 plotted minimum values. Likewise, the smallest maximum value (identified in FIG. 7 min (ω _offi + Δω _offi )) is determined from the 4 maximum values plotted. In the event that the largest minimum value is smaller than the smallest maximum value, there is a common intersection of all 4 methods with the largest minimum value as the lower limit and the smallest maximum value as the upper limit.

By averaging the minimum of all the maximum values and the maximum of all the minimum values, the center of the intersection is determined and used as the corrected offset value ω _{ο ffkorr} . This method can also be used in the event that no common intersection exists, ie that the largest minimum value is greater than the smallest maximum value.

Alternatively, the corrected offset value can also be obtained by weighted averaging from all four determined offset values. It is advantageous here that the weight of the offset value whose error range is smallest is the largest in the weighted averaging. This can be done, for example, using the following weighted averaging:

The determination of the corrected offset value of the rotation rate sensors should be in regular or non-regular repeated at intervals. It turns out in in practice time intervals of the order of 0.1 seconds to 1 Second as beneficial.

Claims (12)

1. Method for determining the offset value (ω _off3 ) of the output signal of a vehicle _sensor ( 21 b),
in which the values of the output signal of the vehicle sensor ( 21 b) are determined,
in which the value range of the output signal of the vehicle sensor ( 21 b) is divided into at least two sub-ranges (intervals, classes),
in which the values of the output signal ( 21 b) are assigned to the partial areas,
in which the number of values (z _i ) representing the number of values of the output signal ( 21 b) assigned to the partial areas are determined,
characterized in that
the number sizes (z _i ) are determined by at least one low pass. 2. The method according to claim 1, characterized in that
a separate low pass is assigned to the at least two sub-areas (intervals, classes),
wherein each of the low-pass filters determines the number size (z _i ) assigned to its sub-area. 3. The method according to claim 1, characterized in that it is a low pass with exponential characteristics delt. 4. The method according to claim 2, characterized in that
that low-pass filter, in the associated sub-range of which the value of the output signal (ω) determined at a sampling time falls, is driven with a first input signal (k = 1) and
those low-pass filters in whose assigned value ranges the value of the output signal determined at the same sampling time does not fall are controlled with a second input signal (k = 0),
the second input signal being different from the first input signal. 5. The method according to claim 5, characterized in that
as a result of the activation of the low pass with a first input signal (k = 1), an increase in the number variable (z _i ) determined by this low pass takes place and
as a result of the activation of the low-pass filters with a second input signal (k = 0), the number variables (z _i ) determined by these low-pass filters are reduced. 6. The method according to claim 1, characterized in
that a first mean value is determined in a first manner from at least two successively determined values of the output signal (ω) of the vehicle sensor and
that instead of the values of the output signal of the vehicle sensor, the mean values of the first type are assigned to the partial areas. 7. The method according to claim 1, characterized in that the vehicle sensor detects the yaw movement of the vehicle. 8. The method according to claim 7, characterized in that the determined offset value (ω _off3 ) for automatic position control and / or control (ACC) is used in the motor vehicle. 9. The method according to claim 1, characterized
that the offset value (ω _off3 ) is determined by forming a weighted average and
that the subarea (interval, class) with the greatest number of assigned values of the output signal (counter reading z _i ) is taken into account in the formation of the weighted mean value, and at least one adjacent interval (class) is taken into account. 10. The method according to claim 9, characterized in that the linear or quadratic values of the number variables (z _i ) go into the weighting factors of the weighted mean. 11. Device for determining the offset value (ω _off3 ) of the output signal of a vehicle _sensor ( 21 b), wherein the sensor ( 21 b) detects at least one movement of a vehicle and means ( 23 c) are provided in which
the values of the output signal of the vehicle sensor (21 b) is determined,
the value range of the output signal of the vehicle sensor ( 21 b) is divided into at least two sub-ranges (intervals, classes),
the values of the output signal ( 21 b) are assigned to the sub-areas and
Number variables (z _i ) are determined which represent the number of values of the output signal ( 21 b) assigned to the partial regions,
characterized in that
the number sizes (z _i ) are determined by at least one low pass. 12. The apparatus according to claim 11, characterized in
that
a separate low pass is assigned to each of the at least two sub-areas (intervals, classes),
wherein each of the low-pass filters determines the number size (z _i ) assigned to its sub-area.