Method for synchronization signal

Abstract

A method for synchronizing signals of a plurality of participants, wherein a correlation is given between the signals by a physical relationship, wherein the signals are filtered by means of a first filter (12) respectively in order to determine a shift (24) between the signals, wherein the determined shift (24) represents a measure of a phase shift between the signals, and wherein the shift (24) is subsequently eliminated in such a way that the signals are filtered by means of a second filter (14) respectively.

Description

Technical Field

[0001] The present invention relates to a method for synchronizing signals and an apparatus for performing the method. Background Technology

[0002] In signal processing, the task of merging asynchronous signals arises in various situations, such as sensor fusion. This often occurs in distributed systems (e.g., in vehicle networks) where measurement signals from multiple sensors must be processed. Thus, methods for sensor fusion and measurement error calibration, for example, presuppose synchronized measurement data. In this sense, synchronization means that the measurement start time is the same for all measurement parameters. However, in distributed systems (e.g., in-vehicle energy networks), this is not given. Instead, in in-vehicle energy networks, measurements are distributed across multiple sub-components, each using its own internal, independent clock generator. Furthermore, signal transmission between components additionally introduces delays, which further increase the asynchronicity. For this reason, computational results based on different signal sources are strongly affected by signal asynchronicity.

[0003] Currently, various methods are being used to solve the asynchronicity problem:

[0004] 1. Use a global clock for synchronization;

[0005] 2. Associate two signals so that their asynchronicity can be determined by their similarity;

[0006] 3. Synchronization signals are transmitted via a bus system to enable adjustment of the internal clock;

[0007] 4. Connect the trigger lines of all sensors in the system so that the synchronous measurement start can be predetermined;

[0008] The methods described in the prior art have a number of drawbacks. Thus, the methods mentioned under 1. and 4. are cost-intensive in most cases. The method mentioned under 2. has proven to be computationally intensive. The method mentioned under 3. requires valuable resources throughout the system. Summary of the Invention

[0009] Against this backdrop, a method having the features of claim 1 and an apparatus according to claim 8 are proposed. Furthermore, a computer program according to claim 10 and a machine-readable storage medium according to claim 11 are also proposed. Embodiments are derived from the description and dependent claims.

[0010] The proposed method is used to synchronize signals from multiple participants (Teilnehmer), which are associated through a mathematical (or physical in configuration) relationship. The signals are filtered using a first filter to determine the shifts between them, where the determined shifts represent a measure of the phase shift. The shifts are then eliminated by filtering the signals using a second filter. This method utilizes the symmetry of the first and second filters to determine and eliminate not only positive but also negative shifts. Furthermore, this approach enables the same attenuation of all signals.

[0011] This method, for example, can be used in cable harness diagnostics within the scope of parameter determination, where parameters are determined based on input and / or measurement parameters. These input or measurement parameters are represented by signals that are asynchronous with each other. The proposed method then enables the signals to be synchronized first, and subsequently, one or more parameters to be determined or estimated in the configuration based on the synchronized signals.

[0012] Therefore, this method involves filtering the signals of multiple (e.g., two) participants separately using filters to determine the temporal shifts between their signals. These determined shifts enable the synchronization of the participants' signals in the distributed system. To synchronize the participants' signals, a mathematical relationship must be established between them. This mathematical relationship can be expressed as an equation, such as U = R * I.

[0013] Because this synchronization scheme uses digital filters, it requires no additional hardware that would incur overhead. This advantage offers significant cost savings potential in cost-critical industries. Furthermore, the filters used are typically simple digital (FIR) filters, thus requiring minimal computation and adding no additional burden to existing systems.

[0014] An FIR filter (finite impulse response) is a discrete, digitally implemented filter with a finite impulse response. It is also known as a transversal filter and is used particularly in the field of digital signal processing.

[0015] Similarly, this method does not require additional resources for the entire system, i.e., no handshake is needed, and no additional messages are needed through the bus system, thus the scalability of the system is unrestricted.

[0016] Specifically, the filter is applied to all signals so that its attenuation affects all signals. This is a positive effect in quotient formation because the attenuation is reduced and thus eliminated in this case. The method is also capable of handling both positive and negative hysteresis. The delay can be estimated, which enables the wide application of this synchronization scheme. Both effects can be achieved through the symmetry of the filter.

[0017] In principle, a fractional delay filter can be used in this method to compensate for transmission behavior. For this purpose, an arbitrary-order Lagrangian filter or other filters could be considered. To determine the hysteresis, parameter estimation methods, such as the Extended Kalman Filter (EKF), can be used.

[0018] This method, in principle, can achieve synchronization of any number of mathematically coupled signals from two or more participants. Optionally, compensation can be made using bias parameters in the EKF, which improves the estimation quality of the delay. Furthermore, the bias parameters can be monitored for reliability testing purposes.

[0019] One possible application of the proposed method is in the field of cable harness diagnostics. Furthermore, this method can also be used in a variety of ways in the field of measurement data fusion.

[0020] The proposed apparatus is used to perform the method. The apparatus is implemented in hardware and / or software. The apparatus can be integrated into or configured as a control device of a vehicle.

[0021] A computer program is also proposed, comprising program code units for performing the steps of the proposed method. This computer program can be stored on a machine-readable storage medium.

[0022] Other advantages and configurations of the invention are apparent from the description and drawings.

[0023] It is understood that the features mentioned above and to be elaborated below can be used not only in combinations described separately but also in other combinations or individually, without departing from the scope of the invention. Attached Figure Description

[0024] Figure 1 The method for estimating parameters is illustrated schematically;

[0025] Figure 2 An implementation scheme of the proposed method is shown;

[0026] Figure 3 A graph is shown to illustrate the filter's behavior;

[0027] Figure 4 This shows the modeling of the path before the filter;

[0028] Figure 5 The functional principle of synchronization is illustrated using a curve diagram;

[0029] Figure 6 A schematic diagram illustrates the apparatus for performing the proposed method. Detailed Implementation

[0030] The following section discusses the issue of asynchronicity between two signals in the context of in-vehicle networks used in motor vehicles. It should be noted that the proposed method is not limited to this application scenario, but can always be applied when two asynchronous signals need to be synchronized.

[0031] Figure 1 A schematic diagram illustrates the method used for parameter estimation. The diagram shows a first block 50 for time updates or predictions, with state prediction 52 and error covariance prediction 54, and a second block 60 for updating measurements or corrections, with Kalman gain calculation 62, measurement estimation update 64, and error covariance update 66. An initial estimate is applied at input 70.

[0032] Figure 2 A possible implementation of the proposed method is illustrated schematically. The diagram shows a physical system 10, a first Kalman filter 12 for estimating the hysteresis, a fractional delay filter 14 for synchronizing the input and measurement parameters, and a second Kalman filter 16 for estimating the parameters of the physical system 10. The measurement parameter z(k) 20 and the input parameter u(k) 22 are coupled to each other through the physical system 10. The physical system 10 is given, for example, by the equation U = R*I, where the voltage U is the measurement parameter z(k) and the current I is the input parameter u(k). The physical coupling of these two parameters is given by a resistor R.

[0033] The input parameter u(k)22 is fed into the physical system 10, the first Kalman filter 12, and the fractional delay filter 14. The first Kalman filter 12 outputs a shift or hysteresis D 24. The fractional delay filter 14 outputs u(k+D / 2*t). s ), z(kD / 2*t s 26 represents the synchronized input and measurement parameters. The second Kalman filter 16 outputs the estimated parameters 28 of the physical system 12.

[0034] As already explained, in a vehicle's onboard network or onboard energy network, diagnostic results are strongly influenced by the asynchronicity of measurement parameters. If this asynchronicity can be identified and eliminated, it has a positive impact on the diagnostic results. For this diagnostic scheme, starting with time-discrete measurement parameters, these parameters exist at the same sampling rate but have a temporal shift relative to each other, where this temporal shift is D*t. s Here, D represents the hysteresis factor between signals, which is a linear factor related to the sampling time (Abtast-bzw.Samplezeit), t s The sampling rate is used for sampling.

[0035] Therefore, among the measurement parameters mentioned, at sampling time k, assuming the shift over time is the same for each measurement parameter determining the participant, the following measurements are available:

[0036] U v (k),I v (k),U Batt (k+D*t s ),I Batt (k+D*t s )

[0037] To determine the lag factor D, the two schemes can be made different. In the first scheme, all signals of one participant are filtered, with the filter shifted in time by an initial factor D. The disadvantage of this scheme compared to the subsequent second scheme is that it only estimates and synchronizes positive delays, and only the participant's signal is attenuated by the filter. Therefore, which signal in the signal leads must be known beforehand in order to apply this scheme.

[0038] According to the second scheme, in order to determine the lag factor D, the input parameters and measurement parameters u(k) and z(k) of all participants are filtered, such as when they are in Figure 4 and Figure 5 As shown in the diagram. Here, the signal of participant 1 is shifted by N / 2 + D / 2 in time, and the signal of participant 2 is shifted by N / 2 - D / 2 in time. Therefore, the difference between the time shifts of the two participants yields the total shift factor D. The time shift of the signal can be performed using a so-called fractional delay filter or fractional hysteresis filter.

[0039] The implementation of the fractional delay filter is described below using a first-order (N=1) Lagrangian filter. Higher-order (N>1) Lagrangian filters are also possible, in which case the filter coefficients change.

[0040] u(kD*t s)=(u(k)*(1-D)+u(k-1)*D)

[0041] Therefore, the estimated measurement parameter z(k-(1 / 2-D / 2)t) is calculated according to the above method. s ) and the estimated measurement parameter h(k-(1 / 2+D / 2)t s ).

[0042] Here:

[0043] k: sampling time

[0044] z: Measurement parameter

[0045] h: The measurement parameter estimated by u through the model equations.

[0046] u: Input parameter (the parameter being measured, which is not synchronized with the measured parameter z)

[0047] To estimate the parameters of the cable bundle's lead resistance and contact resistance, the previously calculated z(k-(1 / 2-D / 2)t) can then be used. s ) and u(k-(1 / 2+D / 2)t s The value of )). Alternatively, it can be, as it is in Figure 2 As shown, a higher-quality additional filter 14, such as a higher-order Lagrangian filter, is implemented using the calculated factor D 24, and its filtered parameters are then provided to the parameter estimate. This has the advantage of achieving high signal quality for the hysteresis signal with low computational overhead for estimating the factor D. This method can also be constructed using only the Kalman filter 12, thereby losing the aforementioned advantages.

[0048] Therefore, synchronization consists of two components: a Kalman filter 12 for estimating the shift and a fractional delay filter 14 for synchronizing the signal. The shift D is a linear factor that describes the time shift between signals as D*t. s , where t s That is the sampling rate. The synchronized signal is then used by the Kalman filter 16 to estimate parameters (e.g., resistance).

[0049] For the synchronization of signals from two participants, the relative shift between them is D*t. s It is decisive. To determine this shift and be able to demonstrate the mentioned advantages, the fractional delay filter (e.g., an Nth-order Lagrangian filter) is remodeled. See [reference needed] Figure 3 .

[0050] Figure 3The behavior of a first-order (i.e., N=1) filter is illustrated in graph 100 based on curve 106. The hysteresis D is plotted on the x-axis 102, and the attenuation |A| is plotted on the y-axis 104. Double arrow 110 shows ΔD of 0.4. First arrow 112 shows z(k-(0.5-D / 2)). Second arrow 114 shows h(k-(0.5+D / 2)). Third arrow 116 shows z(k-(0.5-D / 2)). Fourth arrow 118 shows h(k-(0.5+D / 2)).

[0051] A D of 0.4 means that the signal of the first participant leads the signal of the second participant by 0.4*t. s , where t s This is equal to the signal sampling time. -0.4D means that the signal of the first participant lags behind the signal of the second participant by 0.4*t. s .

[0052] The new model utilizes the symmetry of the filter, which means that the attenuation |A| is the same with respect to the hysteresis of D and ND. Here, N corresponds to the order of the filter. This new modeling and simultaneous filtering of the signals from both participants yield two advantages: achieving the same attenuation, and enabling both positive and negative hysteresis.

[0053] In the following text, a first-order (N=1) Lagrange filter is used as an example to illustrate the filtering of the measured parameter z(k) and the estimated measured parameters, which are calculated from the filtered input parameter u(k) and the model equation h(k).

[0054] z(k-(1 / 2-D / 2)t s )=z(k)(1-(1 / 2-D / 2))+z(k-1)(1 / 2-D / 2)

[0055] h(k-(1 / 2+D / 2)t s )=h(k)(1-(1 / 2+D / 2))+h(k-1)(1 / 2+D / 2)

[0056] If the measurement equation is implemented into the parameter estimator, the lag D can be estimated using noisy signals u(k) and z(k) and the model equation h(k) explained below.

[0057] D(k)=D(k-1)+[z(k-(1 / 2-D / 2))-h(k-(1 / 2+D / 2)t s )]*

[0058] [h(k-1-(1 / 2+D / 2)t s)-h(k-(1 / 2+D / 2)t s )] -1

[0059] Figure 4 A fractional delay filter is illustrated in the diagram. Figure 2 Modeling of the path preceding reference number 14 in the diagram. This figure shows adder 150, subtractor 152, first fractional delay filter 154, and second fractional delay filter 156. A value N / 2 160 and a value D / 2 162 are applied to the input of adder 150, derived by multiplying hysteresis D 164 by a factor of 0.5 166. Similarly, a value D / 2 162 and a value N / 2 168 are applied to the input of subtractor 152.

[0060] Therefore, the adder outputs N / 2 + D / 2 = 170. The subtractor 152 outputs N / 2 - D / 2 = 172. h(k) 180 and N / 2 + D / 2 = 170 are input into the first fractional delay filter 154. z(k) 182 and N / 2 - D / 2 = 172 are input into the second fractional delay filter 156. The first fractional delay filter 154 outputs:

[0061] h(k-(N / 2+D / 2)*t s )

[0062] Output of the second fractional delay filter 156:

[0063] z(k-(N / 2-D / 2)*t s )

[0064] Modeling the shifts preceding fractional delay filters 154 and 156 allows for the mapping of positive and negative shifts between components. Due to the symmetry of fractional delay filters 154 and 156, all signals are attenuated by the same filters.

[0065] Figure 5 Graph 300 illustrates the principle of synchronization as interpolation between two measurement parameters represented by the first signal 310 and the second signal 312. k is plotted on the horizontal axis 302 of this graph, and voltage U [V] is plotted on the vertical axis 304. These signals 310 and 312 are asynchronous to each other. The correlation of this measurement is given by mathematical correlation, in this case:

[0066] U1 = U2 + I2 * R,

[0067] z(k) h(k)

[0068] In this equation, the left side is represented by the first signal 310, and the right side is represented by the second signal 312. Filtering now means interpolation between the measurement points marked by the points in the diagram. Interpolation is indicated by the straight lines in the diagram. Then, deviations in the mathematical correlation are corrected by shifting the measurements. Interpolation of the measurement parameters (i.e., U1, U2, I2) of the two components enables signal shifting. Here, the interpolated measurement parameter (U1) of the first component is shifted by a factor D / 2 + N / 2, and the interpolated measurement parameters (U2, I2) of the second component are shifted by a factor N / 2 - D / 2.

[0069] Figure 6 A schematic, highly simplified diagram illustrates an apparatus for performing the method, which is designated as reference numeral 200. In this case, apparatus 200 is configured as a control device for a vehicle.

[0070] Device 200 is connected to a first participant or control device 202 and a second participant or control device 204, wherein the first control device 202 sends a first signal 206 to device 200, and the second control device 204 sends a second signal 208 to the device. The two signals 206 and 208, each carrying a measured value as information, should be combined for analysis and processing within device 200, taking into account that the two signals 206 and 208 are asynchronous with each other. Now, the two signals 206 and 208 can be synchronized within device 200 using the method proposed herein, so that subsequent analysis and processing of signals 206 and 208 can also be performed within device 200 in the same configuration.

[0071] This method can obviously also be executed with more than two participants or control devices. Here, the control devices can be synchronized with each other. However, synchronization between one or more control devices and device 200 is also possible.

[0072] This method can be applied in a variety of ways when the following requirements are met:

[0073] The participants' signals must be mathematically correlated.

[0074] The participants' signals must have the same sampling rate, which can be addressed using a resampling filter.

Claims

1. A method for synchronizing signals (206, 208, 310, 312) of at least one participant (202, 204) of a decentralized system of vehicles, wherein, A mathematical relationship is given between the signals (206, 208, 310, 312), wherein the signals (206, 208, 310, 312) are filtered by a first filter (12) to determine a shift (24) between the signals (206, 208, 310, 312), wherein the determined shift (24) represents a measure of the phase shift between the signals (206, 208, 310, 312), and the shift (24) is subsequently eliminated by means of a second filter. (14) Filter the signal (206, 208, 310, 312), wherein a first or higher-order first fractional delay filter and a second fractional delay filter are used as the second filter (14), wherein the symmetry of the first filter (12) and the symmetry of the second filter (14) are utilized to determine and eliminate not only positive shifts but also negative shifts (24), wherein the symmetry of the second filter (14) includes the symmetry of the first fractional delay filter and the second fractional delay filter.

2. The method according to claim 1, in which the first filter (12) is implemented as a first Kalman filter, wherein The measurement equations are implemented using a first-order fractional delay filter.

3. The method of claim 2, wherein, The fractional delay filter is implemented using a Lagrange filter.

4. The method according to any one of claims 1 to 3, wherein the first Kalman filter for estimating the shift and the Lagrange filter for synchronizing the signals (206, 208, 310, 312) are combined into a single filter.

5. The method according to any one of claims 1 to 4, wherein, after the synchronization of the signals (206, 208, 310, 312), at least one parameter (28) is estimated based on the synchronized signals.

6. The method according to claim 5, wherein at least one parameter (28) is estimated by means of a second Kalman filter (16).

7. The method according to any one of claims 1 to 6, wherein the method is performed in the vehicle's in-vehicle network.

8. The method according to any one of claims 1 to 7, wherein, for the cases where the signals (206, 208, 310, 312) have different sampling rates, the signals (206, 208, 310, 312) are processed by means of a resampling filter.

9. An apparatus for synchronizing signals of a plurality of participants (202, 204), wherein, The signals (206, 208, 310, 312) are associated by a mathematical relationship, wherein the device (200) is configured to perform the method according to any one of claims 1 to 8.

10. The apparatus according to claim 9, wherein the apparatus is configured as a vehicle control device.

11. A computer program product having program code units, wherein when the computer program product is implemented on a computing unit, the program code units are configured to perform the method according to any one of claims 1 to 8.

12. The computer program product according to claim 11, wherein the computing unit is the computing unit in the apparatus (200) according to claim 9 or 10.

13. A machine-readable storage medium having stored thereon the computer program product according to claim 11 or 12.

Images