Method for processing sensor signal

Abstract

The invention relates to a method for processing sensor signals, in which two respective sensor signals of the sensor signals are first compared with each other at a first time point and a second time point in order to determine a difference between the two sensor signals (40, 42, 44) at the first time point and a difference between the two sensor signals (40, 42, 44) at the second time point, a relative deviation between the two differences is determined in each case, which relative deviations are compared to each other, and an evaluation of the sensor signals (40, 42, 44) is carried out on the basis of the comparison.

Description

TECHNICAL FIELD

[0001] The present application relates to a method for processing a sensor signal and an assembly performing the method. BACKGROUND

[0002] In vehicles, in particular in motor vehicles, a series of different sensors are used to detect different quantities and to convert them into electrical signals which carry information about these quantities, which can in turn be evaluated and further processed. So-called inertial sensors are used, for example, for measuring translational and rotational accelerations. By means of a combination of a plurality of inertial sensors in an inertial sensor unit, accelerations and rotations about three axes can be measured during driving. Such an inertial sensor unit is also referred to as an inertial measurement unit (IMU).

[0003] High-performance inertial measurement units, i.e. inertial measurement units (IMU) with high performance, measure the physical or body motion of a vehicle in terms of acceleration and angular velocity or angular rate. These units are widespread in the field of applications of automatic or autonomous driving, i.e. in AD applications (AD: Autonomous Driving), in which high signal safety is required and high demands are placed on accuracy.

[0004] Many applications, such as autonomous driving (AD), require the use of reliable sensor signals. Therefore, redundant IMU sensors measuring the same physical event, for example three angular rate sensors applied on the same circuit board, are often used to detect sensor faults or safety-relevant deviations by comparing the closeness between the redundant signals. Effective redundant signals are combined, which is also referred to as fusion, or a selection mechanism can be implemented in order to select the "best" final signal from all possible signals.

[0005] An offset drift is an error image or error pattern which can be seen as an offset in the sensor signal, which increases or decreases over time. Usually, this sensor error pattern is detected by a monitoring algorithm after a long period of time when the accumulated offset is large enough and significant.

[0006] It can be very useful to detect an offset drift before it becomes safety-relevant in order to respond faster and to make more appropriate decisions when redundant signals are fused. If, for example, a signal is affected by an offset drift, this signal can be assigned a smaller weight in a weighted decision algorithm in order to select the final signal.

[0007] The document DE 10 2022 212 433 A1 describes a method for monitoring sensor values, in which sensor quantities of a first and a second sensor are detected and compared with each other. Depending on the comparison, an additional sensor quantity of a third sensor is detected and also taken into account for the comparison.

[0008] From the document DE 10 2022 209 917 A1 a method for operating a vehicle is known, in which redundant sensor signals are read. Subsequently, a reference signal is calculated using the sensor signals. Then, a difference of each sensor signal and the reference signal is calculated in order to produce a difference code which is subsequently evaluated. SUMMARY

[0009] Against this background, the method according to the invention and the assembly according to the invention are proposed. Embodiments can be derived from the description.

[0010] The proposed method serves to process redundant sensor signals, for example sensor signals of an IMU, typically in a vehicle, in particular in a motor vehicle. In the method, respective two of the sensor signals are first compared with each other at a first point in time and at a second point in time, whereby a difference between the two sensor signals at the first point in time and a difference between the two sensor signals at the second point in time are calculated respectively. Here, each sensor signal is typically compared with any other sensor signal at the first point in time and at the second point in time. This is subsequently continued. Thus, three sensors or sensor signals produce three sensor pairs or three sensor signal pairs.

[0011] Subsequently, a relative deviation between the two differences is calculated respectively, that is to say, a relative deviation is calculated for each sensor pair. The relative deviations are then compared with each other. Then, based on the comparison or in consideration of the comparison, an evaluation of the sensor signals can be carried out. It can be determined, for example, whether one of the sensor signals shows an offset drift. If such an offset drift is identified, then the sensor signal can be directly identified. It can also be provided, however, that such an identification first leads to a setting of a qualification counter, that is to say, a decrement or an increment. If a threshold value is reached, then an identification of the sensor signal takes place.

[0012] It is thus proposed to implement a mechanism, in particular in order to detect an offset drift in the sensor signals, for example in the IMU signals. As described above, this information can be used to make a decision on the final signal used, which is calculated on the basis of all redundant signals.

[0013] The offset drift information can also be used to influence the activated / deactivated time period of the sensors in the concept, wherein these sensors can be in a low energy consumption mode if they are not used. If, for example, an offset drift is detected in a sensor, then the other redundant sensors can be kept in the activated state, that is to say, not passed into the mode with low energy consumption, since a possible safety-relevant deviation can occur in the next cycle and a plurality of sensors is required to make a better decision. Thus, the evaluation of the sensor signals influences the operation of the sensors.

[0014] The proposed component is used to carry out the described method, and the component or an evaluation unit arranged in the component is designed to carry out the method. The component can be implemented in hardware and / or software. Furthermore, the component can be integrated in a control device of the vehicle or be configured as such.

[0015] Further advantages and design solutions of the application result from the description and the drawings

[0016] It is to be understood that the features mentioned above and those to be explained below can be used not only in the respective combinations indicated, but also in other combinations or in isolation, without departing from the scope of the present application. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A functional block diagram illustrates an embodiment for carrying out the proposed method.

[0018] Figure 2 A graph illustrates the course of the value of the eligibility counter and the error detection time.

[0019] Figure 3 A graph illustrates the course of a sensor signal for illustrating an offset drift.

[0020] Figure 4 A purely schematic, very simplified diagram illustrates a vehicle with a component for carrying out the proposed method.

[0021] Figure 5 A flow chart illustrates a possible flow of the proposed method. DETAILED DESCRIPTION

[0022] The application is illustrated schematically in the drawings by embodiments and described in detail below with reference to the drawings.

[0023] Figure 1 A functional block diagram illustrates an embodiment for carrying out the proposed method. The diagram illustrates a first low-pass filter 10, a second low-pass filter 12, a third low-pass filter 14, a first block 20 for forming an average value, a second block 22 for forming an average value, and a third block 24 for forming an average value. Furthermore, the diagram also illustrates a fourth block 26 for calculating a difference, a fifth block 28 for detecting a relative deviation, a sixth block 30 for updating an eligibility counter, and a seventh block 32 for evaluation and finding a decision.

[0024] The input is the sensor signals S1 40, S2 42, and S3 44. The blocks 20, 22, 24 output the average values Avg1 50, Avg2 52, and Avg3 54. The fourth block 26 outputs the difference of the respective two average values, i.e. Dif1_2 60 (i.e. this corresponds to Avg1 - Avg2), Diff 1_3 62 (i.e. this corresponds to Avg1 - Avg3) and Diff 2_3 64 (i.e. this corresponds to Avg2 - Avg3).

[0025] The low-pass filters 10, 12, 14 and the three blocks 20, 22, 24 implement a pre-processing of the sensor signals S1 40, S2 42 and S3 44.

[0026] Thus, Figure 1 The diagrammatic description of Fig. 1 illustrates important parts or components of the proposed mechanism implemented by the described method. It is assumed in this example that there are three redundant sensor signals, i.e. S1 40, S2 42 and S3 44.

[0027] First, the signals 40, 42, 44 are filtered (low-pass filters 10, 12, 14) and then for this purpose an average value is periodically formed during a certain time period, e.g. 1 s, respectively (blocks 20, 22, 24). In this way it is possible to remove noise and high-frequency parts in the sensor signals 40, 42, 44, respectively, which are irrelevant in order to detect slow offset drifts. Subsequently, the differences between the redundant average signals (Diff 1_2 , Diff 1_3 and Diff 2_3 ) are calculated. In a fifth block 28, the relative differences between the signals are evaluated and possible deviations between the signals are detected.

[0028] It is generally checked here whether the actual absolute difference is significantly different from a previous deviation value which was calculated at a previous point in time (moment) and which is stored in an internal memory, in particular. In this way it is possible to detect a possible drift of one signal with respect to the other signals. If, for example, it is checked whether the Diff Ta 1_2 stored at a point in time Ta (Diff 1_2 has changed with respect to the Diff Tb 1_2 calculated at a point in time Tb (Diff 1_2

[0029] The following table describes the logic implemented here and the expected response based on these checks:

[0030]

[0031]

[0032] ​In this context, granting eligibility to a counter means incrementing the counter. Disqualifying a counter means decrementing the counter.

[0033] In the mathematical formulas in the table, Ta refers to the last point in time when the difference is stored in internal memory. Tb refers to the current time. The expression Delta is a parameter that defines the minimum detectable drift.

[0034] The response list shows how a counter for a signal is qualified when a possible drift is detected in the signal. For example, the counter should be qualified in the following way:

[0035] Updated counter value = previous counter value + increased step size

[0036] Conversely, if the counter is disqualified, then subsequent calculations can be performed:

[0037] Updated counter value = previous counter value - decreased step size.

[0038] The increasing and decreasing step sizes can be calibrated and can be completely different. In the example described here, the increasing step size is set to be greater than the decreasing step size.

[0039] It should be considered that this scheme allows only the detection of signals with errors (where drift exists), and assumes that the other two signals are error-free.

[0040] Therefore, three combinations are defined in the table. However, it is possible to expand the table and add some other combinations to enable better error handling. If, for example, drift is detected in all three checks, then an appropriate response, such as an error message, can be defined. This application scenario may occur when, for example, S1 has positive drift and S2 has negative drift.

[0041] If no potential drift is detected in the signal, the table shows the dequalification of the counters for that signal. If, for example, a potential drift in S1 is detected, then the other counters associated with S2 and S3 are dequalified.

[0042] If any one of the counters is granted eligibility, then the difference currently calculated in Tb, for example, Diff, is... Tb 1_2 Diff Tb 1_3 and Diff Tb 2_3 The information is stored in memory, where, in Ta, for example, Diff T1 1_2 Diff T1 1_3 and DiffTa 2_3 The previously stored values ​​are overwritten so that they can be used in the next iteration, where all checks are repeated.

[0043] If the counter is only disqualified because no possible drift of the signal was detected, then the existing difference, such as the diff, stored in memory will be lost. Ta 1_2 Diff Ta 1_3 and Diff Ta 2_3 It was not covered and can be used during the next evaluation check.

[0044] When updated once per second, the counter is updated at a specific frequency, such as 1 / s, by performing the above checks, i.e., granting / revoking eligibility.

[0045] If a counter reaches a threshold for a signal, then that signal can be identified as invalid, for example, because a significant offset drift or other response can be defined. Figure 2 This illustrates an exemplary operation of the counter.

[0046] The diagram shows graph 100, with time plotted on its horizontal axis 102 and the value of the qualification counter plotted on its vertical axis 104. In graph 100, curve 110 illustrates the change in the value of the qualification counter. The first curly brace 112 indicates an increasing or expanding step size, and the second curly brace 114 indicates a decreasing step size. Therefore, these step sizes differ in their absolute values.

[0047] Furthermore, the diagram illustrates a threshold 120 for the qualification counter value. If curve 110 exceeds this threshold, the signal involved is flagged as erroneous. An error detection time 130 is generated using a selected value for threshold 120.

[0048] It should be considered whether the detected drift is positive or negative. To do this, the sign of the difference should be taken into account, and additional logic should be calculated. The table above shows the absolute values. If the sign of the difference is considered, then the direction of the offset (positive or negative) can be estimated.

[0049] If, for example, a drift is detected in S3, then subsequent additional checks can be performed to determine whether a positive or negative drift exists:

[0050] If [(Diff Tb 1_3 <Diff Ta 1_3 ) and (Diff Tb 2_3<Diff Ta 2_3 If )] is true, then we obtain a positive drift from S3 to S1 and S2. If not, then

[0051] If [(Diff Tb 1_3 >Diff Ta 1_3 ) and (Diff Tb 2_3 >Diff Ta 2_3 If )] is true, then we get the negative drift from S3 to S1 and S2.

[0052] The following illustrations show examples to illustrate the proposed method in detail. Here, S3 shows the positive offset drift.

[0053] Figure 3 An example is shown to illustrate the proposed method in detail. Here, S3 shows the positive offset drift. The illustration, plotted as graph 150 (with time plotted on its horizontal axis 152 and signal values ​​plotted on its vertical axis 154), shows the first curve 160 (reflecting the change in the value of signal S1), the second curve 162 (reflecting the change in the value of signal S2), and the third curve 164 (reflecting the change in the value of signal S3).

[0054] The time points Ta 170 and Tb 172 are plotted in the chart, and braces are used to indicate:

[0055] Diff 1_3 =-4 (see attached diagram, label 180),

[0056] Diff 2_3 =+2 (see attached figure 182),

[0057] Diff 1_3 =-8 (see attached figure 184),

[0058] Diff 2_3 =-2 (see attached figure 180).

[0059] It is important to note that Diff Tb 1_3 = -8 and therefore less than Diff Ta 1_3 =-4. Additionally, Diff also applies. Tb 2_3 = -2 and therefore less than Diff Ta 2_3 =+2.

[0060] Figure 4 A vehicle 200 is shown in a very simplified form, having a component 202 for performing the methods presented herein. Three sensors 204 are also shown, each providing a sensor signal 206. Redundant sensor signals 206 are involved here. An evaluation unit 210 is provided within component 202, designed to perform an implementation of the methods presented herein. A qualification counter 212 is assigned to the evaluation unit 210.

[0061] Figure 5 A flowchart illustrates a possible flow of the described method. In the first step 250, the method begins and redundant sensor signals are detected. In the next step 252, these sensor signals are preprocessed, i.e., filtered using a low-pass filter, and an average value is formed.

[0062] The preprocessed sensor signals are then processed. In step 254, corresponding two sensor signals are compared with each other. Thus, three sensor signal pairs are generated from the three sensor signals. This comparison is performed at a first time point and then at a second time point, where this process continues. In step 256, the relative deviation in the difference between the sensor signal pairs is detected. The relative deviation is then evaluated, and one or more qualification counters are updated if necessary in step 258. Finally, an evaluation is performed in step 260. If, for example, it is determined in the evaluation that there is an offset drift in one of the sensor signals, then that signal is considered with only a small weight or even not considered at all when obtaining the final signal.

Claims

1. A method for processing redundant sensor signals (40, 42, 44, 206), wherein, The corresponding two sensor signals (40, 42, 44, 206) are first compared with each other at a first time point and a second time point, so as to calculate the difference between the two sensor signals (40, 42, 44, 206) at the first time point and the difference between the two sensor signals (40, 42, 44, 206) at the second time point. The relative deviation between the two differences is calculated separately. The relative deviations are compared with each other, and The evaluation of the sensor signals (40, 42, 44, 206) is performed based on the comparison.

2. The method according to claim 1, wherein, The sensor signals (40, 42, 44, 206) are preprocessed before the actual processing.

3. The method according to claim 1, wherein, The preprocessing includes filtering using low-pass filters (10, 12, 14) and generating an average value.

4. The method according to any one of claims 1 to 3, wherein, The method is continued iteratively.

5. The method according to any one of claims 1 to 4, wherein, The final signal is obtained from the redundant sensor signals (40, 42, 44, 206).

6. The method according to claim 5, wherein, The evaluation of the sensor signals (40, 42, 44, 206) is taken into account when obtaining the final signal.

7. The method according to any one of claims 1 to 6, wherein, The evaluation of the sensor signals (40, 42, 44, 206) is taken into account when the sensor (204) is in operation.

8. The method according to any one of claims 1 to 7, wherein the method is used to detect offset drift in one of the sensor signals (40, 42, 44, 206).

9. The method according to any one of claims 1 to 8, wherein, The relative deviation obtained affects the value of the qualification counter (212), which is then assigned a threshold (120).

10. A component for processing redundant sensor signals (40, 42, 44, 206) using an evaluation unit (202), the component being designed to perform the method according to any one of claims 1 to 9.

Images