Method, device and computer program for estimating the speed of a wheeled vehicle

Abstract

The invention relates to a method for estimating the motion speed (v) of a wheeled vehicle, wherein (E5) a frequency trajectory (ξ t ) representative of the speed (v) of a wheel (RO) of the vehicle (VE) is estimated in a filtered spectrogram (S(f t ,t)) as follows: (E51) the observation probability (pz t |ξ t ) of the trajectory (f t ) is estimated based on the calculated amplitudes of the filtered spectrogram (S(f t ,t)); (E52) the posterior observation law (pz ξ |ξ ξ|Z ) proportional to the product of the probability (p t ) and the probability (p t )' is estimated; and (E53) the trajectory (f ξ|Z ) is estimated based on the law (p t ), and (E6) the motion speed (v) of the wheel of the vehicle is estimated based on the trajectory (ξ t ).

Description

Technical Field

[0001] The present invention relates to a method, apparatus and computer program for estimating the speed of a vehicle having wheels for rolling on the ground.

[0002] The invention is applied, in particular, to rolling land vehicles and aircraft with landing gear having wheels that roll on the ground during takeoff or landing. Background Technology

[0003] In the field of inertial navigation, the instantaneous speed of a vehicle is crucial data. Currently, speed signals are typically measured by additional speed sensors (odometers) placed on the vehicle's wheels. Installing such sensors is costly, and in some applications, particularly in aviation and military land navigation, it is not always feasible to install them.

[0004] The present invention seeks to eliminate direct speed sensors such as odometers or position sensors.

[0005] One object of the present invention is to develop a method, apparatus, and computer program for estimating the speed of a vehicle having wheels for rolling on the ground. The method is capable of accurately estimating the instantaneous speed of a wheeled vehicle based on at least one inertial or vibration sensor (which may be, for example, an accelerometer, gyroscope, or microphone) by eliminating direct speed sensors such as odometers or position sensors.

[0006] In existing technologies, such as the 2017 paper "The Future of Automotive Localization Algorithms: Available, reliable, and scalable localization: Anywhere and anytime" by Rickard Karlsson and Fredrik Gustafsson, attempts have been made to overcome the lack of direct velocity sensors by utilizing information present in inertial signals. According to this paper, instantaneous velocity is estimated from the inertial signal by selecting frequencies with higher energy in the spectrum at each moment. This type of method is effective in some simple applications with low noise, but not in the aforementioned application areas where the noise-to-signal ratio is typically very high. The method described by Rickard Karlsson and Fredrik Gustafsson in this paper is not robust to environmental noise such as engine combustion and requires a spectrum where peaks are easily identifiable, which is not possible in the presence of an internal combustion engine. Furthermore, the spectrum of inertial or vibration signals is rich in harmonics, which undergo considerable energy leakage, making it difficult to identify peaks via Fourier transform. The vast majority of existing inertial navigation systems cannot be used with predominantly ambient noise.

[0007] In some applications, particularly in aviation and military land vehicles where the number of sensors is very limited, installing position sensors is extremely difficult, even impossible from an aeronautical perspective. The presence of instantaneous velocity signals is crucial for navigation systems to estimate the vehicle's position with high accuracy. Therefore, it is necessary to measure this instantaneous velocity without adding hardware to the vehicle's wheels. Summary of the Invention

[0008] The present invention aims to provide a method, apparatus, and computer program for estimating the speed of a vehicle having wheels that roll on the ground, thereby overcoming the aforementioned disadvantages and achieving the aforementioned objective.

[0009] Therefore, a first objective of the present invention is a method for estimating the speed of a vehicle having at least one wheel for rolling on the ground, the method comprising: The inertial or vibration signal is measured by at least one inertial or vibration sensor of the vehicle; At least one calculator calculates multiple spectra of the signal over multiple consecutive time windows defined relative to consecutive moments, the multiple spectra forming at least one spectrogram, the at least one spectrogram being a function of the consecutive moments; At least one noise attenuation and / or deterministic feature extraction filter is used to filter the spectrogram to obtain at least one filtered spectrogram of the signal at the consecutive time points. The calculator estimates the frequency trajectory representing the speed of the wheel of the vehicle in the filtered spectrum at consecutive time points, including: The calculator estimates the observation probability of the frequency trajectory from the calculated amplitude at each moment of the filtered spectrogram; The calculator estimates the posterior law of the observed frequency trajectory, which is proportional to the product of the prescribed transition probability of the frequency trajectory and the observed probability of the frequency trajectory; and The calculator estimates the frequency trajectory at consecutive moments from the posterior law of the observation of the frequency trajectory; and The calculator estimates the speed of the vehicle's wheel at any given moment based on the previously estimated frequency trajectory.

[0010] Therefore, this invention proposes a method for estimating inertial or vibration signals that is robust to noise and utilizes frequency information present in the measured (inertial or vibration) signal, along with a method based on specific probabilistic approaches. Unlike existing technologies, this invention enables the extraction of vehicle speed signals from environmental noise without the use of an odometer, allowing for the utilization of data in real-world conditions. This is expressed as a function of amplitude and transfer laws at each point on the time-frequency plane, making it possible to improve the continuity, direction, and smoothness of the trajectory when utilizing the system's kinematics (such as harmonics). The inertial signal is always present in the presence of a vehicle navigation system. In the case of vibration sensors, the vibration frequency emitted by the vehicle's wheels is proportional to the vehicle's speed (simply multiplying the vehicle's speed by the wheel circumference).

[0011] According to one embodiment of the invention, the calculator is also provided to: determine the support of the frequency values; and at each time step, the calculator estimates the frequency trajectory as the frequency value in the support, until at that time step, the frequency value maximizes the posterior law of the observation of the frequency trajectory.

[0012] According to one embodiment of the invention, the calculator is also provided to: determine the support of frequency values; and the calculator to estimate the frequency at each moment as the centroid of the posterior law of the observation at that frequency, the centroid being obtained at the frequency value of the support and influenced by those frequency values.

[0013] According to one embodiment of the invention, the calculator is also provided to determine the support level of frequency values. ; and the probability of observation is estimated by a calculator when the known trajectory depends on the calculated amplitude of the spectrum, which is filtered and normalized for each time step.

[0014] According to one embodiment of the invention, the calculator is also provided to estimate the observation probability. It is based on the following equation:

[0015] in This is the spectrum after filtering at each time t. The calculated amplitude, It is the lowest frequency trajectory measured. It is the highest frequency trajectory measured. and , Is when Between and The value is 1 when it is between 1 and 1. Below or higher A function that takes the value 0.

[0016] According to one embodiment of the invention, the calculator is also provided to estimate the observation probability. It is based on the following equation:

[0017] in It is a positive integer or a real number that satisfies , in Represents the spectrum of each filter. At that frequency at time t, in This is used as the filter spectrum diagram. The threshold frequency is determined in advance as a function of the function.

[0018] According to one embodiment of the invention, the calculator determines the support for frequency values. The calculator estimates the frequency trajectory. The probability of this observation It is based on the following equation:

[0019] for , The calculated amplitude depends on the filtered spectrum at each time t.

[0020] in, It is the observed binary vector, and its size is... This is the moment. The regulations make hour ,otherwise hour 0.

[0021] According to one embodiment of the present invention,

[0022] in, This is a spectrum of the filtered data at each time t. The calculated amplitude,

[0023] It is the lowest frequency trajectory measured. It is the highest frequency trajectory measured. and satisfy , Is when Between and The value is 1 when it is between 1 and 1. Below or higher A function that takes the value 0.

[0024] According to one embodiment of the invention, the calculator also provides: the support level for determining frequency values. ; The time is specified by the calculator The frequency trajectory The transition probability The transition probability is derived from the given initial probability at time 1. In the beginning, And each frequency The equation for the transition probability is as follows:

[0025] in It is a specified integer number of consecutive specified times t that are greater than or equal to 2. in It is the frequency of the trajectory at the current time t. The frequency of the trajectory up to the next time t+1 The transition probability.

[0026] According to one embodiment of the present invention, the frequency of the frequency trajectory at the current time t The frequency of the trajectory at the next time t+1. The transition probability Following the frequency of the filtered spectrum at time t Gauss's law with the center as the condition.

[0027] According to one embodiment of the present invention, at this frequency The condition for Gauss's law to be centered has a standard deviation. ,in, This is the computational time resolution of the spectrum at time t. It is the frequency trajectory The calculated average velocity that evolves between time t and the next time t+1.

[0028] According to one embodiment of the invention, it is also provided that: the calculator multiplies the frequency trajectory by... To estimate the speed of the vehicle's wheel, where R is the predetermined or estimated radius of the wheel.

[0029] According to one embodiment of the invention, the noise attenuation and / or deterministic feature extraction filter includes a sliding median filter and / or a deterministic feature separation filter relative to random features.

[0030] The second subject of this application is a device for estimating the speed of a vehicle having at least one wheel for rolling on the ground, the device comprising: The vehicle's inertial or vibration sensors are used to measure inertial or vibration signals; A calculator configured to calculate multiple spectra of the signal over multiple consecutive time windows defined relative to consecutive moments, the multiple spectra forming at least one spectrogram, the at least one spectrogram being a function of the consecutive moments; A filter is used to attenuate noise and / or extract deterministic features from the spectrum to obtain at least one filtered spectrum of the signal at consecutive time points. The calculator is configured to estimate the frequency trajectory representing the vehicle's wheel speed in the filtered spectrum at consecutive time points, including: The calculator estimates the observation probability of the frequency trajectory based on the calculated amplitude at each moment of the filtered spectrum. The calculator estimates the posterior law of the observed frequency trajectory, which is proportional to the product of the prescribed transition probability of the frequency trajectory and the observed probability of the frequency trajectory; and The calculator estimates the frequency trajectory at consecutive moments based on the posterior law of the observed frequency trajectory; and The calculator is configured to estimate the speed of the vehicle's wheel at consecutive moments based on the frequency trajectory.

[0031] The third subject of the invention is a computer program, comprising, when executed in a calculator, code instructions for implementing the method described above for estimating the speed of a vehicle. Attached Figure Description

[0032] The invention will be better understood by reading the following description, which is given only by reference to the following figures and is a non-limiting example.

[0033] Figure 1 An example flowchart illustrating a process for estimating the speed of a vehicle according to an embodiment of the present invention is provided, the vehicle having wheels for rolling on the ground.

[0034] Figure 2 An example of the calculation steps of a spectrum of a signal for a method, apparatus, and procedure for estimating the speed of a vehicle having wheels for rolling on the ground is illustrated schematically according to an embodiment of the present invention.

[0035] Figure 3 An example of a spectrum of a time signal in the time-frequency plane is shown for a method, apparatus, and procedure for estimating the speed of a vehicle having wheels for rolling on the ground, according to an embodiment of the present invention.

[0036] Figure 4 An example of a device for estimating the speed of a vehicle having wheels for rolling on the ground is illustrated schematically according to an embodiment of the invention.

[0037] Figure 5 An example of a raw signal that can be measured by a method, apparatus, and procedure for estimating the speed of a vehicle having wheels for rolling on the ground is shown according to an embodiment of the invention.

[0038] Figure 6 This illustrates that, according to embodiments of the present invention, the speed of a vehicle can be calculated by a method, apparatus, and program for estimating vehicle speed. Figure 5 The spectrum of the signal, the vehicle has wheels for rolling on the ground.

[0039] Figure 7 This illustrates that, according to embodiments of the present invention, the speed of a vehicle can be calculated by a method, apparatus, and program for estimating vehicle speed. Figure 5 An example of a filtered spectrum of a signal, the vehicle having wheels for rolling on the ground.

[0040] Figure 8 This illustrates that, according to embodiments of the present invention, a method, apparatus, and procedure for estimating the speed of a vehicle can be used. Figure 5 An example of the observed probability of the frequency trajectory calculated from the signal, the vehicle having wheels for rolling on the ground.

[0041] Figure 9 This illustrates that, according to embodiments of the present invention, a method, apparatus, and procedure for estimating the speed of a vehicle can be used. Figure 5 An example of the frequency trajectory transition probability calculated from the signal, the vehicle having wheels for rolling on the ground.

[0042] Figure 10 This illustrates that, according to embodiments of the present invention, a method, apparatus, and procedure for estimating the speed of a vehicle can be used. Figure 5 An example of the frequency trajectory calculated from the signal, the vehicle has wheels for rolling on the ground.

[0043] Figure 11 An example of a device for estimating the speed of a vehicle having wheels for rolling on the ground is illustrated schematically according to an embodiment of the invention. Detailed Implementation

[0044] refer to Figures 1 to 10 The following describes in more detail embodiments and exemplary models of the method, apparatus 1, and computer program for estimating the instantaneous travel speed v of a vehicle VE, the vehicle VE having wheels RO for rolling on the ground.

[0045] In the aforementioned application areas of this invention, when using vibration signals from complex mechanical systems such as land vehicles or aviation systems, these signals are heavily influenced by noise (ambient noise, combustion noise, and aerodynamic noise, etc.), especially by sources of interference and asynchronous rotation. Therefore, under low signal-to-noise ratio conditions, identifying the correct trajectory representing the vehicle's speed v in the time-frequency plane is particularly difficult: only a few peaks related to the instantaneous frequency can be distinguished, and spurious peaks caused by noise can locally substitute for peaks. This first difficulty can be addressed by adding step E4, which preprocesses the time-frequency representation through noise filtering to improve the appearance of wheel characteristics relative to background noise. The proposed denoising method E4 not only attenuates background noise but also attenuates quasi-stationary components in the time-frequency plane when the sought component is strongly non-stationary (rising or falling); conversely, it attenuates non-stationary components when the sought component is quasi-stationary.

[0046] Then, the multi-component nature of the signal and the possibility of asynchronous components in the engine (such as in an aircraft engine) mean that several harmonics can coexist, intersect, be very close, or even be confused in the same frequency band. Therefore, simply following the maximum value at each moment may deviate from the correct trajectory and thus provide an incorrect estimate of the velocity v. To overcome this problem, in addition to the amplitude of the spectrogram, the present invention provides additional constraints on the optimal trajectory during the estimation process to obtain an estimate of the transition probability. Considering the directions of several harmonics and the trajectory of interest makes it possible, for example, to give more weight to the trajectory closer to the sought instantaneous frequency, which is advantageous for the coupling torque between different components. Adding these regularity constraints by estimating the transition probability also makes it possible to avoid jumps and discontinuities in the presence of isolated peaks associated with residual noise after preprocessing step E4.

[0047] The goal is to search for the optimal frequency trajectory among all possible trajectories in the spectrum over time. The optimal frequency trajectory satisfies the considered constraints and represents the speed v of the wheel RO of the vehicle VE. The optimality of one trajectory relative to another is measured by a cost function or, equivalently, by probability density. For this purpose, the information of the spectrum and constraints is represented in probabilistic form. More specifically, it can be obtained by using a transition density... and given initial probability Markov chains for frequency trajectories or Modeling. The transfer density is defined as a function of additional constraints (continuity, harmonics, direction, etc.). At each time step... The probability depends on the amplitude of the spectrum. This provides an observation of the spectrum. The frequency trajectory of the vehicle's speed v (VE) can be visualized. Choosing to make the posterior law The frequency trajectory that is maximized (or by calculating its average value). The uncertainty of this estimate can also be calculated by focusing on the second moment. The user can input the number of harmonics to be considered on the physical data input interface INT1.

[0048] In the accompanying drawings, a method for estimating the speed v of a vehicle having one (or more) wheels RO for rolling on the ground includes a first step E1: measuring the time signal X of inertia or vibration from one (or more) inertial or vibration sensors CAP of the vehicle. The inertial or vibration sensor CAP may be located outside the vehicle's engine. The inertial or vibration sensor CAP may be present on the wheels RO of the vehicle VE for rolling on the ground. Due to extreme environments, sensors often malfunction or exhibit measurement errors. The sensor CAP may be or include one (or more) accelerometers and / or one (or more) gyroscopes and / or one (or more) acoustic sensors and / or one (or more) microphones. The invention can use inertial sensors not located in the engine. Furthermore, in the case where the vehicle VE includes an inertial unit consisting of six inertial sensors (three accelerometers and three gyroscopes), the invention can utilize the redundancy of the sensor CAP formed by some or all of these six inertial sensors to improve the accuracy of the speed information v (the risk of simultaneous measurement errors on all six sensors is lower). Figure 5 An example is shown where the original signal X is a function of time on the horizontal axis.

[0049] During the second step E2 following the first step E1, one (or more) calculators CAL operate within several consecutive (or sliding) time windows W. t The above calculates several spectra SP of the time signal X. t These time windows W t It is defined relative to continuous time t, for example, Figure 2 As shown. Each time window W t It can be centered at time t and have a predetermined finite time width relative to time t (e.g., a continuous time window W). t (They can have the same time width). The calculator CAL calculates several consecutive partial-time signals Y (or time intervals Y), which are respectively equal to the signal X multiplied by the continuous time window W of the consecutive time t. t For example, consecutive times t belong to the set ,in This is a predefined number greater than 1. Time t can be a multiple of the same sampling step size to the nearest specified constant, and is represented below by an integer representing the sampling step size to the nearest specified constant. The sampling step size and / or sampling frequency can be entered by the user on the physical data input interface INT1, or can be pre-recorded in the calculator CAL. The sampling step size and / or sampling frequency define the number of samples collected per second and time t.

[0050] A calculator CAL can be or include one or more computers, one or more servers, one or more machines, one or more processors, or one or more microprocessors. The calculator CAL can include one or more physical data input interfaces INT1 and one or more physical data output interfaces INT2. This or these physical data input interfaces INT1 can be or include one or more computer keyboards, one or more physical data communication ports, one or more touchscreens, etc. This or these physical data output interfaces INT2 can be or include one or more physical data communication ports, one or more screens, etc. A device 1 for estimating the speed v of a vehicle having one or more wheels RO for rolling on the ground specifically includes a sensor CAP, a calculator CAL, and a filter FIL, such as... Figure 11 The example is shown below. The filter FIL can be included in the calculator CAL or separate from the calculator CAL. The sensor CAP is configured to perform step E1 mentioned above. The filter FIL is configured to perform step E4 mentioned below. The calculator CAL is configured to perform steps E2, E3, E5, E6 (and possibly E4) and sub-steps E51, E52, E53 mentioned below. The calculator CAL and / or filter FIL and / or sensor CAP can be programmed by a computer program PR, which includes code instructions to implement a method for estimating the vehicle's travel speed v, i.e., performing steps E1 and / or E2 and / or E3 and / or E4 and / or E5 and / or E6 and sub-steps E51 and / or E52 and / or E53. The calculator CAL includes a non-volatile memory MEM (or a non-volatile recording medium MEM) in which the computer program PR is permanently recorded. The non-volatile memory MEM or non-volatile recording medium MEM is readable by a computer and includes instructions recorded thereon that, when executed by a processor, cause the processor to perform steps E1 and / or E2 and / or E3 and / or E4 and / or E5 and / or E6, as well as sub-steps E51 and / or E52 and / or E53. The signals X and data used by the method, device 1, and program PR mentioned above and below are recorded in the memory MEM.

[0051] The device 1 for estimating the travel speed v of a vehicle VE with wheels RO can be embedded in the vehicle VE. The device 1 for estimating the travel speed v can be an inertial navigation aid and can be integrated into an inertial navigation device included in an inertial unit present on the vehicle. The inertial unit can include three accelerometers in three non-parallel directions, three gyroscopes around the three non-parallel directions, an inertial navigation device, and navigation functions.

[0052] According to one embodiment of the present invention, during the second step E2, the calculator CAL calculates the spectrum SP of each partial time signal Y, which is equal to the Fourier transform (e.g., a fast Fourier transform, FFT) of that partial time signal Y. t The Fourier transform is limited to the corresponding time window W. t (Short-time Fourier transform, also known as STFT). Therefore, the calculator CAL calculates the sliding window W of the time signal X. t The Fourier transform on the time interval is used to calculate multiple spectra SP at multiple consecutive time points t. t Each spectrum SP t Give the spectral amplitude as a function of frequency f.

[0053] The resulting spectral sequence SP t To provide the measured signal X, each sensor CAP generates a spectrum SP at consecutive time t. The spectrum SP is calculated by the calculator CAL. Each spectrum SP... t One column of the spectrum SP is formed, such as Figure 3 As shown in the example. Therefore, if a large number of sensors CAP provide multiple signals X, the calculator CAL will correspondingly calculate multiple spectrograms SP and multiple spectrograms. In several spectrograms SP and In this case, these spectrograms SP and The same sensor can undergo different processing operations.

[0054] Representations in the time-frequency plane, such as those from, for example, Figure 5 The signal Figure 3 and Figure 6 The representation is shown in the form of an image: On the horizontal axis (time): continuous time points t; On the vertical axis: the corresponding spectrum SP at each time t. t frequency f t ; In the grayscale and logarithmic scale on the vertical time t axis: the corresponding spectrum SP at each time t. t The absolute value of the amplitude, which is related to the grayscale scale ECH.

[0055] Figure 3 As an example, the frequency spectrum SP of the inertial signal X obtained by the inertial sensor CAP, formed by the lateral accelerometer recorded on the vehicle VE, is shown in the time-frequency plane. t . Figures 5 to 10The example recorded is obtained from the inertial sensor CAP on a car VE. The response of a mechanical system depends heavily on its operating state, so analysis of the signal X, adjusted at velocity v at each moment, is advantageous. The time-frequency representation makes it possible to reveal characteristics associated with the mechanical source. This representation provides an idea of ​​how the frequency content evolves over time. Other time-frequency representations, such as the Wigner representation, can also be used.

[0056] In a vehicle VE with wheels RO, the wheels RO can emit vibrations that pass through a series of peaks (frequency trajectories). This is manifested in the time-frequency plane. Sources related to the wheel exhibit a continuous, smooth trajectory (frequency trajectory) of velocity v without sudden jumps. Fundamental frequency (frequency trajectory) The energy of the component X and its harmonics (other frequency components COMPF) is maximized in the time-frequency plane. The frequency of the wheel RO velocity v of vehicle VE is represented in the continuous spectrum by a smooth trajectory that varies according to the system state (acceleration, deceleration, stopping, etc.). Therefore, a simple method for estimating the instantaneous frequency of signal X is equivalent to estimating the frequency of the component to be followed in the frequency band (frequency trajectory). The location of the maximum energy peak is found within the range. Considering the basic multiharmonics of the wheel, as well as the addition of continuity constraints and directional information, it becomes possible to attenuate the continuous trajectory related to disturbance components (such as combustion of the vehicle engine).

[0057] During the fourth step E4 following the second step E2, the spectrum SP is processed by one (or more) noise attenuation and / or deterministic feature extraction filters FIL. t Filtering is performed to obtain the filtered spectrum of signal X at continuous time t. In the spectrum In this context, the frequency at time t is represented by f. t This preprocessing step E4 enables a clearer time-frequency representation. In step E4, it becomes possible to distinguish between two cases where the target component (velocity v) is quasi-stationary or highly non-stationary. In the case of a quasi-stationary target component (velocity v), the filter FIL can include a separation filter (DRS: deterministic-random separation) of deterministic (and therefore predictable) characteristics relative to random (and therefore unpredictable) characteristics, including background noise and non-stationary disturbance components. In the case of a highly non-stationary target component (velocity v), the filter FIL can be configured to reduce background noise and non-time-varying frequencies (e.g., ventilation noise, electrical components), and a sliding median filter can be included for this purpose. Figure 5 Spectrum diagram of the filtered signal Examples such as Figure 7 As shown. Filter spectrum. The maximum frequency variation (to ensure signal continuity) can be determined by the calculator CAL, or pre-recorded in the calculator CAL, or specified by the user on the physical data input interface INT1.

[0058] During the fifth step, E5, following the fourth step E4, the frequency trajectory is estimated by the calculator CAL. This frequency trajectory represents the filtered spectrum at consecutive time t. The instantaneous velocity v of wheel RO of vehicle VE. Step E5 includes sub-steps E51, E52, and E53, which will be described below. In the following text, the frequency trajectory... Also represented as .

[0059] In fact, the speed v of the wheels RO of vehicle VE generates a spectrum. A frequency component in the time-frequency plane, which may have a frequency that varies with time (equal to the frequency trajectory). (The following content) and a certain amplitude that changes over time. The frequency in the spectrum It changes with time t, therefore in the spectrum In (and therefore in) Figure 3 In the time-frequency plane, the frequency trajectory of the vehicle VE's wheel RO velocity v as a function of time t is formed. The frequency trajectory With spectrum The trajectory of other frequencies of COMPF is mixed with other frequency components, such as Figure 3 As shown. According to one embodiment, for a wheel RO rolling on the ground, the instantaneous rotational frequency (revolutions per second) of the wheel RO at consecutive times t is the frequency trajectory as a function of time t. According to another embodiment, for a wheel RO rolling on the ground, the tangential velocity of the wheel RO at consecutive time t (revolutions per second multiplied by the radius of the wheel RO and multiplied by...) () is the frequency trajectory as a function of time t. .

[0060] set up For the spectrum ( Figure 2 The number of time intervals or consecutive moments t (greater than or equal to 2) on the horizontal axis (in the diagram), that is, the number of moments. .set up Filter spectrum diagram Number of intermediate frequency transmitter boxes (≥2). Spectrum diagram for each time t. The frequency is expressed as f t The true unknown frequency trajectory still needs to be estimated. Represented as The number of consecutive time intervals t It can be determined by the calculator CAL, or it may have been pre-recorded in the calculator CAL, or it may be specified by the user on the physical data input interface INT1.

[0061] According to one embodiment of the present invention, the frequency value is determined by the calculator CAL during the second step E2. support ,Right now .value support It is a spectrum diagram The support of frequencies, and can include all frequency values ​​contained between zero and the maximum frequency, such as uniformly distributed frequencies, which can be multiples of a specified frequency step size to the nearest specified constant. Support The number of frequency values ​​L s It is finite and can be determined by the calculator CAL; or pre-recorded in the calculator CAL; or specified by the user on the physical data input interface INT1. The frequency step size can be determined by the calculator CAL based on the sampling frequency and / or based on the window W. t It is calculated based on the duration.

[0062] The frequency trajectory to be estimated Represented as Therefore, for consecutive time intervals t from 1 to Ns, the frequency trajectory... Equal to a set of frequencies In other words, It is a simulation of the trajectory to be estimated. A random variable. Frequency trajectory. Can be found in the spectrum diagram frequency support downtake frequency value .

[0063] Frequency Trajectory Continuous frequency chain with time t Transition probability can be used and given initial probability To describe.

[0064] According to one embodiment, during the third step E3 after the second step E2 and before the fourth step E4, from a given initial probability... Initially, the calculator CAL determines the frequency of the trajectory from the current time t. The frequency of this trajectory up to the next time t+1 The frequency trajectory is defined by a Markov chain linked by the transition probabilities. transition probability This transition probability It follows a priori law. In the following text, frequency trajectories... transition probability Represented as .from Figure 5 The transition probability calculated from the signal Examples such as Figure 9 As shown, the transition probabilities describing the possible evolution of velocity over time are specified in advance. Specifically, larger frequency changes have lower probabilities, while smaller changes have higher probabilities. These transition probabilities define what is known as a Markov chain.

[0065] According to one embodiment of the present invention, for and for each frequency (i.e., each) During step E3, the calculator CAL calculates the given initial probability from time 1 according to the following equation. The beginning frequency trajectory transition probability :

[0066] It is the initial probability given the time t=1. In this equation, It is the frequency of the trajectory at the current time t. Frequency of the trajectory up to the next time step t+1 The transition probability.

[0067] Therefore, frequency trajectory The joint law is given by the following formula: (A priori law (2)) According to one embodiment of the present invention, during step E3, the calculator CAL calculates the frequency spectrum at time t based on the filtered spectrum at time t. ) frequency Using the conditional Gaussian law N centered at the center, calculate the frequency of the trajectory from the current time t. Frequency of the trajectory up to the next time step t+1 transition probability ,Right now: ,in, It is the variance of a conditional Gaussian distribution N, and the frequency is... It is the average value of the conditional Gaussian distribution N. This allows us to maintain a certain regularity of the harmonics to be followed and adds a continuity constraint between consecutive times t and t+1.

[0068] According to one embodiment of the present invention, the calculator CAL calculates the standard deviation of conditional Gaussian law N during step E3. , it is ,in, It is a filter spectrum diagram ( The computational time resolution, It is a frequency trajectory The average velocity is calculated to evolve between time t and the next time t+1. Therefore, by reducing the variance... The smoothing effect will be stronger when a frequency trajectory is provided at each time t. When the direction of evolution (e.g., rising, falling, stable), the transition law is calculated by the calculator CAL. This direction can also be considered. For example, the calculator CAL can use a transition distribution. The transfer distribution is located on the frequency trajectory. Cut off when rising And in frequency trajectory Cut off in the case of descent In frequency trajectory When things become constant (stop falling and rising), for example at zero (the stopping of vehicle VE corresponds to zero velocity v), it is necessary to... Set to a fairly low value.

[0069] In the first sub-step E51 of step E5, the calculator CAL uses the filtered spectrum for each time t. The frequency trajectory of the vehicle's speed v is estimated by calculating the amplitude. Observation probability In the following text, the frequency trajectory of the vehicle VE's speed v is described. Observation probability Represented as .from Figure 5 Observation probability calculated from the signal Examples such as Figure 8 As shown.

[0070] According to one embodiment of the invention, the calculator CAL, during sub-step E51, depends on the filtered and normalized spectrogram for each time t based on the known trajectory. Calculate the observation probability when calculating the amplitude. .

[0071] According to one embodiment of the present invention, the calculator CAL calculates the following equation during sub-step E51 for: and for each frequency (i.e., each) Frequency Trajectory Observation probability :

[0072] The calculator CAL calculates the time during substep E51. (and for each frequency) That is, each ) Specified size The observed binary vector or And make the observed binary vector coordinates Each of the binary vectors in the image (and therefore the observed binary vector) coordinates for )equal: if ,but ; Otherwise if or ,but 0.

[0073] In the aforementioned section, It is observation.

[0074] Observation density or probability Written as follows: (Observation Law (1)) Observation probability It is based on the spectrum diagram ,in , amplitude Defined. In fact, it is possible to define the target trajectory. There is a concept, and then based on that point ( Other points ( , The amplitude levels of the compared spectrograms are used to confirm whether it is highly likely to be observed. .

[0075] The main idea of ​​the method used is that, given each moment... Target frequency The probability density is calculated based on the spectrum describing the observed distribution. In fact, for a given... ,frequency coefficient At any moment It is very likely that it equals 1, which means the true frequency. It is very likely equal to .

[0076] According to one embodiment of the present invention, the calculator CAL calculates the frequency trajectory according to the following equation during sub-step E51. Observation probability :

[0077] for , The filtered spectrum depends on each time step (t). The calculated amplitude of ) ,in .

[0078] According to one embodiment of the present invention, the calculator CAL calculates the observation probability during sub-step E51 according to the following equation. :

[0079] in, This is the spectrum after filtering at each time t. The calculated amplitude, It is the lowest frequency trajectory measured. It is the highest frequency trajectory measured. and satisfy , Is when Between and The value is 1 when it is between 1 and 1. Below or higher A function that takes the value 0.

[0080] According to one embodiment of the present invention, the calculator CAL performs the following calculations during sub-step E51.

[0081] in This is a spectrum of the filtered data at each time t. The calculated amplitude.

[0082] The calculator CAL calculates the divisor coefficient during substep E51. It is located in the minimum frequency trajectory and maximum frequency trajectory The maximum value of the frequency f between ,Right now:

[0083] The calculator CAL measures or determines the minimum frequency trajectory during substep E51. The calculator CAL measures or determines the maximum frequency trajectory during substep E51. The frequency trajectory of vehicle VE at its speed v. Included in the minimum frequency trajectory and maximum frequency trajectory Between. Therefore. .

[0084] The calculator CAL will use the function during substep E51. Calculated as between and The single-slot function between ) is, that is, for Between and Between these values, the function takes the value 1. Below or higher The function takes the value 0.

[0085] Minimum frequency trajectory and maximum frequency trajectory This allows us to consider only the frequency trajectory being followed. The peak values ​​of nearby components are identified, and the frequency trajectory is defined. The frequency band in which it is located becomes possible.

[0086] According to one embodiment of the present invention, the minimum frequency trajectory It is fixed over time t, that is to say .

[0087] According to one embodiment of the present invention, the maximum frequency trajectory It is fixed over time t, that is to say .

[0088] According to another embodiment of the invention, the minimum frequency trajectory It changes with time t.

[0089] According to another embodiment of the invention, the maximum frequency trajectory It changes with time t.

[0090] In the target frequency trajectory When the frequency band is spread across a wide frequency band (e.g., a ramp) where other interfering harmonics may coexist, it is preferable to choose a frequency band that varies with time t. This is equivalent to defining a frequency trajectory. Two trajectories and For example, the selected frequency band contains only one complete continuous trajectory over the entire period of the time t under consideration. According to the transfer law The introduced continuity constraint is given by the frequency trajectory The peaks linked to the smooth curve have a greater weighting effect. According to one embodiment of the invention, the trajectory... and It can be automatically calculated by the calculator CAL. According to one embodiment of the present invention, the user can randomly select a spectrum. Spectrum diagram surrounding the peak sequence (maximum amplitude) to be followed The points are then interpolated to specify the trajectory for the CAL calculator. and The definition of the frequency band is optional, but it makes it possible to significantly improve the performance of the method and minimize computational costs, especially when the observed signal contains a large number of harmonics. Stop intervals can also be identified and incorporated into the calculation. and The definition of , and the subsequent transfer law. If this information is unavailable, from the spectrogram A simple way to estimate the stopping time is to use the spectrum. At all frequencies Every moment on The sum of the amplitudes is compared with a given threshold. Then at time Stop, among which This is the threshold to be set (e.g., the median of all values ​​across the entire timeline). Minimum frequency trajectory. and / or maximum frequency trajectory Data can be input by the user on the physical data input interface INT1, or it can be pre-recorded in the calculator CAL. In the case where device 1 is integrated into the navigation device of the vehicle VE, prior possible data, such as... and The data can be entered by the user on the physical data input interface INT1, or it can be pre-recorded in the calculator CAL for real-time navigation.

[0091] Given the target frequency At that time, the calculator CAL calculates each moment. Observation probability .

[0092] According to one embodiment of the present invention, in the case of a single-component signal having an intermediate frequency (IF), if the spectrum of the noise-free signal is... If available, then In other words, only frequency can be observed. Spectrum .otherwise, For the spectrum The amplitude at time All frequencies that are the maximum value Both are important. Therefore, in The moment in When several peaks appear (usually related to noise), the function It can be multimodal. That is, the same observed spectrum can correspond to solutions at several possible frequencies. By considering the spectrum of several times the target harmonics. The amplitude of the amplitude may be biased towards one mode rather than another. According to one embodiment of the invention, the calculator CAL calculates the observation probability during sub-step E51 based on the following equation. (In this case, the probability is written as a product of several probabilities):

[0093] in, It can be any positive integer or real number, satisfying , in, It is based on the spectrum diagram And the predetermined threshold frequency.

[0094] In the second sub-step E52 of step E5, the frequency trajectory is estimated by the calculator CAL. or frequency posterior laws of observation The posterior law of this observation With frequency trajectory transition probability and frequency trajectory Observation probability The product is directly proportional, that is to say This frequency trajectory posterior laws of observation Also expressed as or .

[0095] According to one embodiment of the present invention, in the second sub-step E52 of the fifth step E5, the calculator CAL calculates the posterior law of the observation. .

[0096]

[0097] According to one embodiment of the present invention, in the second sub-step E52 of the fifth step E5, the calculator CAL calculates the posterior law of the observation. .

[0098]

[0099] According to one embodiment of the present invention, in the second sub-step E52 of the fifth step E5, the calculator CAL calculates the posterior law of the observation. The posterior law of this observation is equal to the frequency locus. transition probability Frequency Trajectory Observation probability and the multiplication factor calculated by the calculator CAL The product of, that is .

[0100] According to one embodiment of the present invention, the multiplication factor It can take the value 1.

[0101] According to one embodiment of the present invention, in the second sub-step E52 of the fifth step E5, the calculator CAL calculates the multiplication factor. equal:

[0102] According to one embodiment of the present invention, in the second sub-step E52 of the fifth step E5, the calculator CAL calculates the posterior law of the observation. equal:

[0103] According to one embodiment of the present invention, in the second sub-step E52 of the fifth step E5, the calculator CAL calculates the posterior law of the observation. .

[0104]

[0105] In the third sub-step E53 of step E5, the calculator CAL calculates the frequency trajectory. posterior laws of observation Estimating the frequency trajectory at continuous time t The calculator CAL uses the observed probabilities... The observations given At that time, from the frequency trajectory posterior laws of observation Calculate the optimal trajectory among all possible trajectories in the spectrogram. As frequency trajectory .from Figure 5The frequency trajectory calculated from the signal Examples such as Figure 10 As shown. The frequency trajectory can be provided by the calculator CAL on the physical data output interface INT2 (e.g., by displaying it on the screen). or and / or frequency trajectories or One or more frequency values ​​(which may be equal to the instantaneous frequency of wheel RO). The calculated confidence interval, already calculated by the calculator CAL, can be provided by the calculator on the physical data output interface INT2 (e.g., by display on a screen). In the case where device 1 is integrated into the navigation device of the vehicle VE, the frequency trajectory... or and / or Frequency trajectory or One or more frequency values ​​can be stored in memory M or another memory and evaluated in real time in another filter used for real-time navigation.

[0106] According to an embodiment of the present invention, in the third sub-step E53 of the fifth step E5, the calculator CAL records the frequency trajectory at each time t. Estimated as support frequency values ​​in One of the support features is the frequency trajectory of the vehicle VE's speed v at time t. posterior laws of observation Maximize. In this case, the calculator CAL will focus on frequency values. support The following selects the frequency trajectory for each instant t. Therefore, in this case, the frequency trajectory equal The calculator includes maximum a posteriori (MAP). Maximize the estimator. This optimization problem can be easily solved, for example, by using dynamic programming.

[0107] According to an embodiment of the present invention, in the third sub-step E53 of the fifth step E5, the calculator CAL observes the probability. With transition probability Convolution with Gaussian kernels (e.g., two-dimensional) is used to compute the posterior law of observation. .

[0108] According to an embodiment of the present invention, in the third sub-step E53 of the fifth step E5, the calculator CAL recursively calculates the posterior law according to the following equation. :

[0109] Where * represents the convolution operator, It represents density. The Gaussian kernel. These operations are not expensive.

[0110] Then, estimate the frequency trajectory at each time t. Or as a posterior law frequency of center of gravity They are respectively based on support frequency value The values ​​were obtained from the location and received from these frequency values ​​respectively. The influence of this. The resulting frequency. In this case, it's support. frequency value A linear combination of these, whose weights are determined by the posterior law. Formation, respectively from support frequency value The frequency trajectory of the vehicle VE's speed v is obtained from the location. Therefore, in this case, the frequency trajectory of the vehicle VE's speed v is... equal:

[0111] Therefore, frequency trajectory It is the posterior mean, which will produce a smoother result.

[0112] During the sixth step, E6, following the fifth step E5, the calculator CAL calculates the frequency trajectory. Estimate the speed v of wheel RO of vehicle VE at consecutive time t. The speed v of wheel RO of vehicle VE can be provided by the calculator CAL on the physical data output interface INT2 (e.g., by displaying on a screen). In the presence of a navigation device, the speed v of wheel RO of vehicle VE can be stored in memory MEM or another memory and evaluated in real time in another filter used for real-time navigation.

[0113] According to one embodiment of the present invention, in the sixth step E6, the calculator CAL measures the frequency trajectory of the vehicle VE's travel speed v. Multiply by the circumference of the wheel RO To estimate the speed v of the wheels RO of vehicle VE at time t, where R is the radius of wheel RO, and this value of R is predetermined or estimated, for example, during navigation by any method. Speed ​​v is also expressed as... And in this case, it is a tangential velocity. Therefore or The radius R and / or diameter and / or circumference of wheel RO. The data can be entered by the user on the physical data input interface INT1, or it can be pre-recorded in the calculator CAL.

[0114] According to another embodiment of the present invention, in the sixth step E6, the calculator CAL estimates the travel speed v of the wheel RO of vehicle VE at time t as the rotational frequency (revolutions per second) of the wheel RO of vehicle VE, which is equal to the frequency trajectory at time t. .

[0115] The present invention has the following particular advantages: Removing the sensors eliminates the need for odometers. This is very interesting, especially in the aerospace field where sensors are extremely expensive, not only because of their resistance to extreme environments (high pressure and high temperature), but also because of coordination issues, where even the slightest sensor can alter the modal characteristics of the structure.

[0116] By using information present in a large number of linearly correlated harmonics, it becomes possible to estimate the minute changes in the velocity v rather than a rough estimate.

[0117] It also considers the issue of frequency or rotational speed truncation (common in aircraft engines), as well as the continuity of this variation.

[0118] This method allows users to provide the speed a priori (e.g., based on...) Figure 4 (If the signal is in an ascending or quasi-stationary state), it provides the possibility of increasingly higher estimation accuracy. Although this information is not mandatory, it can significantly improve the quality of the obtained signal.

[0119] Of course, the above embodiments, features, possibilities, and examples can be combined with each other or selected independently.

Claims

1. A method for estimating the speed (v) of a vehicle (VE), said vehicle having at least one wheel (RO) for rolling on the ground, characterized in that, The method includes: (E1) Measure inertial or vibration signals using at least one inertial or vibration sensor (CAP) of the vehicle; (E2) At least one calculator (CAL) calculates the signal over multiple consecutive time windows (W) defined relative to consecutive moments (t). t Multiple spectra (SPs) on ) t The multiple spectrums (SP) t The system forms at least one spectrogram (SP), which is a function of the time (t); the calculator (CAL) determines the frequency value (t). ) support Among them, L s It is the number of the frequency values, the number L s Greater than or equal to 2; (E3) The calculator (CAL) specifies the frequency trajectory at a given time. transition probability The transition probability is derived from the given initial probability at time (1). In the beginning, And each frequency The equation for the transition probability is as follows: in It is a specified integer number of times (t) that are greater than or equal to 2. in The frequency trajectory from the current time (t). Frequency trajectory to the next time step (t+1) The transition probability; (E4) At least one noise attenuation and / or deterministic feature extraction filter (FIL) filters the at least one spectrogram (SP) to obtain at least one filtered spectrogram of the signal at the time (t). ; (E5) The calculator (CAL) estimates the at least one filtered spectrogram at said time (t). The frequency trajectory representing the travel speed (v) of the wheels (RO) of the vehicle (VE) ,include: (E51) The calculator (CAL) is based on the at least one filtered spectrogram. The frequency trajectory is estimated by calculating the amplitude at each time point (t). Observation probability ( ); The calculator (CAL) described in (E52) estimates the frequency trajectory. posterior laws of observation ( The frequency trajectory The posterior law of the observation ( ) and the frequency trajectory The specified transition probability ( ) and the frequency trajectory Observation probability ( The product of ) is proportional; and (E53) The calculator (CAL) is based on the frequency trajectory The posterior law of the observation ( Estimate the frequency trajectory at the given time (t). ;as well as (E6) The calculator (CAL) calculates the frequency trajectory based on the already estimated frequency trajectory. Estimate the speed (v) of the vehicle's wheels at time (t).

2. The method according to claim 1, characterized in that, The method further includes: (E53) The calculator (CAL) at each time (t) will record the frequency trajectory Estimated as support Frequency values ​​in The frequency value at each time point (t) The frequency trajectory The posterior law of the observation ( )maximize.

3. The method according to claim 1, characterized in that, The method further includes: (E53) The calculator (CAL) will calculate the frequency trajectory at each time step (t). The estimated frequency trajectory The posterior law of the observation ( The center of gravity of the support The frequency value ( Obtained from ( ).

4. The method according to any one of the preceding claims, characterized in that, The method further includes: (E51) The calculator (CAL) estimates the probability of the observation ( ), where the observation probability ( (This depends on the at least one filter spectrogram) The calculated amplitude after normalization at each time step (t).

5. The method according to claim 4, characterized in that, The method further includes: (E51) the calculator (CAL) estimating the observation probability. It is based on the following equation: in It is the at least one filtered spectrum diagram The calculated amplitude at each time (t), It is the lowest frequency trajectory measured. It is the trajectory of the highest frequency measured. and satisfy , Is when Between and The value is 1 when it is between 1 and 2. Below or higher A function that takes the value 0.

6. The method according to claim 4, characterized in that, The method further includes: (E51) the calculator (CAL) estimating the observation probability. It is based on the following equation: in It is a positive integer or a real number that satisfies , in Represents the at least one filter spectrum diagram The frequency of each of them at time (t), in It is as the at least one filter spectrum diagram The predetermined threshold frequency, Among them, the It is the divisor coefficient.

7. The method according to any one of claims 1 to 3, characterized in that, The frequency trajectory from the time (t) The frequency trajectory up to the next time point (t+1) The transition probability Following the at least one filtered spectrum at said time (t) The frequency Gauss's law with the center as the condition.

8. The method according to claim 7, characterized in that, At the frequency The conditional Gaussian law centered on the center has a standard deviation ,in, It is the at least one filtered spectrum at the time (t). The computation time resolution, It is the frequency trajectory The calculated average velocity that evolves between the stated time (t) and the next time (t+1).

9. The method according to any one of claims 1 to 3, characterized in that, The method further includes: (E6) The calculator (CAL) uses the frequency trajectory Multiply To estimate the travel speed (v) of the wheels of the vehicle, where R is the predetermined or estimated radius of the wheels (RO).

10. The method according to any one of claims 1 to 3, characterized in that, The noise attenuation and / or deterministic feature extraction (FIL) filter includes a sliding median filter and / or a deterministic feature separation filter relative to random features.

11. A device (1) for estimating the travel speed (v) of a vehicle (VE), said vehicle having at least one wheel (RO) for rolling on the ground, characterized in that, The device includes: The vehicle's inertial or vibration sensor (CAP) is used (E1) to measure inertial or vibration signals; A calculator (CAL) is configured (E2) to calculate the signal over multiple consecutive time windows (W) defined relative to consecutive moments (t). t Multiple spectra (SPs) on ) t The multiple spectrums (SP) t Form at least one spectrogram (SP), the at least one spectrogram (SP) being a function of the time (t); and The calculator (CAL) is configured (E2) to determine the frequency value ( ) support Among them, L s It is the number of the frequency values, the number L s Greater than or equal to 2; The calculator (CAL) is configured to track the frequency trajectory at a specified time (E3). transition probability The transition probability is derived from the given initial probability at time (1). In the beginning, And each frequency The equation for the transition probability is as follows: in It is a specified integer number of times (t) that are greater than or equal to 2. in The frequency trajectory from the current time (t). Frequency trajectory to the next time step (t+1) The transition probability; A filter (FIL) is used to attenuate noise and / or extract deterministic features from the at least one spectrogram (SP) to (E4) obtain at least one filtered spectrogram of the signal at the time (t). , The calculator (CAL) is configured (E5) to estimate the at least one filtered spectrogram at the time (t). The frequency trajectory representing the travel speed (v) of the wheels (RO) of the vehicle (VE) ,include: (E51) The calculator (CAL) is based on the at least one filtered spectrogram. The frequency trajectory is estimated by calculating the amplitude at each time point (t). Observation probability ( ); The calculator (CAL) described in (E52) estimates the frequency trajectory. posterior laws of observation ( The frequency trajectory The posterior law of the observation ( ) and the frequency trajectory The specified transition probability ( ) and the frequency trajectory Observation probability ( The product of ) is proportional; and (E53) The calculator (CAL) is based on the frequency trajectory The posterior law of the observation ( Estimate the frequency trajectory at the given time (t). ;as well as The calculator (CAL) is configured (E6) to calculate based on the frequency trajectory. Estimate the speed (v) of the vehicle's wheels at time (t).

12. A computer program product comprising, when executed in a computer (CAL), code instructions for implementing the method for estimating the driving speed (v) of the vehicle according to any one of claims 1 to 10.

Images