Method for evaluating redundant signals
By conducting statistical analysis of redundant signals in motor vehicles and applying random walk models, the threshold is dynamically adjusted to mark the correlation of signals, solving the problem of low performance in redundant signal monitoring in the prior art, achieving more accurate error recognition and improved adaptability.
Patent Information
- Application Number
- CN202380079018.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-11-14
- Filing Date
- 2023-10-18
- Publication Date
- 2025-06-20
AI Technical Summary
The existing redundant signal monitoring schemes exhibit low performance when detecting errors, and cannot identify most errors that have been injected into redundant signals, resulting in signal uncorrelatedness, and error conditions and thresholds are not representative.
By statistically analyzing the redundant signals in motor vehicles, dynamically adjusting the threshold value in combination with driving condition information, a random walk model is used to estimate the signal deviation and compare it with the defined threshold value to mark the correlation of the signal.
It improves the performance of redundant signal monitoring, can more accurately identify errors, adapt to different driving scenarios and vehicle dynamics, and enhances the signal correlation verification ability.
Smart Images
Figure CN120188115A_ABST
Abstract
Description
Technical field
[0001] The invention relates to a method for evaluating redundant signals and a component for performing the method. Background art
[0002] In a motor vehicle, a plurality of signals are detected during operation, each of which carries information about a technical variable. The information is analyzed and evaluated in the motor vehicle. To ensure safe operation, it is necessary to check whether the detected signals are correct or contain errors. For this purpose, the use or detection of redundant signals provides the possibility. This means that the technical variable is collected multiple times, i.e., there are at least two signals representing the technical variable. In particular, the credibility of these signals can be verified by comparing the two signals. Therefore, errors that occur during the detection and / or transmission of these signals can also be identified.
[0003] A possible use of the method proposed herein can be given, for example, in the form of an association with an inertial measurement unit. An inertial measurement unit is understood as a spatial arrangement composed of a plurality of inertial sensors, such as acceleration sensors and angular velocity sensors. Such a measurement unit is used, for example, as a sensing measurement unit of an inertial navigation system.
[0004] Six possible kinematic degrees of freedom can be detected using an inertial measurement unit. For this purpose, three acceleration sensors that are respectively orthogonally arranged to each other are used to detect translational motion, and three rotational speed sensors that are orthogonally mounted to each other are used to detect rotational or circular motion. In this case, the inertial measurement unit provides three linear acceleration values of the translational motion and three angular velocity values or angular rates of the rotational speed as measurement values.
[0005] If at least two signals are detected for one variable at this time, they are called redundant signals. A scheme for processing these redundant signals in order to particularly verify the credibility of these redundant signals, i.e., in order to identify signals that may contain errors, is called a redundancy scheme.
[0006] A currently used redundancy scheme implemented in an inertial measurement unit is called a correlation monitor, and this redundancy scheme is based on a comparison of measurements of the same physical variable, such as acceleration or angular rate, from different sources. Two parameters, namely Delta and gradient, are calculated for a given pair of redundant signals. Delta is the difference between the current or present samples. The gradient is the difference between the signal change rates, i.e., the discrete derivative. These parameters are calculated for each signal, where the current or present samples and the previous samples are used.
[0007] Once Delta and the gradient are calculated, they are compared with thresholds. In cases where these thresholds are exceeded, the counter is incremented. The conditions for incrementing this counter can be combined in three ways. These ways are: Delta only, gradient only, Delta and gradient. If two conditions for incrementing the counter are used, these two conditions are combined or associated, where the logical "OR" is used. This means that each satisfied condition causes the counter to increment. However, if both conditions are satisfied, the counter is incremented only once, not twice.
[0008] Assume: The signals are relevant until the counter reaches a boundary value. Once this boundary value is reached, the counter value is set to a holding value above this boundary value. Additionally, a relevance flag is set. This means: These signals are marked as irrelevant. As long as the conditions for incrementing are satisfied, the counter value remains at this holding value. When these conditions are no longer satisfied, the counter starts to decrement until the value reaches zero, and only then is the relevance flag removed or cleared. Thus, the signals are marked as relevant again.
[0009] This known process is executed, where the last measured value and the value immediately preceding it are used. In this sense, this approach is practically instantaneous. Thus, in this comparison, there is no memory or storage operation per se. However, this is currently provided for the correlation monitor by using a counter for debouncing the flag.
[0010] The adjustable parameters for this implementation of the redundant signal pair are implemented as follows:
[0011] - The threshold of Delta,
[0012] - The threshold of the gradient,
[0013] - The conditions for incrementing, Delta only, gradient only, or both,
[0014] - The counter boundary value,
[0015] - The counter holding value. Summary of the Invention
[0016] Against this background, a method having the features of claim 1 and a component according to claim 11 are proposed. The embodiments follow from the dependent claims and the description.
[0017] The proposed method is used for evaluating redundant signals, such as redundant signals detected by an inertial measurement unit, for example. However, the method is not limited to this implementation.
[0018] One implementation of the proposed method is based on the insights gained from the following challenges that occur during the operation of the inertial measurement unit.
[0019] If the redundancy monitoring scheme is tested by injecting errors using vehicle measurements from a test catalog, the redundancy monitoring scheme currently only shows low performance. Known implementations have so far been unable to identify most of the errors injected into the redundant signals. This causes signal irrelevance. Additionally, in this case, potential customers strive to lower the correlation threshold.
[0020] Furthermore, it has been recognized that the error scenarios and thresholds are not representative. Not only the error types and their magnitudes, but also the thresholds in the correlation monitor implementation, are defined under the strong influence of the most recent generations of inertial measurement units, where it is expected that these signals are used in the braking system (ESP: Electronic Stability Program) in most cases. The inertial measurement unit is mainly focused on autonomous driving applications, which makes it necessary to rethink the schemes for error definition and threshold definition.
[0021] At this time, the correlation monitor in the inertial measurement unit is a static solution with fixed programming. This means that once the sensor is flash programmed, the parameters of the redundancy monitoring basically remain application-independent. Not being adjustable is equivalent to fixed programming. Additionally, this feature does not consider vehicle dynamics in the current implementation. This means that the threshold also remains independent of the driving scenario. Therefore, the lack of adaptability means it is static. "Not being able to adjust the parameters and make them adapt to the environment" goes against the current trend in the automotive sensor market, especially in autonomous driving applications.
[0022] The proposed method is used to evaluate redundant signals in a motor vehicle. In this method, a first signal is compared with a second signal, and the result of this comparison is compared with a threshold. Here, the threshold is evaluated based on a statistical analysis of the signals and determined based on information about the driving condition of the motor vehicle.
[0023] The proposed method redefines the way for monitoring signal redundancy. To this end, error scenarios related to autonomous driving scenarios can be examined together with vehicle dynamics and signal noise characteristics. The basis is to find safety-critical decisions mainly based on signals, such as inertial measurement unit signals, and define the maximum allowable signal deviation. Here, the current driving condition and the influence of noise on the integrated signal are considered.
[0024] In the proposed solution, in one configuration, a similar approach is applied to the acceleration signal and the angular rate signal within a typical autonomous vehicle architecture. First, the random or probabilistic walk (random walk), i.e., the angle or position, is estimated based on the signal. Then, this estimate is incorporated into the formula for the maximum allowable signal deviation. Finally, the measured signal deviation is compared with a defined threshold. The information thus obtained is converted into a flag.
[0025] In different steps of the process, external parameters are used for calculations, which enables the OEM (Original Equipment Manufacturer) to configure the characteristics of the target vehicle and adapt the target vehicle to the corresponding driving conditions. Additionally, the signal characteristics are estimated, where the previous samples are used to dynamically evaluate the estimation quality.
[0026] Subsequent discussion will be about the implementation of the proposed method:
[0027] - Random Walk
[0028] The first step is to define the error caused by the integration of the noise signal. For the angular rate signal, only one integration step is required to obtain the random angle walk. In the case of the acceleration signal, two integrations need to be performed to obtain the random position walk. In both cases, the input parameters for the calculation are the integration time, the noise characteristics of the original signal, and the signal sampling frequency. For this, reference is made to Figure 1 and Figure 2 . This step only considers the noise in the signal to estimate the random walk. By varying the parameters, the characterization surface can be defined, see Figure 3 and Figure 4 .
[0029] - Acceleration Signal Deviation
[0030] In the case of using longitudinal control, i.e., the control via the accelerator pedal and braking, as a reference, the scenario considered here assumes that the vehicle is driving on a straight road and suddenly has to perform an emergency brake, where coupled positioning navigation is used. The goal is to stop the vehicle within a predefined longitudinal safety boundary. In this case, this is not possible. The vehicle should either perform an emergency avoidance maneuver or trigger or activate a damage mitigation system.
[0031] After performing a mathematical transformation on the equations for linear motion and discrete integration, and combining some worst-case assumptions, the maximum allowable deviation is represented by the following equation:
[0032]
[0033] Here, V0 is the initial longitudinal speed of the vehicle, A lonis the expected longitudinal deceleration during braking, S lim is the longitudinal safety margin, and dt is the signal sampling time, see Figure 5 .
[0034] - Angular rate signal deviation
[0035] Using the lateral control, i.e., the control by steering, as a reference, the scenarios investigated herein assume that the vehicle is driving on a curved road and must stay in the lane, where coupled positioning navigation is used. The aim is to keep the vehicle within a predefined lateral safety margin. In this case, this is not possible. The vehicle should either initiate decoupling (Abkopplung) or perform an emergency brake.
[0036] After mathematical transformation of the equations for circular motion and discrete integration, combined with some worst-case assumptions, the maximum allowable deviation is represented by the following equation:
[0037]
[0038] Here, V lon is the longitudinal speed of the vehicle, T dr is the coupling time, S lim is the lateral safety margin, and dt is the signal sampling time, see Figure 6 .
[0039] - Debouncing and final marking components
[0040] The final step in implementing this method is to estimate the value of the random walk using a threshold formula and compare this estimate with the measured signal deviation. The implementation for acceleration and angular rate is slightly different, but the strategy is the same. Before transforming the result into a correlation mark, the comparison needs to be debounced, see Figure 7 and Figure 8 .
[0041] If necessary, some of the parameters described in this method can be fixed-programmed. But the idea is to have a general description that can adapt to each specific situation.
[0042] The described components are set to perform the method proposed herein and can be implemented, for example, in hardware and / or software. Additionally, the components can be integrated in the vehicle's control device or can itself be constructed as such a control device.
[0043] Further advantages and configurations of the present invention result from the description and the drawings.
[0044] Of course, without departing from the scope of the present invention, the features described above and those to be explained below can be used not only in the given combinations respectively, but also in other combinations or alone. Description of the Drawings
[0045] Figure 1 The calculation of the random position walk is shown in the diagram.
[0046] Figure 2 The calculation of the random angular walk is shown in the diagram.
[0047] Figure 3 The position walk estimation surface is shown in the graph.
[0048] Figure 4 The angular walk estimation surface is shown in the graph.
[0049] Figure 5 The maximum deviation of the acceleration signal is shown in the graph.
[0050] Figure 6 The maximum deviation of the angular rate signal is shown in the graph.
[0051] Figure 7 An embodiment of the proposed method is shown in the flowchart.
[0052] Figure 8 Another embodiment of the proposed method is shown in the flowchart. Detailed Description of the Invention
[0053] The present invention is schematically illustrated by way of embodiments in the accompanying drawings and will be described in detail below with reference to the drawings.
[0054] The method will be described hereinafter in connection with an inertial measurement unit, but the method is not limited to such an implementation and can be used anywhere in a motor vehicle, particularly where signals that exist redundantly are to be evaluated.
[0055] Figure 1 The calculation of the position walk is illustrated in Chart 10. It should be noted here that:
[0056] A motor vehicle uses an inertial measurement unit for coupled positioning navigation for a duration of T dr seconds, and the only difference between the redundant signals is given by the noise parameters. The noise in the redundant signals is integrated twice. The standard deviation of the resulting position, i.e., the linear walk, is a function of the noise characteristics in A lon (i.e., σ A and σ B ) in the redundant signals, T dr and the signal sampling time dt.
[0057] The input variable 12 is the integration time T dr , the sampling frequency 1 / dt, and the noise σ of the acceleration sensor Acc . The output variable 14 is the position random walk σ Pos . The first graph 16 shows the acceleration change curve, the second graph 18 shows the velocity change curve, the third graph 20 shows the position, the fourth graph 22 shows the histogram of the noise, the fifth graph 24 shows the histogram of the end point, and the sixth graph 26 also shows the histogram of the end point
[0058] Figure 2 The calculation of the angular random walk is described in Chart 40. It should be noted here that:
[0059] The motor vehicle uses the inertial measurement unit for coupled positioning navigation for a duration of T dr seconds, and the only difference between the redundant signals is given by the noise parameter. The noise in the redundant signals is integrated once. The standard deviation of the resulting position, i.e., the angular random walk, is a characteristic of the noise in the yaw rate (i.e., σ A and σ B ), T dr , and the function of the signal sampling time dt
[0060] The input variable 42 is the integration time T dr , the sampling frequency 1 / dt, and the rate noise σ Rate . The output variable 44 is the angular random walk σ Ang . The first graph 46 shows the angular rate, the second graph 48 shows the angle, the third graph 50 shows the histogram of the noise, and the sixth graph 52 shows the histogram of the end point
[0061] Figure 3 The position random walk estimation surface at a sampling rate of 1000 Hz is shown in Graph 80. The signal noise is plotted on one abscissa 82 of this graph, the integration time is plotted on the other abscissa 84 of this graph, and the position random walk is plotted on the ordinate 86 of this graph
[0062] Figure 4 The angular random walk estimation surface at a sampling rate of 1000 Hz is shown in Graph 100. The signal noise is plotted on one abscissa 102 of this graph, the integration time is plotted on the other abscissa 104 of this graph, and the angular random walk is plotted on the ordinate 106 of this graph
[0063] Figure 5The surface showing the maximum deviation of the acceleration signal is shown in graph 120, where the initial vehicle speed is plotted on one abscissa 122 of the graph, the expected maximum deviation is plotted on another abscissa 124 of the graph, and the maximum allowable deviation is plotted on the ordinate 126 of the graph.
[0064] Figure 6 The surface showing the maximum deviation of the angular rate signal is shown in graph 140, where the initial vehicle speed is plotted on one abscissa 142 of the graph, the integration time is plotted on another abscissa 144 of the graph, and the maximum allowable deviation is plotted on the ordinate 146 of the graph.
[0065] Figure 7 A possible flow of the proposed method is shown in the flow chart, and the implementation of redundant monitoring of the acceleration signal in component 198 for executing the method is schematically illustrated.
[0066] The input variables are acceleration signal (dt) A 200 and acceleration signal (dt) B 202. First, a statistical analysis and evaluation 204 is performed, which includes: calculating the absolute difference (Delta) in step 206, calculating the moving standard deviation based on the previous N samples of input variable A 200 in step 208, calculating the moving standard deviation based on the previous N samples of input variable B 202 in step 210, and calculating the moving average based on the previous N samples in step 212.
[0067] In addition, the OEM input 222, i.e., the integration time T, is considered for the driving condition 220 dr 230, safety margin S lim 232, maximum expected deceleration A lon 234 and vehicle speed V0 236.
[0068] When considering the integration time T dr 230 and the moving standard deviation 208, the first expected drift 240 is obtained. When considering the integration time T dr 230 and the moving standard deviation 210, the second expected drift 242 is obtained. In step 250, the worst-case drift combination is obtained from these two drifts 240 and 242. Then, in step 254, when considering the pre-given safety margin S lim 232, the safety margin is adjusted to include this random drift.
[0069] In step 260, when considering the maximum expected deceleration A lon and the vehicle speed V0, the maximum allowable deviation Δ is obtained using equation (1) maxThen, in step 262, the counter performs debouncing based on approximately 4. Finally, the correlation monitoring status 270 is output.
[0070] Figure 8 A possible flow of the proposed method is shown in the flowchart, and the implementation of redundant monitoring of the angular rate signal in the component 298 for executing the method is schematically illustrated.
[0071] The input variables are the acceleration signal (dt)A 300 and the acceleration signal (dt)B 302. First, a statistical analysis evaluation 304 is performed, which includes: calculating the absolute difference (Delta) in step 306, calculating the sliding standard deviation based on the previous N samples of the input variable A 300 in step 308, calculating the sliding standard deviation based on the previous N samples of the input variable B 302 in step 310, and calculating the sliding average based on the previous N samples in step 312.
[0072] In addition, the OEM input 322 is considered for the driving condition 320, i.e., the integration time T dr 330, the vehicle speed V lon 332, and the safety margin S lim 334.
[0073] The first expected walk 340 is obtained considering the integration time T dr 330 and the sliding standard deviation 308. The second expected walk 342 is obtained considering the integration time T dr 330 and the sliding standard deviation 310. In step 350, the worst-case walk combination is obtained from these two walks 340 and 342. Then, in step 354, the arccos value is obtained to include this random walk.
[0074] In step 360, considering the vehicle speed V lon and the safety margin S lim 334, equation (2) is used to obtain the maximum allowable deviation Δ max . Then, in step 362, the counter performs debouncing based on approximately 4. Finally, the correlation monitoring status 370 is output.
Claims
1. A method for evaluating redundant signals in a motor vehicle, wherein, Compare a first signal with a second signal, compare the result of the comparison with a threshold value, wherein, the threshold value is determined based on a statistical analysis evaluation (204, 304) of the signals and based on information about the driving situation (220, 320) of the motor vehicle.
2. The method according to claim 1, wherein, Within the scope of the statistical analysis evaluation (204, 304), a random walk is first determined by considering the integral of the signal over time.
3. The method according to claim 1 or 2, wherein, The statistical analysis evaluation (204, 304) includes at least one calculation step selected from the group consisting of the following calculation steps: calculating an absolute difference (206), calculating a sliding standard deviation (208, 210), calculating a sliding average (212).
4. The method according to any one of claims 1 to 3, wherein, The information about the driving situation (220, 320) is selected from the group consisting of the following information: integration time (230), vehicle speed (236), safety margin (232), maximum expected deceleration (234).
5. The method according to any one of claims 1 to 4, wherein, The method is carried out in combination with an inertial measurement unit.
6. The method according to claim 5, wherein, Evaluate the signals of at least one acceleration sensor.
7. The method according to claim 5 or 6, wherein, Evaluate the signals of at least one rotational speed sensor.
8. The method according to any one of claims 1 to 7, wherein, Use the method in an autonomously operating vehicle, wherein error situations related to an autonomous driving scenario are taken into account.
9. The method according to any one of claims 1 to 8, wherein, When the threshold value is exceeded, a counter is incremented.
10. The method according to claim 9, wherein, When the threshold value is undershot, the counter is decremented.
11. A component for evaluating signals redundantly present in a motor vehicle, the component being arranged to perform the method according to any one of claims 1 to 10.
12. The component according to claim 11, the component being arranged to be used in combination with an inertial measurement unit.