DE102017211816A1 - METHOD FOR CALCULATING A RADAR SPACING WIDTH AND A DEVICE FOR CARRYING OUT THE METHOD - Google Patents

Abstract

A method of calculating a radar distance sighting distance of a radar sensor (2), comprising the steps of: S1 creating a mathematical model and indicating a predicted value of a distance sighting distance from a previous cycle to a current cycle; S2 constructing a new histogram using the real distance of all objects (4, 5) from a field of view (3) of the radar sensor (2) and identifying a scenario; and S3 taking into account the mathematical model from the first method step (S1) and the identified scenario from the second method step (S2), and obtaining an estimated value of the distance sight distance of the radar sensor (2); and an apparatus for carrying out the method.

Description

The invention relates to a method for calculating a radar distance sight distance. The invention further relates to an apparatus for carrying out the method.

Radar devices are used in many applications, for example for the detection of objects in a detection area and for distance measurement. In motor vehicles, radars are increasingly used to determine distances to vehicles in front and fixed objects.

In this context, radar devices in motor vehicles form components of driver assistance systems (ADAS Advanced Driver Assistance Systems) for increasing the safety of the vehicle, driver and road users.

The presence of foreign matter or objects can affect the accuracy and reliability of radars, particularly radar sensors. Such foreign material may e.g. be caused by environmental conditions, such as temperature, humidity, ice, snow, rain or dirt / slush. Such an environmental condition can impair the intended operating behavior of the automatic radar sensors, in extreme cases even prevent them.

For an analysis of performance degradation of radar sensors, a so-called sensor visibility is used. This involves a statistical approximation of the sensor's ability to detect solid objects in the entire field of view.

• - Bewegliche Objekte
• - Stationäre Objekte
• - Erste/Letzte Erfassung von Objekten

To obtain a statistical range visibility, three histograms are used in which the following data is entered:

• - Moving objects
• - Stationary objects
• - First / last capture of objects

The average values of each histogram are processed by means of fusion and finally the sensor distance range is determined from this. When the distance sight distance of the radar sensor 5 is below a certain threshold, there is a notification of a performance decrease.

One of the problems of the conventional methods is that updating the histograms does not take into account the following fact. As objects leave the field of view of the radar sensor, and there is no attenuation of the radar sensor (i.e., the weather conditions remain good), the average of the above-mentioned histogram decreases while the radar sensor performance is not compromised. This reduced value does not represent a real decrease in performance and may lead to a false or erroneous blocking message.

An object of the invention is therefore to provide an improved method for calculating a radar distance sighting distance.

The object is achieved by an article having the features specified in claim 1.

Another object is to provide a device for such a method.

This further object is achieved by an article having the features specified in claim 9.

The invention thus provides a method of calculating a radar distance sighting distance of a radar sensor, comprising the steps of, namely, creating a mathematical model and indicating a predicted value of a distance sighting distance from a previous cycle to a current cycle; Building a new histogram using the real distance of all objects from a field of view of the radar sensor and identifying a scenario; and taking into account the mathematical model from the first method step and the identified scenario from the second method step, and obtaining an estimated value of the distance sighting distance of the radar sensor.

The method advantageously makes it possible to prevent messages about a power attenuation of a radar sensor in difficult environmental conditions. This enhances security by preserving and not disabling radar functionality.

A device according to the invention for carrying out the method according to the invention for calculating a radar distance sighting range comprises the radar sensor, a statistical block, a prediction value block, an identification block, a linking block with a filter and an estimation block, and an output block.

It is advantageous that an implementation of the device in an existing radar system is possible.

The possibilities for messages in case of false blockages of the radar sensor can be advantageously reduced.

In one embodiment, when constructing a histogram, the histogram is used to compare its mean gradient with a mean gradient of an existing extrapolated distance, and differentiates between a real attenuation and an object that leaves the field of view of the radar sensor. This can make an effective distinction.

In another embodiment, a dynamic Kalman filter is used. This is a proven technique and easy to use.

A further embodiment provides that the creation of the mathematical model the sub-steps T1.1. comprising: viewing last N values last distance values = {R ₀ , R ₁ , ..., R _N-1 } of provided distance sight distances; and T1.2 Nonlinear joining of the last N values Last distance values = {R ₀ , R ₁ , ..., R _N-1 } to a curve (R) R = R _Last distance values (t) and predictions of a distance sight distance. In this case, a Levenberg-Marquardt algorithm can be used for the non-linear connection, which results in an effective evaluation.

• T2.1 Erstellen einer Varianz des Systems oder des Mess-Rauschens und Erzeugen eines Histogramms, welches in Beziehung mit real erfassten Abständen der Objekte, die sich in dem Sichtbereich des Radarsensors befinden;
• T2.2 Ermitteln von Gradienten ∇MeanReal_k und ∇MeanExtrapol_k nach den folgenden Gleichungen: $$\nabla {\text{MeanReal}}_{\text{k}}=\frac{{\text{MeanReal}}_{\text{k}}-{\text{MeanReal}}_{\text{k-1}}}{\text{ECU Zykluszeit}}$$
  
  $$\nabla {\text{MeanExtrapol}}_{\text{k}}=\frac{{\text{MeanExtrapol}}_{\text{k}}-{\text{MeanExtrapol}}_{\text{k-1}}}{\text{ECU Zykluszeit}};$$
  
  und
• T2.3 Identifizieren eines zugehörigen Szenarios durch Analysieren der Gradienten ∇MeanReal_k und ∇MeanExtrapol_k.

In a further embodiment, the second method step, namely the construction of the histogram, comprises the following three sub-steps.

• T2.1 creating a variance of the system or measurement noise and generating a histogram related to real detected distances of the objects located within the field of view of the radar sensor;
• T2.2 Determining Gradients ∇MeanReal _k and ∇MeanExtrapol _k according to the following equations: $$\nabla {\text{Mean Real}}_{\text{k}}=\frac{{\text{Mean Real}}_{\text{k}}-{\text{Mean Real}}_{\text{k-1}}}{\text{ECU cycle time}}$$
  
  $$\nabla {\text{MeanExtrapol}}_{\text{k}}=\frac{{\text{MeanExtrapol}}_{\text{k}}-{\text{MeanExtrapol}}_{\text{k-1}}}{\text{ECU cycle time}};$$
  
  and
• T2.3 Identify an associated scenario by analyzing the gradients ∇MeanReal _k and ∇MeanExtrapol _k .

This advantageously enables a fast and effective identification of the scenario by comparing these gradients.

• T3.1 Vorhersagen der Abstandssichtweite basierend auf den letzten N Werten R_k= R_{LetzteAbstandsWerte} (t_k);
• T3.2 Messen eines Wertes MeasR_k, welcher von dem Mittelwert des Histogramms der extrapolierten Mittelwerte im Zyklus k bereitgestellt wird;
• T3.3 Anpassen einer Kalmanverstärkung K_k eines Kalman-Filters in Abhängigkeit von dem identifizierten Szenario; und
• T3.4 Einschätzen eines Zustands gemäß folgender Gleichung EstR_k = R_k + K_k (R_k - MeasR_k) und entsprechendes Ausgeben oder Unterlassen einer Meldung.

In a further embodiment, it is provided that the third method step, namely taking into account the mathematical model, has the following sub-steps.

• T3.1 predicting the distance visual range based on the last N values of R _k = R _{Last distance values} (t _k);
• T3.2 measuring a value MeasR _k which is provided by the mean value of the histogram of the extrapolated mean values in the cycle k;
• T3.3 adjusting a Kalman gain K _{k of} a Kalman filter as a function of the identified scenario; and
• T3.4 estimating a state according to the following equation estr _k = R _k + K _k (R _k - MeasR _k) and outputting corresponding or omission of a message.

In this way, an effective adaptation of the filter and an effective assessment of a condition can be achieved. In addition, the calculated distance sight distance can be specified with a higher accuracy.

A further embodiment provides that in sub-step T3.3, the Kalman gain K _{k is} adjusted in the case of the damping scenario such that more weight is placed on the measurement, and in the case of objects which enter into the field of view of the radar sensor or leave it, the Kalman gain K _k adjusted so that more emphasis is placed on the prediction. Thus, an advantageous weighting for a decision is possible.

In one embodiment of the device, the filter may comprise a Kalman filter.

The method and apparatus may be applied to any radar detection system, e.g. also with wide-range radar systems.

The invention will be further described in connection with the figures with reference to embodiments.

• 1 eine schematische Darstellung eines Fahrzeugs mit einem Radarsensor;
• 2 ein schematisches Flussdiagramm eines Ausführungsbeispiels eines erfindungsgemäßen Verfahrens;
• 3 eine schematische grafische Darstellung von Abstandswerten;
• 4-6 schematische Histogramme von Abständen; und
• 7 ein schematisches Blockschaltbild eines Ausführungsbeispiels einer erfindungsgemäßen Vorrichtung.

Hereby shows:

• 1 a schematic representation of a vehicle with a radar sensor;
• 2 a schematic flow diagram of an embodiment of a method according to the invention;
• 3 a schematic graphical representation of distance values;
• 4-6 schematic histograms of distances; and
• 7 a schematic block diagram of an embodiment of a device according to the invention.

In 1 is a schematic representation of a vehicle 1 with a radar sensor 2 shown.

The vehicle 1 drives on a road in a direction indicated by an arrow. The radar sensor 2 is located at the front of the vehicle 1 , There can also be several radar sensors 2 to be available.

The radar sensor 2 has a field of view 3 on, the forward in the direction of travel of the vehicle 1 is directed. Different objects 4 . 5 in this field of view 3 be from the radar sensor 2 recorded, with their distances are measured. The object 5 moves out of sight 3 out, which is indicated by another arrow.

These measured values are then further processed in a method according to the invention for calculating a radar distance sighting distance.

In addition shows 2 a schematic flow diagram of an embodiment of the method according to the invention.

In a first process step S1 A mathematical model is created to indicate a predicted value of a distance sighting distance from a previous cycle to a current cycle.

Then in a second process step S2 a new histogram using the real distance of all objects 4 . 5 within the field of view 3 of the radar sensor 2 built up. The histogram is used to compare its mean gradient with the average gradient of the extrapolated distance present, between a real attenuation and an object 5 which the field of vision 3 leaves to distinguish.

A third process step S3 includes a consideration of the mathematical model from the first step S1 and identifying the scenario from the second step S2 by applying a dynamic Kalman filter, which then obtains an estimated value of the distance sighting distance.

The following are the process steps S1 to S3 further explained.

To create the mathematical model in the first step S1 in a first step T1 .1 considers the last N values of the distance sight distance which are provided by the already existing histograms described above in the introduction to the description. The last N values are named as follows $$\text{LetzeAbstandsWerte}=\left\{{\text{<figure-callout id="0" label="R" filenames="DE102017211816A1\_0013.png,DE102017211816A1\_0014.png" state="{{state}}">R</figure-callout>}}_0.{\text{R}}_1.\mathrm{...,}{\text{R}}_{\text{N}-1}\right\}$$

These last distance values 6 are in a schematic graph in 3 applied over time t with small circles.

Then, in a further sub-step T1.2, a non-linear connection process, eg a Levenberg-Marquardt algorithm, is applied to the set of the last distance values 6 applied. This results in a connected curve R over the time t with the designation $$R=R_{\text{Last spacing values}}\left(\text{t}\right)$$

The curve R _{Last Distance Values} (t) is used to predict a distance _sight distance for the next step S2 to meet.

In the second process step S2 For generating the variance of the system or the measurement noise in a further sub-step T2.1, a new histogram is generated, which is related to real recorded distances of the objects 4 . 5 that are in the field of vision 3 of the radar sensor 2 are located. This is in 4 as a schematic histogram 7 showing patterns represented in real pitch behavior observed from the field data measurements.

5 and 6 represent the behavior over time t of the histogram 7 with the real distances. It shows 5 a histogram 8th for extrapolated distances without attenuation, and 6 shows a histogram for extrapolated distances with a weakening.

When real attenuation occurs, this results in the attenuation of the extrapolated distance of all objects in the range of visibility 3 of the radar sensor 2 , this shows 6 as a histogram 9 , In this case the real attenuation will be a shift of all extrapolated data of the histogram 9 in the 6 to the left (indicated by arrow) observed.

In the case where a situation of recovery from an attenuation scenario (by environmental influences, as described above) occurs and the reflectivity of all objects increases, a shift of all the data of the histogram 8th according to 5 be determined to the right.

$$\nabla {\text{MeanExtrapol}}_{\text{k}}=\frac{{\text{MeanExtrapol}}_{\text{k}}-{\text{MeanExtrapol}}_{\text{k-1}}}{\text{ECU Zykluszeit}};$$

The shift of the extrapolated data is reflected in the gradient (see the following equations (2) and (3)) of the mean value of all the data in the histogram. $$\nabla {\text{Mean Real}}_{\text{k}}=\frac{{\text{Mean Real}}_{\text{k}}-{\text{Mean Real}}_{\text{k-1}}}{\text{ECU cycle time}}$$

$$\nabla {\text{MeanExtrapol}}_{\text{k}}=\frac{{\text{MeanExtrapol}}_{\text{k}}-{\text{MeanExtrapol}}_{\text{k-1}}}{\text{ECU cycle time}};$$

In a further sub-step T2.2, the two gradients ∇MeanReal _k and ∇MeanExtrapol _{k are} determined or calculated.

In the case of a stepwise damping scenario or a recovery from damping, the value ∇MeanReal _{k is} in the neighborhood of zero and ∇MeanExtrapol _k is strictly correlated with the power decrease, ie there is a fast change of MeanExtrapol _k compared to MeanReal _k .

In the case in which objects 5 the field of vision 3 leave or enter into it, both gradients, ∇MeanReal _k and ∇MeanExtrapol _{k are} strictly correlated.

Thus, in a sub-step T2.3, the associated scenario can be identified by analyzing the gradients VMeanReal _k and ∇MeanExtrapol _k .

In order to obtain the estimated value for the distance view distance, in the application of the Kalman filter and the above data in the third process step S3 the following sub-steps T3.1 to T3.4 performed.

In a sub-step T3.1, a prediction of the distance sighting distance based on the last N values from equation (1) is made: $$R_k=R_{\text{Last spacing values}}\left(t_k\right)$$

In a further sub-step T3.2, the measurement MeasR _k , which is provided by the mean value of the histogram of the extrapolated mean values in the cycle k, is used.

In a further substep T3.3, the Kalman gain K _{k of} the Kalman filter is adjusted as a function of the identified scenario.

In the case of the damping scenario, the variance of the system increases, and therefore the Kalman gain K _k is adjusted to place more emphasis on the measurement.

In the case of objects 5 which is in the field of vision 3 enter or leave it, thereby lowering the variance of the system, the Kalman gain K _{k is} adjusted such that more emphasis is placed on the prediction.

In a substep T3.4, the estimated state is given $$est{R}_k=R_k+K_k\left(R_k-Meas{R}_k\right)$$

Based on this status, a corresponding message can be issued or omitted.

In this way, messages due to false blocks can be reduced, making the calculated distance range more accurate.

In 5 is a schematic block diagram of an embodiment of a device according to the invention 10 shown.

The device 10 here includes the radar sensor 2 , a statistics block 11 , a predictive value block 12 , an identification block 13 , a shortcut block 14 with a filter 15 and an estimation block 16 , and an output block 17 ,

The statistics block 11 generates the mathematical model of the first process step S1 , The prediction value block 12 is designed to predict a value of the distance sight distance. The identification block 13 builds the histograms to examine and compare the mean gradients and identifies a state or scenario by distinguishing between a real attenuation and an object 5 which the field of vision 3 leaves.

The link block 14 serves to link the mathematical model from the statistics block 11 and identifying the scenario from the identifier block 13 , In this case, the link block 14 the filter 15 and the estimation block 16 on. The filter 15 may have, for example, a Kalman filter.

The output block 17 performs an output of a message, eg if an identified blockage of the radar sensor 2 due to environmental conditions, such as ice, rain, dirt, etc.

The device 10 can be implemented in an existing radar sensor operating system.

The method and apparatus may be used with any radar detection system.

The embodiment described above does not limit the invention, but it is within the scope of the claims modifiable.

LIST OF REFERENCE NUMBERS

1

vehicle

2

radar sensor

3

field of view

4, 5

object

6

distance value

7, 8, 9

histogram

10

contraption

11

statistics block

12

Predictive value block

13

identification block

14

link block

15

filter

16

estimation block

17

output block

R

Curve

S

step

T

partial step

Claims (10)

A method for calculating a radar distance sighting distance of a radar sensor (2), comprising the method steps Constructing (S1) a mathematical model and specifying a predicted value of a distance sighting distance from a previous cycle to a current cycle; Constructing (S2) a new histogram using the real distance of all objects (4, 5) from a field of view (3) of the radar sensor (2) and identifying a scenario; and Taking into account (S3) the mathematical model from the first method step (S1) and the identified scenario from the second method step (S2), and obtaining an estimated value of the distance sight distance of the radar sensor (2). Method according to Claim 1 characterized in that in the second method step (S2) the histogram is used for a comparison of its mean gradient with an average gradient of an existing extrapolated distance, and distinguishing between a real attenuation and an object (5) representing the field of view (3) of the Radar sensor (2) leaves, takes place. Method according to Claim 1 or 2 , characterized in that in the third method step (S3), a dynamic Kalman filter is applied. Method according to Claim 3 characterized in that the first method step (S1) comprises the substeps of: T1.1 Viewing Last N Values Last Distance Values = {R ₀ , R ₁ , ..., R _N-1 } of provided distance sight distances; and T1.2 Nonlinear joining of the last N values Last distance values = {R ₀ , R ₁ , ..., R _N-1 } to a curve (R) R = R _{Last distance values} (t) and predictions of a distance _sight distance. Method according to Claim 4 , characterized in that in the sub-step T1.2 a Levenberg-Marquardt algorithm is used. Method according to Claim 4 or 5 characterized in that the second method step (S2) comprises the substeps of: T2.1 creating a variance of the system or the measurement noise and generating a histogram related to real detected distances of the objects (4, 5) that are located in the viewing area (3) of the radar sensor (2); T2.2 Determining gradients ∇MeanReal _k and VMean extrapol _k according to the following equations $$\nabla {\text{Mean Real}}_{\text{k}}=\frac{{\text{Mean Real}}_{\text{k}}-{\text{Mean Real}}_{\text{k-1}}}{\text{ECU cycle time}}$$

$$\nabla {\text{MeanExtrapol}}_{\text{k}}=\frac{{\text{MeanExtrapol}}_{\text{k}}-{\text{MeanExtrapol}}_{\text{k-1}}}{\text{ECU cycle time}};$$

and T2.3 identifying an associated scenario by analyzing the ∇MeanReal _k and VMean extrapol _k gradients. Method according to Claim 6 , characterized in that the third method step (S3) comprises the sub-steps: T3.1 predicting the distance sighting distance based on the last N values R _k = R _Last distance values (t _k ); T3.2 measuring a value MeasR _k which is provided by the mean value of the histogram of the extrapolated mean values in the cycle k; T3.3 adjusting a Kalman gain K _{k of} a Kalman filter as a function of the identified scenario; and T3.4 estimating a condition according to the following equation EstR _k = R _k + K _k (R _k - MeasR _k ) and corresponding output or omission of a message. Method according to Claim 7 , characterized in that in the substep T3.3 the Kalman gain K _{k is} adjusted in the case of the damping scenario such that more weight is placed on the measurement, and in the case of objects (5) which are projected into the field of vision (3 ) of the radar sensor (2) or leave the field of view (3), the Kalman gain K _k adjusted so that more weight is placed on the prediction. Device (10) for carrying out a method for calculating a radar distance sight distance according to one of the preceding claims, characterized in that the device (10) comprises the radar sensor (2), a statistics block (11), a prediction value block (12), an identification block (12). 13), a linking block (14) having a filter (15) and an estimation block (16), and an output block (17). Device (10) according to Claim 9 , characterized in that the filter (15) is a Kalman filter.

Images