Method for monitoring a sensor signal

Abstract

Method for monitoring at least one sensor signal in a vehicle, wherein, when a fault is detected in the at least one sensor signal, a bidirectional qualification counter is set, and when the qualification counter reaches one of at least two threshold values, the sensor signal is marked as faulty, taking into account the type of deviation when setting the bidirectional qualification counter.

Description

[0001] The invention relates to a method for monitoring a sensor signal and an arrangement for carrying out the method. State of the art

[0002] Vehicles, especially motor vehicles, use a variety of sensors to detect different quantities and convert them into electrical signals that carry information about these quantities. These signals can then be evaluated and further processed. For example, inertial sensors are used to measure translational and rotational accelerations. By combining several inertial sensors in an inertial sensor unit, accelerations and rotations around three axes can be measured while driving. Such an inertial sensor unit is also called an inertial measurement unit (IMU).

[0003] High-performance inertial measurement units (IMUs) measure the physical motion of a vehicle with respect to acceleration and angular velocity or rate. Such units are widely used in automated driving (AD) applications, where high signal reliability and accuracy are essential.

[0004] Many applications, such as autonomous driving (AD), require the use of a reliable sensor signal. Therefore, redundant IMU sensors measuring the same physical event—for example, three angle rate sensors mounted on the same circuit board—are typically used to detect sensor failures or safety-related deviations by comparing how close the redundant signals are to each other. The valid redundant signals are then combined, a process also known as fusion, or a selection mechanism can be implemented to choose the "best" final signal from all possible signals.

[0005] To detect a faulty signal, detection methods are employed. In one known method, the following applies: If a large deviation between redundant signals is detected, a qualification counter is incremented. If no deviation is present, the counter is disqualified, i.e., it is decremented if it was previously incremented or qualified. When the qualification counter reaches a threshold, the signal is considered faulty. The reason for implementing such a counter is to obtain more robust detection in the event that the detected deviation is only a sporadic occurrence, but the sensor continues to provide valid results. By changing the qualification threshold, the fault detection time can also be modified. In this context, it is referred to as... Fig. 1 referred. Disclosure of the invention

[0006] Against this background, a method with the features of claim 1 and an arrangement according to claim 9 are presented. Embodiments are described in the dependent claims and in the description.

[0007] A method for monitoring at least one sensor signal in a vehicle is presented, in which a bidirectional qualification counter is set when a fault is detected in the at least one sensor signal. A bidirectional qualification counter is a counter configured to be decremented or incremented depending on the type of deviation. Such a bidirectional qualification counter is also capable of accepting positive and negative values.

[0008] When the bidirectional qualification counter reaches one of at least two threshold values, the sensor signal in question is marked as faulty, taking into account the type of deviation when setting the bidirectional qualification counter.

[0009] For example, the sign of the deviation can be used to determine whether the qualification counter is incremented or decremented. This sign indicates, for instance, whether a positive or negative offset error is present. Furthermore, the sign can determine whether a gain or loss error is present.

[0010] It was thus recognized that qualification counters without a sign have been used so far, meaning that the counter starts at 0 and is incremented whenever a sensor deviation occurs. The counter is decremented back to 0 when the deviations disappear.

[0011] The present method is based on the understanding that some statistical error patterns, such as offset errors and sensitivity errors, have two states. This means that the offset error can be either positive or negative, but both offset errors cannot be present in the signal simultaneously. The same applies to the sensitivity error. It can be either an attenuation error or a gain error, but both sensitivity errors cannot be present in the signal at the same time. In known methods, this requires two qualification counters for the detection of different types of errors, e.g., one counter that responds to gain errors and another counter that responds to attenuation errors.

[0012] It is now proposed to use a bidirectional signed counter, i.e., a counter with a sign that can take both positive and negative values. If the qualification counter is used to detect offset deviations, the following applies: positive values ​​indicate a potential positive offset error, while negative values ​​indicate a negative offset error. The same approach can be used with a monitoring algorithm that detects sensitivity errors: positive counter values ​​indicate an amplification error in the sensor signal, while a negative qualification counter value indicates an attenuation error.

[0013] The advantages for offset and sensitivity monitoring algorithms that arise when a bidirectional qualification counter is used are shown below: The information about whether the offset deviation is positive or negative can only be obtained by checking the sign of the counter. The same can be done when the bidirectional counter is used to measure sensitivity errors. In this case, a positive counter indicates a gain error, and a negative counter indicates a loss error.

[0014] The overall robustness of the detection system increases if it is not possible to have both a positive offset error (amplification) and a negative offset error (decrease) in the signal simultaneously. For example, if the counter was registering positive values ​​because a positive offset deviation was detected during previous checks, and suddenly a negative offset is detected, this could indicate that a permanent positive offset does not actually exist and that it was only a temporary issue. Therefore, in this case, the counter can be decremented more quickly to prevent it from reaching the threshold in future cycles due to a false positive detection.

[0015] It should be taken into account that, especially when the qualification threshold is set at low values ​​due to the need for a short data collection time, the robustness of the data collection plays a very important role.

[0016] The presented arrangement serves to carry out the procedure described herein and includes an evaluation unit for this purpose. The arrangement and / or the evaluation unit is / are implemented in hardware and / or software. The arrangement can be integrated into a vehicle control unit, in particular a motor vehicle, or be designed as such a control unit.

[0017] Furthermore, the arrangement can include a bidirectional qualification counter. This counter, in turn, can be implemented as a purely software solution, a purely hardware solution, or a hardware / software system.

[0018] Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawings.

[0019] It is understood that the features mentioned above and those to be explained below can be used not only in the combinations specified, but also in other combinations or on their own, without leaving the scope of the present invention. Brief description of the drawings Fig. Figure 1 shows in a graph the determination of the error detection time using a threshold value. Fig. Figure 2 shows in a graph the progression of the value of a qualification counter and two threshold values. Fig. Figure 3 shows, in a highly simplified, purely schematic form, a vehicle with one embodiment of the presented arrangement. Fig. Figure 4 shows a flowchart illustrating an embodiment of the presented method. Embodiments of the invention

[0020] The invention is schematically illustrated with reference to embodiments in the drawings and is described in detail below with reference to the drawings.

[0021] Fig.Figure 1 shows a graph 10, where time is plotted on the abscissa 12 and the value of a qualification counter is plotted on the ordinate 14. A curve 20 illustrates the progression of the qualification counter's value, which is incremented with each error or triggering event (i.e., a signal value deemed incorrect), resulting in a step-like curve 20. Specifically, an error is detected when the monitored sensor signal deviates too significantly from an expected pattern. If multiple redundant sensor signals are present, an error in one sensor signal can be detected if it deviates too much from the other signals, particularly from the average of all sensor signals. The required deviation for an error to be detected can be predefined.This can also be adjusted adaptively during operation, especially if an initially set deviation proves to be unsuitable, i.e., too low or too high.

[0022] If an error is detected in the sensor signal, this does not usually lead to the signal being immediately marked as faulty. This would mean that a single error, even one caused by a measurement error, would result in the sensor signal being used only to a limited extent or not at all in the future. Therefore, it is stipulated that the signal is only marked as faulty when the signal 20 reaches or exceeds a threshold value 22 (reference digit 24). The threshold value 22 determines the fault detection time 26. If the threshold value 22 is increased, the fault detection time 26 also increases, and vice versa.

[0023] The basic principle is that in the event of an error, the qualification counter is set, i.e., incremented or decremented. If the qualification counter is incremented, it can be configured that reaching or exceeding the threshold value triggers the marking of the sensor signal as faulty. If the qualification counter is decremented, it can be configured that reaching or falling below the threshold value triggers the marking of the sensor signal as faulty.

[0024] Fig.Figure 2 shows an example of a case where a significant sensitivity error, either gain or loss, has been detected. The diagram shows a graph 50, with time plotted on its abscissa 52 and the counter value on its ordinate 54. The diagram also shows a first threshold 60, a second threshold 62, and a curve 64 representing the value or state of the qualification counter over time.

[0025] In this example, it is assumed that initially there was an amplification error (reference digit 70), but this was only temporary; later, a significant reduction in the signal is detected (reference digit 72). This continues over time (reference digit 74). Therefore, the qualification counter eventually reaches the second threshold value of 62, and the signal is set to invalid.

[0026] Fig.Figure 3 shows a purely schematic view of a vehicle designated by the reference numeral 100. This vehicle 100 has an inertial measuring unit 102 in which a number of sensors 104 are provided, each of which delivers a sensor signal 106. These redundant sensor signals 104 are transmitted to an arrangement 108, in which an evaluation unit 110 is provided for checking the sensor signals 104. Furthermore, a bidirectional qualification counter 112 is provided.

[0027] Fig.Figure 4 shows a possible flowchart of the presented procedure. In step 150, the procedure begins with the starting of a vehicle. In step 152, sensor signals from an inertial measurement unit are acquired, and at least one of these is monitored. Typically, all signals are monitored, and if a deviation is detected in one of these signals, a bidirectional qualification counter is set in step 154, i.e., incremented or decremented, depending on the deviation, i.e., its magnitude and sign. If a threshold value is reached by the qualification counter in step 156, the affected signal is marked as faulty in step 158.

[0028] In principle, each monitored signal can be assigned a qualification counter. Alternatively, a qualification counter is provided for all monitored signals, especially redundant ones. In summary, it can be stated that:

[0029] Using a bidirectional qualification counter is helpful in detecting offset and sensitivity errors in redundant signals. Employing different incremental and decremental steps in the counter increases its robustness.

[0030] The table below shows an example of how the qualification counter can be updated. Counter value Deviations detected during current review Is this deviation relevant? Final step used to update the counter >0 Positive offset (or amplification) Yes + Step pain no -Step Disqualification 0 Not applicable -Disqual Negative offset (or reduction) Yes -Grand School Dis no =0 Positive offset (or amplification) Yes +StepQuarantine no 0 0 Not applicable 0 Negative offset (or reduction) Yes -StepQuarantine no 0 <0 Positive offset (or amplification) Yes +LargeSchrDis no 0 Not applicable +Step Disqualification Negative offset (or reduction) Yes -StepTorture no +Step Disqualification

[0031] The first step is to check whether there is a deviation, positive or negative offset, gain or loss, in the signal. If so, it is necessary to evaluate whether this deviation is significant, i.e., large enough, which may indicate that the signal is faulty.

[0032] Typical methods for detecting a deviation are based on comparisons between redundant signals. For example, if one signal shows a deviation, a difference is observed between the affected signal and the other redundant signals.

[0033] The first column of the table indicates the previous sign and value of the qualification counter. This can be positive, negative, or 0. The second column indicates the type of deviation detected during the current check: positive, negative, or 0. A value of 0 means no deviation is detected. Similarly, in the case of a sensitivity monitoring system, an amplitude or attenuation variation in the evaluated signal can be detected.

[0034] The counter is set or updated depending on whether the detected deviation is significant or not, and according to the sign of the counter and the deviation. The counter is updated as follows: Final counter = previous counter + final step.

[0035] The value of the final step depends on: 1) the previous value of the counter, first column in the table, 2) the type of deviation, second column in the table, 3) Whether the recorded deviation is considered relevant or not. One way to do this is, for example, to compare the recorded deviation with a threshold value. If the deviation is above the threshold value, the deviation is considered relevant; see column 3 in the table.

[0036] The last column in the table concerns the final step used to update the counter. Three different types of steps are defined: a) StepQualify (StepQual): if a qualification of the counter is required, this value is used. b) StepDisqualify (StepDisqual): If a counter disqualification is required, this value is used. c) LargeStepDisqualify (LargeStepDis): In some cases, e.g., when the sign value differs from the sign of the deviation, a large step is used to disqualify the counter, thus increasing the robustness of the data collection.

[0037] It should be taken into account that in the previous example, which was in the Fig. As shown in 2, the absolute value of LargeStepDisqualify is twice as large as the absolute value of StepQualify.

Claims

[1] Method for monitoring at least one sensor signal (106) in a vehicle (100) wherein, when a fault is detected in the at least one sensor signal (106), a bidirectional qualification counter (112) is set and when the qualification counter (112) reaches one of at least two threshold values ​​(60, 62), the sensor signal (106) is marked as faulty, taking into account the type of deviation when setting the bidirectional qualification counter (112). [2] Method according to claim 1, wherein the sign of the deviation is used to determine whether the qualification counter (112) is incremented or decremented. [3] Method according to claim 1 or 2, wherein a plurality of redundant sensor signals (106) are monitored. [4] Method according to claim 3, wherein sensor signals (106) of an inertial measuring unit (102) are monitored. [5] Method according to any one of claims 1 to 4, wherein positive and / or negative offset errors are detected. [6] Method according to any one of claims 1 to 4, wherein sensitivity errors, e.g. gain errors and / or attenuation errors, are detected. [7] Method according to any one of claims 1 to 6, wherein the step height is adjusted when setting the qualification counter (112). [8] Method according to any one of claims 1 to 7, wherein two threshold values ​​(60, 62) are used. [9] Arrangement for monitoring at least one sensor signal (106) with an evaluation unit (110) configured to perform a method according to any one of claims 1 to 8. [10] Arrangement according to claim 9, comprising a bidirectional qualification counter (112).

Images