Road surface friction coefficient estimation method based on intelligent tire and related equipment

Abstract

The invention provides a road surface friction coefficient estimation method based on an intelligent tire and related equipment, and relates to the technical field of state estimation and control of intelligent automobiles, in particular to the field of road surface friction coefficient estimation.The method comprises the steps that a circle of rotation of the intelligent tire serves as a period, sampling is conducted according to a preset angle serving as a sampling interval, and a sampling interval is obtained; the method comprises the steps of obtaining acceleration signals, including longitudinal, lateral and vertical acceleration signals, of an intelligent tire in multiple periods, performing spatial short-time Fourier transform on the acceleration signals to obtain transformation results of the acceleration signals in three directions in an angular space dimension, and inputting the transformation results into a convolutional neural network to obtain a convolutional neural network; according to the road surface type, the corresponding confidence coefficient and the road surface type friction coefficient relation table, the first friction coefficient of the road surface is determined. The method is not easily influenced by external environmental factors, estimation can be completed without depending on vehicle speed information and large excitation, and the problem of state coupling does not exist.

Description

Technical Field

[0001] The present disclosure relates to the field of intelligent vehicle state estimation and control technology, in particular to the field of road friction coefficient estimation, and more particularly to a road friction coefficient estimation method based on intelligent tires and related equipment. Background Art

[0002] As autonomous vehicle technology continues to advance toward higher levels of automation and the boundaries of vehicle performance are gradually expanded, achieving robust vehicle safety control in various driving scenarios, especially in emergency situations, has become a major challenge. Accurate and timely estimation of the road friction coefficient is crucial in vehicle motion control. Advanced chassis control systems require this information to improve vehicle handling and stability control.

[0003] In related technologies, methods for estimating the friction coefficient include first obtaining tire inputs (such as normal load and longitudinal and lateral slip states) and tire responses (tire forces and moments), and then identifying the friction coefficient parameters in the tire model using certain methods. Because the road friction coefficient primarily affects the nonlinear region of tire forces, such methods typically require the tire to be subjected to significant longitudinal or lateral excitation to estimate the friction coefficient. However, sufficient tire excitation usually indicates that the vehicle is approaching the edge of stability. In contrast, it is more ideal and safer to obtain the friction coefficient under conditions where the tire is subjected to only small or near-zero excitation.

[0004] In addition, the calculation of the tire slip state in the input of such methods requires vehicle speed information, and the general model-based vehicle speed estimation also requires friction coefficient information. Therefore, the above-mentioned method of estimating the friction coefficient still has a state coupling problem.

[0005] In addition to the mainstream dynamic methods mentioned above, there are also road surface classification methods based on on-board cameras. Such methods do not require a large degree of tire excitation, but are easily affected by environmental factors such as rain, snow, and lighting conditions.

[0006] Based on this, the related art has the following problems in estimating the friction coefficient: it not only requires vehicle speed information and large longitudinal or lateral excitation, but also has problems of state coupling and is easily affected by environmental factors.

[0007] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute prior art known to ordinary technicians in the field. Summary of the Invention

[0008] The present disclosure provides a method and related equipment for estimating the road friction coefficient based on smart tires, which, at least to a certain extent, overcomes the problems in related technologies such as the need for vehicle speed information and large longitudinal or lateral excitation when estimating the friction coefficient, the existence of state coupling, and susceptibility to environmental factors.

[0009] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by practice of the present disclosure.

[0010] In a first aspect, embodiments of the present disclosure provide a method for estimating a road friction coefficient based on a smart tire, the method comprising: Taking one rotation of the smart tire as a cycle, sampling is performed at a preset angle as the sampling interval to obtain acceleration signals of the smart tire for multiple cycles; the acceleration signals include: longitudinal acceleration signal, lateral acceleration signal and vertical acceleration signal; Perform spatial short-time Fourier transform on the acceleration signal to obtain the transformation results of the acceleration signals in three directions in the angular space dimension; The transformation results are input into the convolutional neural network to obtain the road surface type and the corresponding confidence level; The first friction coefficient of the road surface is determined according to the road surface type, the corresponding confidence level and the road surface type friction coefficient relationship table.

[0011] In a possible embodiment, the method further includes: The first friction coefficient and the first confidence level corresponding to the first friction coefficient are processed by adaptive Kalman filtering to determine the second friction coefficient.

[0012] In one possible embodiment, the adaptive Kalman filter includes a prediction update process and a measurement update process; Processing the first friction coefficient and a first confidence level corresponding to the first friction coefficient by using an adaptive Kalman filter to determine a second friction coefficient includes: During the prediction update process, the current friction coefficient prediction value and the current error covariance prediction value are determined; During the measurement update process, the Kalman gain is determined based on the current error covariance prediction value and the first confidence level; The second friction coefficient of the current time is determined according to the first friction coefficient, the Kalman gain, and the predicted value of the friction coefficient of the current time.

[0013] In a possible embodiment, during the measurement update process, determining the Kalman gain according to the current error covariance prediction value and the first confidence level includes: Determining a measurement covariance according to the first confidence level; wherein the greater the first confidence level, the smaller the measurement covariance; The Kalman gain is determined based on the measurement covariance and the current error covariance prediction value.

[0014] In a possible embodiment, determining the error covariance prediction value includes: Obtaining an error covariance update value of a measurement update process of a previous adaptive Kalman filter; wherein the error covariance update value is determined based on a previous Kalman gain and an error covariance prediction value of a prediction update process of a previous adaptive Kalman filter; According to the error covariance update value, the current error covariance prediction value is determined.

[0015] In a possible embodiment, the method further includes: In a process of processing the first friction coefficient and the first confidence level corresponding to the first friction coefficient by using an adaptive Kalman filter, determining whether there is a valid first friction coefficient in the current cycle; If it exists, the measurement update process is performed; If it does not exist, the measurement update process is skipped, and the current friction coefficient prediction value is used as the current second friction coefficient, and the current error covariance prediction value is used as the current error covariance update value.

[0016] In a second aspect, an embodiment of the present disclosure provides a device for estimating a road friction coefficient based on a smart tire, comprising: The sampling unit is configured to sample the acceleration signals of the smart tire at a predetermined sampling interval, with one rotation of the smart tire as a cycle, to obtain acceleration signals of the smart tire for multiple cycles; the acceleration signals include longitudinal acceleration signals, lateral acceleration signals, and vertical acceleration signals; The Fourier transform unit is used to perform a spatial short-time Fourier transform on the acceleration signal to obtain the transformation results of the acceleration signals in three directions in the angular space dimension; An estimation unit, used to input the transformation results into a convolutional neural network to obtain the road surface type and the corresponding confidence level; The determining unit is used to determine a first friction coefficient of the road surface according to the road surface type, the corresponding confidence level and the road surface type friction coefficient relationship table.

[0017] In a possible embodiment, a filtering unit is further included, configured to process the first friction coefficient and the first confidence level corresponding to the first friction coefficient through adaptive Kalman filtering to determine the second friction coefficient.

[0018] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute the method in the first aspect above by executing the executable instructions.

[0019] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which implements the method in the first aspect when the computer program is executed by a processor.

[0020] In a fifth aspect, according to another aspect of the present disclosure, a computer program product or computer program is further provided, the computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform any of the above methods.

[0021] Embodiments of the present disclosure provide a method and related equipment for estimating road friction coefficients based on smart tires, relating to the field of smart vehicle state estimation and control technology, and more particularly, to the field of road friction coefficient estimation. The method comprises: taking one rotation of the smart tire as a cycle, sampling at a preset angle as a sampling interval, obtaining multiple cycles of smart tire acceleration signals, the acceleration signals comprising longitudinal acceleration signals, lateral acceleration signals, and vertical acceleration signals; performing a spatial short-time Fourier transform on the acceleration signals to obtain the transformation results of the acceleration signals in the three directions in the angular space dimension; inputting the transformation results into a convolutional neural network to obtain the road surface type and the corresponding confidence level; and determining the first road friction coefficient based on a relationship table between the road surface type, the corresponding confidence level, and the road surface type friction coefficient. The method is not easily affected by external environmental factors, can complete the estimation without relying on vehicle speed information or large excitations, and does not suffer from state coupling issues. By integrating acceleration information from three directions into the model to provide rich features, the model achieves feature complementarity between different road types. Direct sampling at preset angles adapts to the dimensional consistency requirements of the convolutional neural network input, and based on spatial short-time Fourier transform processing, more intuitive transformation results in the spatiotemporal dimensions are obtained to improve prediction accuracy.

[0022] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification, are used to explain the principles of the present disclosure. Obviously, the drawings described below are only some embodiments of the present disclosure, and those skilled in the art can derive other drawings based on these drawings without inventive effort.

[0024] Figure 1A flow chart of a method for estimating road friction coefficient based on smart tires in an embodiment of the present disclosure is shown; Figure 2 A schematic diagram showing a transformation result in an embodiment of the present disclosure; Figure 3 A flow chart showing another method for estimating road friction coefficient based on smart tires according to an embodiment of the present disclosure is provided; Figure 4 A schematic diagram showing a process for estimating a road friction coefficient in an embodiment of the present disclosure is shown. Figure 5 A flow chart showing an adaptive Kalman filtering process according to an embodiment of the present disclosure is shown; Figure 6 A schematic diagram of the structure of a convolutional neural network in an embodiment of the present disclosure is shown; Figure 7 A schematic diagram of a cross-validation embodiment of the present disclosure is shown; Figure 8 A flow chart showing an adaptive Kalman filtering process according to an embodiment of the present disclosure is shown; Figure 9 A schematic diagram showing an evaluation result in an embodiment of the present disclosure; Figure 10 A schematic diagram showing another evaluation result in an embodiment of the present disclosure; Figure 11 A schematic structural diagram of a road friction coefficient estimation device based on a smart tire in an embodiment of the present disclosure is shown; Figure 12 A schematic structural diagram of an electronic device in an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0025] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be embodied in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art. The described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0026] In addition, the accompanying drawings are merely schematic illustrations of the present disclosure and are not necessarily drawn to scale. Identical reference numerals in the figures denote identical or similar parts, and thus repetitive descriptions thereof will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities that do not necessarily correspond to physically or logically separate entities. These functional entities may be implemented in software, in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0027] This exemplary implementation is described in detail below with reference to the accompanying drawings and examples.

[0028] First, a method for estimating a road friction coefficient is provided in an embodiment of the present disclosure. The method can be executed by any electronic device with computing and processing capabilities. In the following process, the electronic device is taken as an example of a terminal device.

[0029] Figure 1 A flow chart of a method for estimating road friction coefficient based on smart tires in an embodiment of the present disclosure is shown. Figure 1 As shown, the method for estimating the road friction coefficient based on the smart tire provided in the embodiment of the present disclosure includes the following steps: S102: Taking one rotation of the smart tire as a cycle, sampling is performed according to a preset angle as a sampling interval to obtain acceleration signals of the smart tire for multiple cycles, where the acceleration signals include: a longitudinal acceleration signal, a lateral acceleration signal, and a vertical acceleration signal.

[0030] In one possible embodiment, an acceleration sensor can be installed within the smart tire. The acceleration sensor processes acceleration signals using each smart tire rotation cycle (one rotation cycle is 360°) as a processing cycle, sampling at a preset angle as a sampling interval. This generates multiple cycles of acceleration signals from the smart tire. The acceleration signals sampled by the smart tire system represent data in the angular space dimension.

[0031] The preset angle may be 1°.

[0032] S104: Performing spatial short-time Fourier transform on the acceleration signal to obtain transformation results of the acceleration signals in three directions in the angular space dimension.

[0033] In a possible embodiment, the acceleration signal sampled in one processing cycle is subjected to spatial short-time Fourier transform.

[0034] In a possible embodiment, in order to improve the accuracy of the friction coefficient output by the convolutional neural network (CNN) and to improve the prediction speed of the convolutional neural network, the input features can be better used when the convolutional neural network performs estimation. In the embodiment of the present disclosure, a spatial short-time Fourier transform is used to process the acceleration signal in the angular space dimension to obtain the transformation results of the acceleration signals in the three directions in the angular space dimension.

[0035] It should be noted that the spatial short-time Fourier transform in the embodiment of the present disclosure is based on the input of the angular space dimension, and a short-time Fourier transform (STFT) is performed in the angular space dimension.

[0036] The three directions are longitudinal, lateral and vertical in the sensor coordinate system within the tire.

[0037] The spatial short-time Fourier transform can overcome the problem that the conventional Fourier transform cannot locate the temporal changes of the frequency components. In the embodiment of the present disclosure, the input signal is an acceleration signal in the angular space dimension. Based on the spatial short-time Fourier transform, the transformation result is obtained. The transformation result can characterize the signal strength at different angular space positions and angular space frequencies.

[0038] Based on this, after obtaining the transformation result, when it is input into the convolutional neural network, it can provide the convolutional neural network with richer information and further improve the classification accuracy of the convolutional neural network.

[0039] Angular spatial frequency refers to the number of spatial variation periods contained within a unit angular range.

[0040] Among them, the parameters of the spatial short-time Fourier transform for processing the acceleration signal in the embodiment of the present disclosure can be set to multiple types, and only two are given as examples, as shown below.

[0041] The first method is to set the acceleration sampling interval to 1 degree, the window length used for the spatial short-time Fourier transform to 64, the window function used to be a Hamming window, and the overlap rate is 75% of the window length. The number of points used in the FFT is twice the window length.

[0042] The second method: The acceleration sampling interval can be set to 0.1 degrees, the window function uses a Hamming window, the window length is 512, the overlap rate is 75%, and the number of points used in the spatial short-time Fourier transform FFT is twice the window length.

[0043] Both of the above methods are acceptable. The second method produces a clearer image than the first method, and the first method can reduce computing power compared to the second method. In this solution, the first spatial short-time Fourier transform is selected to process the acceleration signal, and the transformation result is input into the convolutional neural network to complete the prediction.

[0044] Figure 2 A schematic diagram showing a transformation result in an embodiment of the present disclosure is shown in FIG. Figure 2 As shown, the coordinates are: angular position (°) and angular spatial frequency (cycle / °), taking the sampling results of three cycles as an example, a total of 1080°. Take the road type of ice as an example for illustration, Figure 2200-1400 represents the intensity, including: 200, 400, 600, 800, 1000, 1200 and 1400 respectively. The grayscale-based color gradient represents the intensity.

[0045] The horizontal axis represents the angular position, with the horizontal axis showing 200, 400, 600, 800 and 1000 respectively, and the vertical axis represents the angular spatial frequency, with the vertical axis showing 10 -2 , 10 -1 , 10 0 .

[0046] S106: Input the transformation result into a convolutional neural network to obtain the road surface type and the corresponding confidence level.

[0047] In a possible embodiment, the output of the convolutional neural network is different road surface types and confidence levels corresponding to the different road surface types.

[0048] Pavement types can include various types, such as fine asphalt, coarse asphalt, brick paving, cement, wet tile, wet basalt, snow, and ice.

[0049] The final road surface type output by the convolutional neural network in this estimation process can be determined by the maximum confidence.

[0050] S108: Determine a first friction coefficient of the road surface according to the road surface type, the corresponding confidence level, and the road surface type friction coefficient relationship table.

[0051] In one possible embodiment, the road surface type friction coefficient relationship table includes the correlation between road surface type and friction coefficient, for example: fine asphalt 1.0, coarse asphalt 1.0, brick road 0.6, cement 0.7, wet tile 0.1, wet basalt 0.4, compacted snow 0.3 and ice road 0.1.

[0052] In a possible embodiment, the final road surface type output by the convolutional neural network is determined by the maximum confidence, and the first friction coefficient is determined based on the final road surface type and the road surface type friction coefficient relationship table.

[0053] The above method in the embodiment of the present disclosure can achieve the following technical effects: (1) It is not easily affected by external environmental factors and can complete the estimation without relying on vehicle speed information and large incentives. There is no problem of state coupling. At the same time, it solves the problems existing in the two different methods in related technologies.

[0054] (2) Various machine learning methods were used in the experiment. The convolutional neural network was most suitable for the acceleration signals in the three directions collected in the embodiment of the present disclosure. The transformation results in the three directions can be input into the convolutional neural network and then feature splicing can be performed in the spatial dimension to form an input tensor, so that the model can perceive richer features and higher accuracy.

[0055] (3) Based on the use of convolutional neural networks, since it requires the input to meet the problem of size consistency, when sampling, if sampling is performed according to time, for example, sampling is performed according to 1 second or several seconds, when the vehicle's moving speed is unstable, effective sampling features cannot be obtained, and the input size consistency requirements cannot be met. The convolutional neural network cannot use the features sampled according to time. Based on this consideration, the inventors thought of sampling according to a fixed preset angle, which can obtain an acceleration signal that meets the network input requirements and adapt to the estimation process of the convolutional neural network.

[0056] (4) After directly sampling the features of the angular space dimension, the conventional Fourier transform cannot meet the specific data analysis process. Based on this consideration, the inventor uses the acceleration signal of the angular space dimension to perform a spatial short-time Fourier transform, and can obtain a transformation result, which is used to characterize the distribution intensity of the angular space frequency of the sampled acceleration signal as the angular position changes.

[0057] Figure 3 FIG. 1 shows a flow chart of another method for estimating road friction coefficient based on smart tires in an embodiment of the present disclosure. Figure 3 As shown, the method includes: S302: Taking one rotation of the smart tire as a cycle, sampling is performed according to a preset angle as a sampling interval to obtain acceleration signals of the smart tire for multiple cycles, where the acceleration signals include: a longitudinal acceleration signal, a lateral acceleration signal, and a vertical acceleration signal.

[0058] S304: Performing spatial short-time Fourier transform on the acceleration signal to obtain transformation results of the acceleration signals in three directions in the angular space dimension.

[0059] S306: Input the transformation result into a convolutional neural network to obtain the road surface type and the corresponding confidence level.

[0060] S308: Determine a first friction coefficient of the road surface and a corresponding first confidence level according to a road surface type, a corresponding confidence level, and a road surface type friction coefficient relationship table.

[0061] S310: Processing the first friction coefficient and the first confidence level corresponding to the first friction coefficient through adaptive Kalman filtering to determine a second friction coefficient.

[0062] In a possible embodiment, the first friction coefficient is further estimated by an adaptive Kalman filter (AKF). The process update of the adaptive Kalman filter can adopt a random walk model to adaptively adjust the first friction coefficient based on the first confidence level and the adaptive Kalman filter.

[0063] Figure 4 A schematic diagram of a road friction coefficient estimation process in an embodiment of the present disclosure is shown. Figure 4 As shown, the transformation result of the acceleration signal is input into the convolutional neural network, the road surface type and the corresponding first confidence level are output, the first friction coefficient is determined through the road surface type friction coefficient relationship table, and the first friction coefficient and its corresponding first confidence level are input into the adaptive Kalman filter, and the second friction coefficient is output.

[0064] In a possible embodiment, the first friction coefficient and its corresponding first confidence level are used in a measurement update process of an adaptive Kalman filter.

[0065] Adaptive Kalman filtering includes: prediction update process and measurement update process.

[0066] Figure 5 FIG. 4 shows a flow chart of an adaptive Kalman filtering process in an embodiment of the present disclosure. Figure 5 As shown, the following steps are included: S502: During the prediction update process, determine the current friction coefficient prediction value and the current error covariance prediction value.

[0067] In a possible embodiment, a method of obtaining the current friction coefficient prediction value may include: determining the current friction coefficient prediction value according to the second friction coefficient obtained in a measurement update process of the last adaptive Kalman filter.

[0068] In a possible embodiment, the method of obtaining the current error covariance prediction value may include: obtaining the error covariance update value of the measurement update process of the last adaptive Kalman filter; wherein the error covariance update value is determined based on the last Kalman gain and the error covariance prediction value in the prediction update process of the last adaptive Kalman filter; and determining the current error covariance prediction value based on the error covariance update value.

[0069] It should be noted that the current time, the previous time, etc. in the embodiment of the present disclosure refer to the number of internal loops in the adaptive Kalman filtering process.

[0070] S504: During the measurement update process, determine the Kalman gain according to the current error covariance prediction value and the first confidence level.

[0071] In a possible embodiment, the Kalman gain may be determined in the following manner: determining a measurement covariance based on a first confidence level; wherein, the greater the first confidence level, the smaller the measurement covariance; and determining the Kalman gain based on the measurement covariance and a current error covariance prediction value.

[0072] S506: Determine the current second friction coefficient according to the first friction coefficient, the Kalman gain, and the current friction coefficient prediction value.

[0073] Through the above process, the first friction coefficient and the first confidence level output by the convolutional neural network are used in the filtering process to re-estimate the friction coefficient. The convolutional neural network estimation process is combined with the adaptive Kalman filtering process to jointly estimate the friction coefficient, thereby improving the accuracy of the estimated friction coefficient. The above process is not a simple combination, where only the friction coefficient is filtered. Instead, the friction coefficient is adaptively adjusted in combination with its confidence level. The confidence level is used to determine the gain. Based on different confidence levels, different gains are dynamically determined, thereby adjusting the first friction coefficient in real time and outputting the second friction coefficient.

[0074] In one possible embodiment, the measurement update process of the adaptive Kalman filter (AKF) primarily processes the first friction coefficient and the first confidence level. The sampling results of the intelligent tire system are not generated at a fixed frequency. Instead, they are sampled based on a preset angle and depend on the tire's rotational speed. The faster the rotational speed, the more sampling results are obtained in the same timeframe. However, the sampling results are typically lower than the sampling rate of the adaptive Kalman filter (AKF). Since the AKF operates at a fixed sampling interval, the sampling results of the intelligent tire system are not always available. Therefore, an update flag is introduced to indicate whether a valid first friction coefficient exists at the current moment.

[0075] Based on this, whether a valid first friction coefficient exists can be determined between the prediction update process and the measurement update process of the adaptive Kalman filter process. This can include the following steps: during the process of processing the first friction coefficient and the first confidence level corresponding to the first friction coefficient through the adaptive Kalman filter, determining whether a valid first friction coefficient exists in the current cycle; if so, executing the measurement update process; if not, skipping the measurement update process, using the current friction coefficient prediction value as the current second friction coefficient, and using the current error covariance prediction value as the current error covariance update value.

[0076] The convolutional neural network in the embodiment of the present disclosure is a trained neural network. Figure 6 A schematic diagram of the structure of a convolutional neural network in an embodiment of the present disclosure is shown in FIG. Figure 6As shown, it includes: the first convolutional layer, the maximum pooling layer, the second convolutional layer, the first fully connected layer, the second fully connected layer and the normalization layer.

[0077] Exemplarily, the transformation results of the acceleration signals in three directions in the angular space dimension can be spliced ​​in the spatial dimension, including: the transformation result of the longitudinal acceleration signal, the transformation result of the lateral acceleration signal, and the transformation result of the vertical acceleration signal, to form an input tensor of size 65×57×1 as the input of the first convolutional layer.

[0078] The first convolutional layer may include: 16 5×5 convolution kernels with a stride of 1. It outputs the first convolution result, performs batch normalization (BN) and ReLU activation on the first convolution result, and then inputs the first processing result into a 2×2 maximum pooling layer. Figure 6 In the above figure, BN and ReLU are used to represent these two processes respectively.

[0079] The stride of the maximum pooling layer can be 2, which can reduce the spatial dimension and output the pooling result.

[0080] The second convolutional layer may include 32 3×3 convolution kernels. The input is the pooling result, and the output is the second convolution result. The second convolution result is batch normalized and ReLU activated to extract features to obtain the second processing result.

[0081] The first fully connected layer may include 128 neurons, the second processed result is input, and the first fully connected result is output. ReLU activation is performed on the first fully connected result to obtain the third processed result.

[0082] The second fully connected layer can include 8 neurons. Its input is the result of the third processing. The second fully connected layer can map the features to 8 output categories, and then perform classification through the normalization (softmax) layer to obtain the corresponding road surface type and confidence.

[0083] During the training process, the stochastic gradient descent with momentum (SGDM) optimizer can be used, the learning rate is set to 0.005, and the mini-batch size is 128.

[0084] In one possible embodiment, part of the collected data is used for training the convolutional neural network, and another part is reserved for subsequent estimation algorithm verification, etc., which can better reflect the generalization ability of the model on unfamiliar data.

[0085] Samples from vehicles on different road types under conventional uniform speed driving conditions and double lane change (DLC) conditions can be collected and divided according to the proportion. The entire dataset is divided into: training set, validation set and test set for training the CNN model.

[0086] Table 1 shows a schematic diagram of a confusion matrix when testing a test set in an embodiment of the present disclosure. As shown in Table 1, it includes multiple road surface types, taking the following 8 as examples, namely: fine asphalt, coarse asphalt, paving bricks, cement, wet ceramic tiles, wet basalt, snow, and ice.

[0087] The specific data information is shown in Table 1.

[0088] Table 1

[0089] The horizontal axis in Table 1 represents the different road surface types output by the convolutional neural network, i.e., the predicted categories. The vertical axis represents the actual road surface types during sampling, i.e., the true categories. The data in the table represents quantities.

[0090] For example, for cement, the number of true cement categories in the vertical direction is 82, which means that the total number of samples on the cement road surface is 82. The number of predicted cement categories in the horizontal direction is 82, which means that the final road surface type output by the convolutional neural network is 82.

[0091] Table 2 shows the accuracy and negative prediction value respectively. The specific data information is shown in Table 2, and the unit is %.

[0092] Table 2

[0093] For example, in the horizontal direction, for the predicted category of fine asphalt, there were 255 outputs of the predicted category of fine asphalt, 251 of which were true as fine asphalt, and 4 of which were true as coarse asphalt. Using a convolutional neural network to estimate the final road surface type, the probability of a positive prediction, or accuracy, was 98.4%, and the probability of a negative prediction, or negative prediction value, was 1.6%.

[0094] Table 3 shows the recall rate and specificity. The specific data information is shown in Table 3, and the unit is %.

[0095] Table 3

[0096] For example, in the longitudinal direction, for an item predicted as fine asphalt, the true category was fine asphalt 252 times, the predicted category was also fine asphalt 251 times, and the total number of predicted categories was other than asphalt 1. Using a convolutional neural network to estimate the final road surface type, the probability of a positive prediction, or recall, was 99.6%, and the probability of a negative prediction, or specificity, was 0.4%.

[0097] Among them, Table 2 and Table 3 both output their corresponding probability values ​​according to the horizontal and vertical sorting of Table 1.

[0098] It can be seen from the above table that after training, the convolutional neural network in the embodiment of the present disclosure has a high accuracy in estimation results. Without relying on vehicle motion state information (vehicle speed information) and large excitation, the road surface type can be accurately obtained and the friction coefficient can be determined.

[0099] In order to evaluate the robustness of the convolutional neural network, 5-fold cross-validation (CV) was used to evaluate the entire dataset.

[0100] Figure 7 A schematic diagram of a cross-validation in an embodiment of the present disclosure is shown. Figure 7 As shown in FIG, the cross-validation results of estimating the longitudinal acceleration signal, lateral acceleration signal and vertical acceleration signal separately, and estimating the acceleration signals in the three directions together.

[0101] The ordinate represents the estimated accuracy, which includes: 0.88, 0.9, 0.92, 0.94, 0.96, 0.98 and 1. The abscissa represents the acceleration signals used, which include: longitudinal acceleration signal, lateral acceleration signal and vertical acceleration signal, as well as acceleration signals in three directions.

[0102] based on Figure 7 The range of accuracy when using different acceleration signals shows that the accuracy is highest when using acceleration signals in three directions for estimation. This is because the three-axis information is fused in the convolutional neural network, providing richer features, enabling the model to achieve feature complementarity between different road types and improve the accuracy of the estimation.

[0103] In a possible embodiment, the processing method of the prediction update process and the measurement update process of the adaptive Kalman filter process may include the following steps, as shown in the following formula.

[0104] The k-th prediction update process includes: (1) (2) in, represents the k-th predicted friction coefficient value, and indicates that it is calculated based on the second friction coefficient of the k-1th time. represents the k-th error covariance prediction value, and indicates that it is calculated based on the k-1-th error covariance update value; Represents the k-th error covariance update value; represents the k-1th second friction coefficient, Q represents the process noise covariance, which can be set to 0.001, and A represents the identity matrix.

[0105] The initial second friction coefficient may be set to an extreme value far from the friction coefficient corresponding to the actual road surface type, which is not specifically limited in the embodiments of the present disclosure.

[0106] The k-th measurement update process includes: (3) (4) (5) in, represents the k-th Kalman gain. represents the k-th second friction coefficient and indicates that it is calculated based on the k-th friction coefficient prediction value, the k-th Kalman gain, and the k-th first friction coefficient. represents the k-th error covariance update value, and indicates that it is calculated based on the k-th error covariance prediction value and the k-th Kalman gain, represents the measurement covariance, represents the kth first friction coefficient, and H and I both represent the unit matrix.

[0107] The measurement covariance can be determined based on formula (6) as shown below.

[0108] (6) in, Indicates the first confidence level corresponding to the first friction coefficient of the kth time.

[0109] When the first confidence level approaches 1, that is, when the confidence level is high, the measurement covariance approaches 0, and the adaptive Kalman filter will assign a larger Kalman gain during the measurement update process. When the first confidence level approaches 0, that is, when the confidence level is low, the first friction coefficient obtained is less reliable, the measurement covariance approaches infinity, and the obtained Kalman gain is very small. The first friction coefficient is actually ignored.

[0110] Figure 8FIG. 4 shows a flow chart of an adaptive Kalman filtering process in an embodiment of the present disclosure. Figure 8 As shown, it includes: prediction update process, judgment process and measurement update process.

[0111] S802: Execute the prediction update process.

[0112] S804: Determine whether upflg is 1. If so, execute S806; if not, execute S808.

[0113] Here, upflg is the abbreviation of the update flag, which is used to determine whether there is a valid first friction coefficient at the current moment.

[0114] S806: Execute the measurement update process.

[0115] S808 uses the current friction coefficient prediction value as the current second friction coefficient, and uses the current error covariance prediction value as the current error covariance update value, and returns to execute S802.

[0116] By using the above method in the embodiment of the present disclosure, when a vehicle is traveling on a rough asphalt road at a constant speed, Figure 9 A schematic diagram showing an evaluation result in an embodiment of the present disclosure is shown in FIG. Figure 9 As shown, there are four coordinate systems, among which the ordinates of three coordinate systems are friction coefficients (μ), namely: true friction coefficient, first friction coefficient and second friction coefficient; the ordinate of another coordinate system is the confidence level corresponding to the first friction coefficient; and the abscissas of the four coordinate systems are time (s).

[0117] The coordinate system shows the friction coefficient from 0 to 1.2, 0, 0.2, 0.4, 0.6, 0.8, 1 and 1.2, the confidence level from 0.5 to 1, and the time includes: 0, 2, 4, 6, 8, 10, 12, 14, 16 and 18 values ​​from 0 to 18s. The time dimension of the horizontal axis in the four coordinate systems is aligned.

[0118] The actual friction coefficient of coarse asphalt is 1. To ensure rigorous evaluation, the initial state of the AKF (the initial secondary friction coefficient) was set to 0.1, significantly lower than the actual friction coefficient. During normal, constant-speed driving, the convolutional neural network's direct output of the first friction coefficient misjudged the first friction coefficient. However, after filtering and adjusting the output of the second friction coefficient through the AKF filter, it quickly converged to the correct value upon initial availability.

[0119] Furthermore, the AKF can adaptively adjust the measurement covariance based on the probability output by the CNN, thereby adjusting the output value of the second friction coefficient. When the time is between 10s and 12s, due to the low confidence level (close to 0.5), directly outputting the second friction coefficient value would cause the AKF to misjudge. However, the AKF effectively reduces the Kalman gain during this measurement update process, thereby maintaining the stability of the estimation result.

[0120] When the vehicle is on the transition road between snow and ice, Figure 10 A schematic diagram showing another evaluation result in an embodiment of the present disclosure is shown. Figure 10 As shown, there are four coordinate systems. The ordinates of three coordinate systems represent the friction coefficient (μ), namely the true friction coefficient, the first friction coefficient, and the second friction coefficient. The ordinate of another coordinate system represents the confidence level corresponding to the first friction coefficient. The abscissas of all four coordinate systems represent time (s). The vehicle moves from snow to ice at 4.7 seconds.

[0121] The coordinate system shows the friction coefficient from 0.2 to 1 as 0.2, 0.4, 0.6, 0.8 and 1, the confidence level from 0.5 to 1, and the time includes the horizontal coordinate corresponding to the friction coefficient: 1, 2, 3, 4, 5, 6, 7, 8 and 9 values ​​from 0 to 9s, and the horizontal coordinate corresponding to the confidence level: 0, 2, 4, 6, 8 and 10 values ​​from 0 to 10s.

[0122] The initial state of AKF is set to 1.0. During the overall estimation process, the convolutional neural network has some individual misjudgments, but the confidence corresponding to these first friction coefficients is low. Therefore, AKF can effectively reduce the Kalman gain of these unreliable first friction coefficients during the measurement update process, making the estimation results stable. Figure 10 It can be seen that when the road surface changes, further using AKF to reprocess the first friction coefficient output by the convolutional neural network can make the estimation result converge quickly to the correct friction coefficient, thereby improving the accuracy of friction coefficient estimation.

[0123] in, Figure 9 and Figure 10 In the diagram, the first friction coefficient is represented by a rhombus, the second friction coefficient is represented by a dotted line, and the actual friction coefficient is represented by a solid line. The rhombuses are connected by lines to make the diagram more clear in terms of how the first friction coefficient changes over time.

[0124] Based on the same inventive concept, the present disclosure also provides a device for estimating the road friction coefficient based on a smart tire, as shown in the following embodiment. Because the principles of this device embodiment are similar to those of the aforementioned method embodiment, the implementation of this device embodiment can be referenced to the implementation of the aforementioned method embodiment, and any repetitions will not be repeated.

[0125] Figure 11 A schematic diagram of the structure of a road friction coefficient estimation device based on a smart tire in an embodiment of the present disclosure is shown. Figure 11 As shown, the road friction coefficient estimation device 110 based on the smart tire includes: a sampling unit 1101, which is used to take one rotation of the smart tire as a cycle and perform sampling according to a preset angle as a sampling interval to obtain the acceleration signal of the smart tire for multiple cycles; the acceleration signal includes: a longitudinal acceleration signal, a lateral acceleration signal and a vertical acceleration signal; a Fourier transform unit 1102, which is used to perform a spatial short-time Fourier transform on the acceleration signal to obtain the transformation results of the acceleration signals in three directions in the angular space dimension; an estimation unit 1103, which is used to input the transformation result into a convolutional neural network to obtain the road surface type and the corresponding confidence level; a determination unit 1104, which is used to determine the first friction coefficient of the road surface according to the relationship table of the road surface type, the corresponding confidence level and the road surface type friction coefficient.

[0126] Those skilled in the art will appreciate that various aspects of the present disclosure may be implemented as systems, methods, or program products. Therefore, various aspects of the present disclosure may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, microcode, etc.), or in a combination of hardware and software, collectively referred to herein as "circuits," "modules," or "systems."

[0127] Refer to the following Figure 12 12 is a diagram to describe the electronic device 1200 according to this embodiment of the present disclosure. Figure 12 The electronic device 1200 shown is merely an example and should not limit the functions and scope of use of the embodiments of the present disclosure.

[0128] like Figure 12 As shown, electronic device 1200 is implemented as a general-purpose computing device. Components of electronic device 1200 may include, but are not limited to, the aforementioned at least one processing unit 1210, the aforementioned at least one storage unit 1220, and a bus 1230 connecting various system components (including storage unit 1220 and processing unit 1210).

[0129] The storage unit stores program code, which can be executed by the processing unit 1210, so that the processing unit 1210 performs the steps described in the "Exemplary Method" section above according to various exemplary embodiments of the present disclosure. For example, the processing unit 1210 can perform the steps of any of the above method embodiments.

[0130] The storage unit 1220 may include a readable medium in the form of a volatile storage unit, such as a random access memory unit (RAM) 12201 and / or a cache memory unit 12202 , and may further include a read-only memory unit (ROM) 12203 .

[0131] The storage unit 1220 may also include a program / utility 12204 having a set (at least one) of program modules 12205, such program modules 12205 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0132] The bus 1230 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0133] Electronic device 1200 can also communicate with one or more external devices 1240 (e.g., a keyboard, pointing device, Bluetooth device, etc.), one or more devices that enable a user to interact with electronic device 1200, and / or any device that enables electronic device 1200 to communicate with one or more other computing devices (e.g., a router, modem, etc.). This communication can occur via input / output (I / O) interface 1250. Furthermore, electronic device 1200 can communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network such as the Internet) via network adapter 1260. As shown, network adapter 1260 communicates with other modules of electronic device 1200 via bus 1230. It should be understood that, although not shown in the figure, other hardware and / or software modules can be used in conjunction with electronic device 1200, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0134] Through the description of the above embodiments, it will be readily understood by those skilled in the art that the example embodiments described herein can be implemented via software or via a combination of software and necessary hardware. Therefore, the technical solutions according to the embodiments of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, or mobile hard drive) or on a network and includes several instructions for enabling a computing device (such as a personal computer, server, terminal device, or network device) to execute the methods according to the embodiments of the present disclosure.

[0135] In particular, according to embodiments of the present disclosure, the processes described above with reference to the flowcharts may be implemented as a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the methods of the above embodiments.

[0136] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided. The computer-readable storage medium may be a readable signal medium or a readable storage medium. A program product capable of implementing the above-mentioned method of the present disclosure is stored thereon. In some possible implementations, various aspects of the present disclosure may also be implemented in the form of a program product, which includes program code. When the program product is run on a terminal device, the program code is used to cause the terminal device to execute the steps according to various exemplary embodiments of the present disclosure described in the "Exemplary Methods" section above of this specification.

[0137] More specific examples of computer-readable storage media in the present disclosure may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), optical fibers, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0138] In the present disclosure, a computer-readable storage medium may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable signal medium may also be any readable medium other than a readable storage medium that can transmit, propagate, or transfer a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0139] Alternatively, the program code contained on the computer-readable storage medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination thereof.

[0140] In particular embodiments, the program code utilized by the program code instructions can be implemented in any of various ways. For example, it can be implemented in various programming languages, including compiled or interpreted languages, and can be implemented using scripting languages such as VBScript, JavaScript, Perl, Python, etc. In some embodiments, different programming languages can be employed in various combinations. In particular embodiments, the program code instructions utilized by the program code can be executed by one or more programmable processors electonically. The input can be supplied to the electronic processor(s) via user input, from sensors, and / or from other electronically stored data. The output can be provided on a display, to a storage device, and / or to other output devices. In particular embodiments, the program code instructions utilized by the program code can be implemented in any of various ways. For example, it can be implemented in various programming languages, including compiled or interpreted languages, and can be implemented using scripting languages such as VBScript, JavaScript, Perl, Python, etc. In some embodiments, different programming languages can be employed in various combinations. In particular embodiments, the program code instructions utilized by the program code can be executed by one or more programmable processors electonically. The input can be supplied to the electronic processor(s) via user input, from sensors, and / or from other electronically stored data. The output can be provided on a display, to a storage device, and / or to other output devices.

[0141] It should be noted that while the foregoing detailed description has set forth a number of specific embodiments of the devices for action performance, this division into modules or units is not mandatory. Indeed, according to embodiments of the present disclosure, the features and functionalities of two or more of the modules or units described above can be embodied in a single module or unit. Conversely, the features and functionalities of one module or unit described above can be further divided into multiple modules or units embodied by multiple modules or units.

[0142] Furthermore, although individual steps of the methods in the present disclosure are described in a particular order in the drawings, this is not required or implied as to the order in which the steps must be performed, nor is it required that all of the steps shown be performed to achieve the desired result. Additionally or alternatively, certain steps can be omitted, multiple steps can be combined into a single step, a single step can be broken up into multiple steps, and / or the like.

[0143] From the above description of embodiments, those skilled in the art will readily perceive that the example embodiments described herein can be practiced by various other than the methods specifically described above. It should be noted that the example embodiments described above can be implemented in software and / or in combinations of software and necessary hardware. The embodiments implemented in software can be implemented in any programming language such as C, C++, Java, etc., and can be implemented in any operating system such as Microsoft Windows®, Linux, etc. The software implementation can be for or part of one or more computer programs, which can execute on a computer. This computer can be a personal computer, a server, a mobile device, a network device, etc.

[0144] Other embodiments of the present disclosure will readily occur to those skilled in the art after considering the specification and practicing the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art not disclosed herein. The description and examples are to be considered as exemplary only, with the true scope and spirit of the present disclosure being indicated by the appended claims.

Claims

1. A method for estimating road friction coefficient based on intelligent tires, characterized in that: The method comprises: Taking one rotation of the smart tire as a cycle, sampling is performed at a preset angle as a sampling interval to obtain acceleration signals of the smart tire for multiple cycles; the acceleration signals include: longitudinal acceleration signals, lateral acceleration signals, and vertical acceleration signals; Performing a spatial short-time Fourier transform on the acceleration signal to obtain transformation results of the acceleration signals in three directions in the angular space dimension; Inputting the transformation result into a convolutional neural network to obtain the road surface type and the corresponding confidence level; A first friction coefficient of the road surface is determined according to the road surface type, the corresponding confidence level and the road surface type friction coefficient relationship table.

2. The method according to claim 1, characterized in that The method further comprises: The first friction coefficient and a first confidence level corresponding to the first friction coefficient are processed by adaptive Kalman filtering to determine a second friction coefficient.

3. The method according to claim 2, characterized in that The adaptive Kalman filter includes a prediction update process and a measurement update process; The step of processing the first friction coefficient and the first confidence level corresponding to the first friction coefficient by using an adaptive Kalman filter to determine the second friction coefficient includes: During the prediction update process, determining a current friction coefficient prediction value and a current error covariance prediction value; During the measurement update process, determining a Kalman gain according to the current error covariance prediction value and the first confidence level; The second friction coefficient for the current time is determined according to the first friction coefficient, the Kalman gain, and the predicted value of the friction coefficient for the current time.

4. The method according to claim 3, characterized in that The step of determining the Kalman gain according to the current error covariance prediction value and the first confidence level during the measurement update process includes: Determining a measurement covariance according to the first confidence level; wherein the greater the first confidence level, the smaller the measurement covariance; The Kalman gain is determined according to the measurement covariance and the current error covariance prediction value.

5. The method according to claim 3, characterized in that Determine the error covariance forecast, including: Obtaining an error covariance update value of a measurement update process of a previous adaptive Kalman filter; wherein the error covariance update value is determined based on a previous Kalman gain and an error covariance prediction value of a prediction update process of a previous adaptive Kalman filter; The current error covariance prediction value is determined according to the error covariance update value.

6. The method according to claim 3, characterized in that The method further comprises: In a process of processing the first friction coefficient and the first confidence level corresponding to the first friction coefficient by using an adaptive Kalman filter, determining whether there is a valid first friction coefficient in the current cycle; If so, executing the measurement update process; If it does not exist, the measurement update process is skipped, and the friction coefficient prediction value of the current time is used as the second friction coefficient of the current time, and the error covariance prediction value of the current time is used as the error covariance update value of the current time.

7. A road friction coefficient estimation device based on intelligent tires, characterized in that: include: a sampling unit, configured to take one rotation of the smart tire as a cycle and perform sampling at a preset angle as a sampling interval, to obtain acceleration signals of the smart tire for multiple cycles; The acceleration signal includes: a longitudinal acceleration signal, a lateral acceleration signal and a vertical acceleration signal; A Fourier transform unit, configured to perform a spatial short-time Fourier transform on the acceleration signal to obtain transformation results of the acceleration signals in three directions in the angular space dimension; an estimation unit, configured to input the transformation result into a convolutional neural network to obtain a road surface type and a corresponding confidence level; The determining unit is used to determine a first friction coefficient of the road surface according to the road surface type, the corresponding confidence level and the road surface type friction coefficient relationship table.

8. An electronic device, characterized in that: include: processor; and a memory for storing executable instructions for the processor; The processor is configured to perform the method according to any one of claims 1 to 6 by executing the executable instructions.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

10. A computer program product comprising: A computer program or instruction, characterized in that when the computer program or instruction is executed by a processor, it implements the method according to any one of claims 1 to 6.

Images