Procedure for checking a selected signal

Abstract

A method for checking a selected signal selected from a plurality of signals provided by an inertial measuring unit, wherein the selected signal is compared with the other signals, and wherein a trend of the selected signal and an accumulator value associated with the selected signal are taken into account during the checking.

Description

[0001] The invention relates to a method for checking a selected signal selected from a plurality of signals provided by sensors of an inertial measuring unit, and to an arrangement for carrying out the method. State of the art

[0002] An inertial measurement unit (IMU) is a spatial combination of multiple inertial sensors, such as acceleration sensors and yaw rate sensors. Such a measurement unit is used, for example, as a sensory measurement unit in an inertial navigation system.

[0003] Inertial measurement units, which typically have high performance, are used, for example, to measure the movement of a vehicle in terms of acceleration and angular rate. IMUs are also widely used in autonomous driving (AD) applications, where high signal reliability and high accuracy requirements are required.

[0004] Many applications, such as the AD applications mentioned above, require a reliable sensor signal. Redundant IMU sensors, such as three angular rate sensors, all measuring the same physical event and mounted on the same printed circuit board (PCB), are typically used to detect sensor errors by comparing how close the redundant signals are to each other.

[0005] The redundant signals identified or classified as valid are then combined or fused. This step is also referred to as fusion. Alternatively, a dropout algorithm can be implemented to select the "best" final signal from all possible signals. This signal is referred to herein as the selected signal.

[0006] Currently, many selection algorithms are available to select the selected signal from among the redundant signals. Most of these algorithms are based on a voting mechanism. The signals are compared with each other. If any signal does not match or differs from the other signals, that signal is not considered for selection. A priority-based scheme can be implemented to define the selected signal.

[0007] In general, due to the tolerances of the IMU sensors, a monitor or voter may select different selected signals consecutively within a short period of time. This can lead to some degree of distortion in the selected signal. Disclosure of the invention

[0008] Against this background, a method according to claim 1 and an arrangement according to claim 11 are presented. Embodiments emerge from the dependent claims and from the description.

[0009] A method for verifying a selected signal from among several signals provided by an inertial measurement unit is presented. The selected signal is compared with the other signals, taking into account a trend of the selected signal and an accumulator value associated with the selected signal during verification.

[0010] The proposed method avoids unnecessary switching during selection and maintains the selected signal for as long as possible despite minor temporal deviations between the signals. At the same time, rapid switching can be performed if the selected signal exhibits a significant deviation from the other redundant signals.

[0011] The selection mechanism implemented in the presented method enables rapid response to a significant deviation in the selected signal by incorporating information regarding the signal's tendency or trend. Furthermore, unnecessary switching in the selected signal is avoided by implementing a selection mechanism based on accumulators and a cost function.

[0012] The presented arrangement is configured to carry out the method described herein and can be implemented, for example, in hardware and / or software. The arrangement can be integrated into a control unit, in particular a control unit in a motor vehicle, or can be designed as such a control unit. This arrangement has an evaluation unit or evaluation means configured to carry out the method.

[0013] Further advantages and embodiments of the invention will become apparent from the description and the accompanying drawings.

[0014] It is understood that the features mentioned above and those to be explained below can be used not only in the combination specified in each case, but also in other combinations or on their own, without departing from the scope of the present invention. Short description of the drawings Fig. 1 shows a flowchart of a possible sequence of the presented procedure. Fig. 2 shows a graph of an accumulator update over time. Fig. Figure 3 shows a comparison between two different increase steps in two graphs. Fig. 4 shows a schematic representation of a vehicle with an arrangement for carrying out the described method. Embodiments of the invention

[0015] The invention is illustrated schematically in the drawings using embodiments and is described in detail below with reference to the drawings.

[0016] Fig. Figure 1 illustrates an embodiment of the presented method in a flowchart. In a first step 10, the redundant IMU signals s_1, s_2, s_3 are preprocessed. For example, a time allocation mechanism is implemented to correct the phase differences between the signals. Subsequently, in a step 12, the mean value or median s_median is calculated from the redundant IMU signals. Each signal and the median are filtered in a step 14 with two low-pass filters "a" and "b" to obtain two different filtered signals. The first filter "a" removes the noise, e.g., at a cutoff frequency of 25 to 30 Hz. The signals after the first filter are referred to as sfa_1, sfa_2, sfa_3, and sfa_median. The second filter is a strong or rigid low-pass filter, e.g. at 3 Hz, which reacts slowly to signal changes and represents the long-term trend of the signal over time.After the second filter, the signals are referred to as sfb_1, sfb_2, sfb_3 and sfb_median, respectively.

[0017] With the exception of the preprocessing and filtering steps, the algorithm can be executed at a lower frequency. For example, if the signals and filters are updated every millisecond, the rest of the algorithm can be executed at a cycle time of 10 to 20 milliseconds.

[0018] In the first algorithm execution cycle, any signal s1, s2, or s3 is taken as the selected or final signal "s_sel." This decision can be based on fixed priorities given to redundant signals (step 16).

[0019] In step 18, the algorithm begins by checking whether the "Selection_Flag" flag is set or not. If the Selection_Flag is 1 (arrow 20), this means a new selection is required because the currently selected signal is not suitable. To enable a selection in the first execution cycle, this flag should be initialized to 1.

[0020] If a new selection is required, i.e., Selection_Flag is equal to 1, then in step 22, certain conditions are checked for each of the redundant signals. The idea behind this is to filter out those redundant signals that are not suitable to be the newly selected signal. The suitable signals are contained in a candidate list.

[0021] A condition that should be fulfilled for each signal, e.g. for s_1, can be, for example: |dfa_sel_m|>|dfa_1_m|

[0022] Where dfa_sel_m = (sfa_sel - sfa_median), this expression represents the difference between the currently selected signal and the median, using the signals obtained after the first filter "a". Similarly, dfa_1_m = (sfa_1 - sfa_median), which refers to the distance between the evaluated signal s_1 and the median, using the signals after the filter "a". If a redundant signal is closer to the median than the currently selected signal, then this signal is a potential candidate to be the new selected signal. If this is not the case, this signal can be removed from the candidate list. Additional conditions can also be implemented, e.g., to confirm that the signals in the candidate list are valid.

[0023] If it is determined at step 24 that no signal meets the conditions (arrow 26), the currently selected signal does not change and a new cycle 28 is executed.

[0024] For each signal in the candidate list that meets the above conditions, a cost function is calculated in step 30 as follows, e.g. for s_1: C_s1=Ka*|dfa_1_m|+Kbn*|dfb_1_m|+Kc*|sfa_sel−sfa_s1|+Kd*|sfb_sel−svb_s1|

[0025] Here, Ka, Kb, Kc, and Kd are constants that can be calibrated and that balance the impact of the four terms within the cost function. The first term of the cost function evaluates how close the currently filtered signal is to the calculated median, taking into account the signals filtered with the first filter "a."

[0026] The second term concerns the same difference, but uses heavily filtered signals. The motivation for this is to consider the history and trend of the signal over time. The heavily low-pass filtered signals are quite robust against temporal fluctuations. The third term of the cost function concerns the difference between the currently selected signal and the evaluated signal s_1, considering both signals after filter "a." The last term concerns the difference between the selected signal and the evaluated signal, considering the signals after the strong low-pass filter "b."

[0027] With this cost function, the signals that are close to the median and at the same time close to the currently selected signal have smaller cost functions.

[0028] Finally, the signal associated with the minimum cost function is used as the new selected or final signal s_sel (step 32). Thus, the flag selection_flag is set to 0, indicating that a new selection is no longer necessary.

[0029] In the next cycle, if selection_flag is equal to 0 (step 18, arrow 40), a check is performed to ensure that the selected signal is appropriate. For this purpose, the absolute difference between the selected signal and the median is calculated in step 42. If the absolute dfa_sel_m is higher than a threshold "Thd1," then an accumulator counter (step 50) is updated using a specific increment step (step 48). If this is not the case, a decrement step, which can be 0, is used (step 58) to update the accumulator counter (step 60).

[0030] The following applies: If|dfa_sel_m=sfa_sel−sfa_median| <Thd1 %% Abstand zu Median u¨bersteigt einem Schwellwert use Inc_Step (increment step) otherwise use Dec_Step (decrement step)

[0031] The threshold Thd1 is calculated as follows: Thd1=Ct+pt*|sfa_median|, where Ct is a constant that can be calibrated. The second term of the equation and the factor pt account for an increment of the threshold, which is proportional to the median signal after filter "a," i.e., sfa_median. This second term is provided to account for the sensitivity error in the threshold definition, whose absolute deviation is higher for large signals.

[0032] The increment step “Inc_Step” can be calculated as follows: Inc_step=A*(|dfa_sel_m|−Thd1)+B*Delta, where Delta=(dfa_sel_m−dfat0_sel_m).

[0033] The first term of Inc_step increases when the actual or current difference of the selected signal from the median is greater than Thd1. The larger the difference, the larger the resulting increment.

[0034] The second term is generated to account for the trend of the signal and includes the delta factor, which is the difference between the current dfa_sel_m and a value that is at a previous time t0, e.g. dfa t0 _sel_m. Therefore, in the case of a short offset jump relative to the median, the delta factor will assume positive values ​​during the leading edge and negative values ​​during the trailing edge. Note that the signals after filter "a" are used. Alternatively, and depending on the cutoff frequency used for filter "b," the signals after the second filter can also be used to calculate the delta factor.

[0035] The second term of the Inc_step step allows for a faster response to sudden large deviations between the median and the selected signal. If this deviation does not change over time, the second term of Inc_step will be close to 0.

[0036] In the Inc_step equation, the constant A can be set higher than B to give greater weight to the current measurement rather than to the values ​​related to the predicted signal trend. Note that the Inc_step increment must be able to take values ​​≥ 0, so saturation to 0 or a minimum small positive value should be implemented. If necessary, the factor B can be a function of Delta and take different values ​​depending on the sign of the delta factor. This can result in different increments during the leading and trailing edges of a deviation from the median.

[0037] As mentioned earlier, the effect of the second term in the Inc_step formula is to increase faster than a displacement step detected in the selected signal, allowing for a quick switch to another signal. Furthermore, if there is only a brief peak and the difference between the selected signal and the median is rapidly decreasing, the Inc_step will decrease to allow for a longer delay before triggering an unnecessary switch.

[0038] Once the Inc_step is calculated, the accumulator Acc is updated as follows: Acc=Acc+Inc_step

[0039] As mentioned above, if |dfa_sel_m| < Thd1, the Dec_step is used to decrement the counter until it reaches 0. The value of Dec_step can be a calibratable constant or a function, which is calculated as follows: Dec_step=C*(|dfa_sel_m|−Thd1)

[0040] This function allows the accumulator to decrement more slowly when |dfa_sel_m| is closer to Thd1. The factor C can be calibrated.

[0041] After Dec_step is defined, the accumulator is updated as follows: Acc=Maximum[0,(Acc−Dec_step)]

[0042] After the accumulator has been updated, either with an incremental or decremental step, if the accumulation is less than the calibratable threshold AccThd (step 70), the currently selected signal retains its state and no new selection is performed in the next cycle. The Selection_Flag flag is set to 0 (step 72).

[0043] In contrast, if the accumulator continuously increments and exceeds the threshold AccThd, the Select_Flag flag is set to 1 (step 76), indicating that a new selection is required in the next cycle. Furthermore, the accumulator is reset (step 74) if Acc > Acc_Thd.

[0044] Fig. Figure 2 shows two graphs of signal waveforms to illustrate the presented method. The waveforms of +Thd1 110, sfa_sel 112, s_median 114, and -Thd1 116 are plotted in a first graph, with time plotted on the abscissa 102 and signal values ​​plotted on the ordinate 104. A second graph 130, with time plotted on the abscissa 132 and accumulator values ​​plotted on the ordinate 134, shows the waveforms of Acc_Thd 140, inc_step#1 142, and inc_step#2 144. Furthermore, the ranges tA 150, tB 152, tC 154, tD 156, and tE 158 are displayed.

[0045] Fig. Figure 2 shows an update of the accumulator over time. No precise calculations are shown; the illustration is for illustrative purposes only. When the accumulator reaches the threshold Acc_Thd 140, a new selection of the selected signal is required.

[0046] Fig. To clarify this, Figure 2 illustrates an example using two different incremental steps. inc_step#1 142 considers only the first term of the increment equation encompassed herein. inc_step#2 144 considers both the first and second terms of the equation as described herein.

[0047] For simplicity, only the selected or final signal sfa_sel 112 and the median sfa_median 114 are shown. During the period "tA" 150, i.e., the leading edge of the deviation, the difference between the signal sfa_1 and the median is higher than Th1, therefore the accumulator is incremented using an increment step.

[0048] It is expected that during "tA" 150, the step inc_step#2 144 is larger than the step inc_step#1 142 and causes a higher increment in the accumulator. During the period "tC" 154, i.e., trailing edge, inc_step#2 144 is smaller than inc_step#1 142 and therefore the accumulator increases more slowly with inc_step#2 144 than when using inc_step#1 142. During the period "tB" 152, the distance between the signal and the median remains more or less constant and therefore both increment steps inc_step#1 142 and inc_step#2 144 are equal. During "tD" 156, the absolute difference between sfa_sel 112 and the median 114 is smaller than Thd1. In this example, a constant decrement step is used and the accumulator is decremented or decreased until it reaches 0.

[0049] Fig. Figure 2 further illustrates that when a sudden large deviation peak is later detected (“tE” 158), it is expected that the use of inc_step#2 144 will result in faster switching of the selected signal compared to the use of inc_step#1 142.

[0050] Fig. Figure 3 shows two different graphs. A first graph 200, with time plotted on the abscissa 202 and accumulation plotted on the ordinate 204, shows the plots of inc_step#1 210 with A equal to 2 and B equal to 0 and inc_step#2 212 with A equal to 1 and B equal to 14. A second graph 230, with time plotted on the abscissa 232 and a difference value plotted on the ordinate 234, shows the plot of |dfa_sel_m| - Thd1 236.

[0051] Fig. Figure 3 shows a comparison between the two different increments. The difference value refers to (dfa_sel_m - Thd1). The accumulator increases according to the calculated increments if (dfa_sel_m - Thd1) is positive.

[0052] Fig. Figure 3 shows another example with two different configurations for Inc_step. Trace 236 concerns the difference between dfa_sel_m and Thd1. As already mentioned, the accumulator begins to increment when the difference is positive. If the difference is 0 or negative, the accumulator decrements. This example uses a constant decrement step.

[0053] Plots 210 and 212 show the accumulator value over time with two different increment steps: inc_step#1 has only the first term of the equation, i.e. B = 0, and the factor A is set to 2.

[0054] For the second increase step inc_step#2 applies: A=1 and B=14.

[0055] Assuming that inc_step#2 can be negative, saturation is additionally implemented to obtain a minimal positive increment. It can be observed how the accumulator increments rapidly during the rising edge of the black curve by using inc_step#2 instead of inc_step#1. It can also be noted that when trace 236 returns to 0 after about 180 ms, inc_step#2 is much smaller than inc_step#1. Therefore, the accumulator decrements more slowly than when using inc_step#1.

[0056] Fig.4 shows, in a highly simplified manner, a vehicle 250 in which an arrangement 252 for carrying out the presented method is provided. Furthermore, an inertial measuring unit 254 is provided in the vehicle 250, in which three sensors 256, for example, yaw rate sensors and acceleration sensors, are provided, which in particular provide redundant signals 260, 262, 264. From these redundant signals 260, 262, 264, a selected signal 270 is selected using the arrangement 252. If, in this case, signal 260 is selected as the selected signal, the selected signal 270 is compared with the other signals 262, 264 according to the presented method to determine whether this selected signal 270 is suitable. If this is not the case, a new signal is selected as the selected signal 270.

Claims

[1] A method for checking a selected signal (270) selected from a plurality of signals (260, 262, 264) provided by an inertial measuring unit (254), wherein the selected signal (270) is compared with the other signals (262, 264), and wherein a tendency of the selected signal (270) and an accumulator value associated with the selected signal (270) are taken into account during the checking. [2] Method according to claim 1, wherein the comparison of the selected signal (270) with the other signals (262, 264) takes into account a median signal obtained by averaging the plurality of signals (260, 262, 264). [3] Method according to claim 1 or 2, wherein, if the comparison of the selected signal (270) with the other signals (262, 264) results in a deviation that exceeds a threshold value, a new signal is selected from the plurality of signals (262, 264) as the selected signal (270), taking into account a cost function of the plurality of signals (260, 262, 264). [4] Method according to claim 2 and 3, wherein the cost function takes into account a distance between the median signal and the selected signal (270). [5] Method according to one of claims 1 to 4, wherein during preprocessing of the signals (260, 262, 264) a time allocation mechanism is implemented to correct phase differences between the signals (260, 262, 264). [6] Method according to one of claims 1 to 5, wherein the signals (260, 262, 264) are filtered with two filters, a first filter serving to suppress noise and a second filter serving to clarify a tendency of the signals (260, 262, 264). [7] Method according to one of claims 1 to 6, in which the accumulator is incremented. [8] Method according to one of claims 1 to 6, in which the accumulator is decremented. [9] Method according to one of claims 1 to 8, wherein at the beginning a signal is selected from the plurality of signals (260, 262, 264) by taking into account priorities of the signals (260, 262, 264). [10] Method according to one of claims 1 to 9, in which a candidate list is used which contains appropriate signals which are suitable to serve as the selected signal (270). [11] Arrangement for checking a selected signal (270) selected from a plurality of signals (260, 262, 264) provided by an inertial measuring unit (254), wherein the arrangement (252) comprises an evaluation unit configured to carry out a method according to one of claims 1 to 10.

Images