Method, device and vehicle for estimating the adhesion coefficient of a road surface

Abstract

The application relates to a road surface adhesion coefficient estimation method, device and vehicle, wherein the estimation method comprises the following steps: obtaining a target data sequence of the vehicle; wherein the target data sequence comprises data points at different time points, and each data point comprises an adhesion coefficient estimation value and a tire slip rate; determining a slope of the adhesion coefficient estimation value relative to the tire slip rate according to the target data sequence, and obtaining a target slope; and determining a current peak adhesion coefficient of a current driving road surface of the vehicle according to the target slope, so as to realize accurate estimation of the peak adhesion coefficient of the road surface.

Description

Technical Field

[0001] This application relates to the field of vehicle technology, and in particular to a method, apparatus and vehicle for estimating the coefficient of adhesion of a road surface. Background Technology

[0002] Peak road surface adhesion coefficient (PGF) is a key parameter that determines a vehicle's braking distance, steering stability, and acceleration performance. Accurate PGF estimation can effectively improve vehicle driving safety and handling stability under extreme conditions, and has significant engineering implications for intelligent vehicle chassis control and autonomous driving safety decisions.

[0003] Existing methods for estimating road surface adhesion coefficient typically employ specialized sensors such as optical, laser, and ultrasonic sensors to scan road surface texture, material composition, and water film status to identify adhesion characteristics. However, these sensors are highly susceptible to interference from environmental factors such as road stains, water accumulation, and dust, making it difficult for this method to accurately estimate the peak adhesion coefficient.

[0004] There is currently no effective solution to the problem that related technologies cannot accurately estimate the peak adhesion coefficient of road surfaces. Summary of the Invention

[0005] This embodiment provides a method, apparatus, and vehicle for estimating the adhesion coefficient of a road surface, aiming to achieve accurate estimation of the peak adhesion coefficient of the road surface.

[0006] Firstly, this embodiment provides a method for estimating the adhesion coefficient of a road surface, the estimation method comprising:

[0007] Obtain a target data sequence for the vehicle; wherein the target data sequence includes data points at different times, and the data points include an estimated coefficient of adhesion and a tire slip ratio;

[0008] Based on the target data sequence, the slope of the estimated adhesion coefficient relative to the tire slip ratio is determined to obtain the target slope;

[0009] The current peak adhesion coefficient of the road surface on which the vehicle is currently traveling is determined based on the target slope.

[0010] Secondly, this embodiment provides a device for estimating the adhesion coefficient of a road surface, the device comprising:

[0011] An acquisition module is used to acquire a target data sequence of the vehicle; wherein, the target data sequence includes data points at different times, and the data points include an estimated value of the adhesion coefficient and a tire slip ratio;

[0012] The analysis module is used to determine the slope of the estimated adhesion coefficient relative to the tire slip ratio based on the target data sequence, and obtain the target slope;

[0013] The processing module is used to determine the current peak adhesion coefficient of the road surface on which the vehicle is currently traveling, based on the target slope.

[0014] Thirdly, this embodiment provides a vehicle including: a control device, the control device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for estimating the road surface adhesion coefficient described in the first aspect above.

[0015] Compared with related technologies, this embodiment provides a method, apparatus, and vehicle for estimating the coefficient of adhesion (COP) of a road surface. The estimation method includes: acquiring a target data sequence of the vehicle; wherein the target data sequence includes data points at different times, and the data points include an estimated COP and a tire slip ratio; determining the slope of the COP relative to the tire slip ratio based on the target data sequence to obtain a target slope; and determining the current peak COP of the road surface on which the vehicle is currently traveling based on the target slope. Thus, by acquiring a data sequence containing the COP and tire slip ratio, accurately estimating the target slope of the COP relative to the tire slip ratio based on the data sequence, and then determining the peak COP of the road surface on which the vehicle is currently traveling based on the target slope, interference from environmental factors can be avoided, and stable and reliable estimation performance can be maintained even in complex road surface environments, thereby achieving accurate estimation of the peak COP of the road surface.

[0016] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is a flowchart of an embodiment of the method for estimating the adhesion coefficient of the road surface provided in this application;

[0019] Figure 2 This is a flowchart of Embodiment 2 of the method for estimating the adhesion coefficient of the road surface provided in this application;

[0020] Figure 3 This is a flowchart of Embodiment 3 of the method for estimating the adhesion coefficient of the road surface provided in this application;

[0021] Figure 4 This is a flowchart of Embodiment 4 of the method for estimating the adhesion coefficient of the road surface provided in this application;

[0022] Figure 5 This is a flowchart of Embodiment 5 of the method for estimating the adhesion coefficient of the road surface provided in this application;

[0023] Figure 6 This is a flowchart of Embodiment Six of the method for estimating the adhesion coefficient of the road surface provided in this application;

[0024] Figure 7 This is a flowchart of Embodiment Seven of the method for estimating the adhesion coefficient of the road surface provided in this application;

[0025] Figure 8 This is a structural block diagram of the device for estimating the adhesion coefficient of the road surface provided in this application. Detailed Implementation

[0026] To better understand the purpose, technical solution, and advantages of this application, the application is described and illustrated below in conjunction with the accompanying drawings and embodiments.

[0027] Unless otherwise defined, the technical or scientific terms used in this application shall have the general meaning understood by one of ordinary skill in the art to which this application pertains. Words such as “a,” “an,” “an,” “the,” “the,” and “these” used in this application do not indicate quantitative limitation and may be singular or plural. The terms “comprising,” “including,” “having,” and any variations thereof used in this application are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or modules (units) is not limited to the listed steps or modules (units) but may include steps or modules (units) not listed, or may include other steps or modules (units) inherent to these processes, methods, products, or devices. Words such as “connected,” “linked,” and “coupled” used in this application are not limited to physical or mechanical connections but may include electrical connections, whether direct or indirect. “Multiple” used in this application refers to two or more. “And / or” describes the relationship between related objects, indicating that three relationships may exist; for example, “A and / or B” can represent: A alone, A and B simultaneously, and B alone. Normally, the character " / " indicates that the objects before and after it are in an "or" relationship. The terms "first," "second," "third," etc., used in this application are merely to distinguish similar objects and do not represent a specific order of objects.

[0028] The following specific embodiments are given to illustrate the technical solution of this application in detail.

[0029] Figure 1 This is a flowchart of an embodiment of the method for estimating the adhesion coefficient of the road surface provided in this application. Please refer to... Figure 1 The method provided in this embodiment may include the following steps:

[0030] Step S101: Obtain the target data sequence of the vehicle; wherein, the target data sequence includes data points at different times, and the data points include the estimated value of the adhesion coefficient and the tire slip ratio.

[0031] It should be noted that the target data sequence is time-series data continuously collected and stored at a preset sampling frequency (e.g., 100Hz) during vehicle operation. Each data point at any given time contains two key parameters: the estimated coefficient of adhesion and the tire slip ratio.

[0032] The coefficient of adhesion can be estimated in real time using the vehicle's existing wheel speed sensors, inertial measurement unit (IMU), and driving force estimation module. Specifically, the coefficient of adhesion estimate... , The driving force at the current moment (which can be estimated from the output torque of the motor or engine). This represents the normal load at the current moment.

[0033] Tire slip ratio can be calculated based on the wheel speed collected by wheel speed sensors and the vehicle's reference speed. Specifically, tire slip ratio... , The wheel angular velocity is measured by the wheel speed sensor, and R is the wheel rolling radius. This is the longitudinal reference speed for the vehicle.

[0034] Understandably, the acquisition duration of the target data sequence can be adaptively adjusted according to actual working conditions to ensure a sufficient number of data points to guarantee the statistical reliability of subsequent slope estimation. Optionally, outlier data points can be removed during the acquisition process to ensure the validity of the data sequence.

[0035] Step S102: Based on the target data sequence, determine the slope of the estimated adhesion coefficient relative to the tire slip ratio to obtain the target slope.

[0036] For example, in one embodiment, a linear fit is performed on the data points in the target data sequence, and the slope output by the fit is determined as the target slope.

[0037] For example, the Random Sample Consensus (RANSAC) algorithm can be used to fit the data points in the target data sequence, solve for the slope of the adhesion coefficient estimate relative to the tire slip ratio, and obtain the target slope. In this way, through the random sampling and voting mechanism of RANSAC, interference from outliers such as sensor impulse noise and signal jumps can be automatically eliminated, effectively improving the robustness of the slope estimation.

[0038] It should be noted that the above method for determining the target slope is merely an illustrative example, and this embodiment does not impose any specific limitations on it. Those skilled in the art can flexibly choose methods such as linear regression and RANSAC to determine the target slope according to actual application needs.

[0039] Optionally, to further improve the robustness of the estimation, the target data sequence can be preprocessed, for example, by applying mean filtering or low-pass filtering to the slip ratio and adhesion coefficient to eliminate the influence of high-frequency noise on the slope estimation.

[0040] Step S103: Determine the current peak adhesion coefficient of the road surface on which the vehicle is currently traveling based on the target slope.

[0041] It should be noted that there is a physical correlation between the target slope and the peak coefficient of adhesion (COP). This embodiment determines the COP of the current driving surface based on this correlation and the estimated slope value. Specifically, the target slope represents the change in the COP estimate when the tire slip ratio changes by a unit, characterizing the sensitivity of the COP estimate to changes in tire slip ratio. In the COP estimate-tire slip ratio curve, the slope is positive in the low slip ratio range, then gradually decreases, approaching 0 near the peak COP.

[0042] In practice, for example, the target slope is substituted into the preset tire-road friction model, and the road surface state parameters in the model are solved by combining the current working condition parameters (e.g., vehicle speed, load, etc.). The peak adhesion coefficient is then determined based on these state parameters.

[0043] It should be noted that the current peak adhesion coefficient can be output to the Advanced Driving Assistance System (ADAS) or the chassis electronic control unit to drive real-time correction of safety parameters. Specifically, the current peak adhesion coefficient can serve as the basis for setting the target slip ratio and optimizing the intervention timing in the Anti-lock Braking System (ABS) and Electronic Stability Control (ESC). When the peak adhesion coefficient is low (e.g., on a wet or slippery surface), ESC can intervene earlier to suppress excessive slippage of the drive wheels. At the same time, the peak adhesion coefficient can also be used for dynamic adjustment of the following distance in the Adaptive Cruise Control (ACC) system, and adaptive correction of the Time-to-Collision (TTC) threshold in the Automatic Emergency Braking (AEB) system. In one example, when the estimated current peak adhesion coefficient is 0.3 (corresponding to a wet and slippery road surface), the TTC threshold of the AEB system can be dynamically corrected from 2 seconds on a dry road surface to 6 seconds, which means that the braking warning and actuator action are triggered 3 times earlier. This allows for more sufficient reaction and braking distance for the driver and the system, effectively reducing the risk of rear-end collisions on wet and slippery roads.

[0044] Thus, by directly using the peak adhesion coefficient estimated in this embodiment as the input for ADAS safety parameter correction, a closed-loop linkage from perceiving road conditions to active safety control is achieved, effectively improving the active safety performance of the vehicle under extreme or low adhesion conditions.

[0045] The estimation method provided in this embodiment specifically includes: acquiring a target data sequence of the vehicle; wherein the target data sequence includes data points at different times, and the data points include the estimated value of the adhesion coefficient and the tire slip ratio; determining the slope of the estimated value of the adhesion coefficient relative to the tire slip ratio based on the target data sequence to obtain the target slope; and determining the current peak adhesion coefficient of the vehicle on the current driving surface based on the target slope. In this way, by acquiring a data sequence containing the estimated value of the adhesion coefficient and the tire slip ratio, accurately estimating the target slope of the adhesion coefficient relative to the tire slip ratio based on the data sequence, and then determining the peak adhesion coefficient of the current driving surface based on the target slope, it is possible to avoid interference from environmental factors and maintain stable and reliable estimation performance even in complex road conditions, thereby achieving accurate estimation of the peak adhesion coefficient of the road surface without the need for additional dedicated hardware equipment, significantly reducing vehicle costs.

[0046] Figure 2This is a flowchart of Embodiment 2 of the method for estimating the adhesion coefficient of the road surface provided in this application. Please refer to... Figure 2 Based on the above embodiments, the current peak adhesion coefficient of the road surface on which the vehicle is currently traveling is determined according to the target slope, including:

[0047] Step S201: Determine the target mapping coefficient corresponding to the target slope based on the target slope and the preset correspondence between the slope and the mapping coefficient; wherein, the preset correspondence between the slope and the mapping coefficient is used to record the correspondence between the slope of the adhesion coefficient and the tire slip ratio and the mapping coefficient, and the mapping coefficient is the proportional coefficient between the slope and the peak adhesion coefficient corresponding to the slope.

[0048] In this step, the target mapping coefficient corresponding to the target slope is found from the preset correspondence between slope and mapping coefficient. The mapping coefficient is defined as the proportionality coefficient between the slope and the peak adhesion coefficient, i.e. , Here, k is the peak adhesion coefficient, and k is the slope. These are the mapping coefficients.

[0049] To facilitate understanding of the physical meaning of the mapping coefficients mentioned above, an exemplary tire-road friction model will be used as an example for illustration.

[0050] The Kiencke model expresses the tire-road friction coefficient as a parameterized function of the slip ratio, as follows:

[0051] (1)

[0052] In equation (1), Peak adhesion coefficient; This represents the tire slip ratio corresponding to the peak adhesion coefficient, and its value typically ranges from 0.1 to 0.2. In the small slip zone (i.e.,...) For example, s is in the range of 0.5% to 3%, because Compared to It can be ignored, in equation (1) The above model can be approximated as:

[0053] (2)

[0054] Therefore, the initial slope of the adhesion coefficient relative to the tire slip ratio can be defined as follows: Solving this equation inversely yields... ,in, .

[0055] It should be noted that the proportional relationship between slope and peak adhesion coefficient varies under different road surface types or tire characteristics, and therefore the corresponding mapping coefficients also differ. In this embodiment, extensive real-vehicle calibration experiments were conducted beforehand to calibrate the mapping coefficients under different road surface types (e.g., dry asphalt, wet asphalt, snow, ice, etc.), different vehicle speeds, and different tire operating conditions, and to establish the correspondence between slope and mapping coefficients.

[0056] Optionally, the correspondence between the slope and the mapping coefficient can be grouped and calibrated according to factors such as tire model and tire wear condition to improve mapping accuracy. Specifically, when a vehicle replaces its tires with non-original equipment brand tires, the original correspondence between the slope and the mapping coefficient may experience systematic deviations due to differences in rubber compound, tread pattern, and stiffness characteristics among different tire brands. Therefore, upon detecting a tire replacement event, the system automatically enters an adaptive learning mode for the μ-s (peak adhesion coefficient - tire slip ratio) parameter table. The tire replacement event can be triggered by user interaction commands, changes in tire pressure monitoring signals, etc. Under a known road surface type (e.g., dry asphalt road surface with a baseline peak adhesion coefficient μ of 0.9), the system continuously collects data points of the estimated adhesion coefficient and tire slip ratio during driving, constructing a μ-s curve sample for the current tire under this known road surface. Based on this, a Bayesian optimization method is used, with the collected sample data as observations, to update the posterior probability distribution of the mapping coefficient. During iterative optimization, the confidence level between prior information and newly collected samples is balanced, gradually converging to the optimal mapping coefficient that adapts to the characteristics of the new tire.

[0057] Step S202: Determine the target peak adhesion coefficient corresponding to the target slope based on the target slope and the target mapping coefficient.

[0058] Specifically, the target slope and the target mapping coefficient are multiplied to obtain the target peak adhesion coefficient corresponding to the target slope. The calculation formula is as follows:

[0059] (3)

[0060] In equation (3), The target peak adhesion coefficient corresponding to the target slope; The target slope; For example, the target mapping coefficients, The value can range from 0.05 to 0.1.

[0061] Step S203: Determine the current peak adhesion coefficient based on the target peak adhesion coefficient.

[0062] In practice, the target peak adhesion coefficient can be directly output as the current peak adhesion coefficient.

[0063] Optionally, a first-order low-pass filter or a weighted moving average can be used to smooth the target peak adhesion coefficient to obtain the current peak adhesion coefficient. For example, the target peak adhesion coefficient at the current time can be weighted and averaged with the current peak adhesion coefficient output at the previous time.

[0064] The method provided in this embodiment determines the target mapping coefficient corresponding to the target slope based on the correspondence between slope and mapping coefficient, and then determines the corresponding target peak adhesion coefficient based on the target slope and target mapping coefficient. Finally, it determines the current peak adhesion coefficient based on the target peak adhesion coefficient. Since this mapping coefficient is obtained directly through calibration experiments under different road surface types and tire combinations, it can adaptively match the differences in adhesion characteristics of different tire brands or models on the same road surface, improving the cross-vehicle generalization ability of the estimation method.

[0065] Figure 3 This is a flowchart of Embodiment 3 of the method for estimating the adhesion coefficient of the road surface provided in this application. Please refer to... Figure 3 Based on the above embodiments, determining the current peak adhesion coefficient according to the target peak adhesion coefficient includes:

[0066] Step S301: Obtain the current residual correlation features of the vehicle, and input the residual correlation features into the pre-trained residual correction model for processing, so that the residual correction model outputs the residual correction value.

[0067] Specifically, the residual-related features of the vehicle are obtained, including but not limited to the residual sequence of the adhesion coefficient estimate output by the physical model within a preset historical time window (e.g., the residual between the physical model estimate and the measured value 5 times before the current time constitutes a 5-dimensional feature vector), the energy distribution features of the vibration spectrum collected by the vehicle within a preset frequency band (i.e., vibration spectrum texture features), prior information on the current road surface type (e.g., represented in the form of 6 types of unique hot codes, covering dry asphalt, wet asphalt, snow, ice, gravel road, mud road, etc.), ambient temperature, tire temperature, rain sensor status and rainfall duration, vehicle load distribution features, and the current tire model identifier (represented in the form of unique hot codes to represent different tire brands or models).

[0068] The aforementioned residual-related features are input into a pre-trained residual correction model for processing, and the model outputs the residual correction value. The residual correction model includes, but is not limited to, extreme gradient boosting (XGBoost), light gradient boosting machine (LightGBM), random forest, or lightweight neural network.

[0069] In one embodiment, the residual correction model can be trained using the XGBoost algorithm. For example, by selecting 200 decision trees with a maximum tree depth of 5, the model file size is approximately 300KB, and the time for a single inference is approximately 1.5ms, which meets the real-time requirements.

[0070] It should be noted that this residual correction model can learn and compensate for the complex nonlinear deviations that are difficult to model in physical models under different road surface types, tire configurations and environmental conditions, thereby improving the estimation accuracy of the peak adhesion coefficient.

[0071] Step S302: Determine the current peak adhesion coefficient based on the target peak adhesion coefficient and the residual correction value.

[0072] In this step, the target peak adhesion coefficient is fused with the residual correction value to obtain the current peak adhesion coefficient. For example, in one possible implementation, the target peak adhesion coefficient is added to the residual correction value to obtain the current peak adhesion coefficient.

[0073] Optionally, to enhance the smoothness and stability of the fusion results, the residual correction value can be limited to a preset reasonable correction range (e.g., ±0.15) to avoid the estimated value jumping due to excessive correction under extreme conditions.

[0074] When there are missing values ​​in the feature input of the residual correction model (for example, some vehicles are not equipped with rain sensors), preset default values ​​(for example, the default value for rain status is 0) can be used to fill them in, so as to ensure the universality of the model on vehicles with various configurations.

[0075] Furthermore, for scenarios where the shape of the road surface μ-s curve (e.g., icy or snowy roads) differs significantly from the aforementioned Kiencke model, a nonlinear correction can be used to correct the target peak adhesion coefficient. This nonlinear correction can be generated based on the current slip ratio and can also be combined with vibration spectrum texture features in the calculation, outputting a nonlinear correction that varies with the slip ratio to capture and compensate for the nonlinear deviation of the physical model in different slip ratio ranges.

[0076] The method provided in this embodiment acquires multi-dimensional residual correlation features such as vibration spectrum texture, road surface type prior, ambient temperature, tire temperature, rainfall status, load distribution, and tire model, and processes them using a pre-trained residual correction model. This effectively learns and compensates for complex nonlinear biases that are difficult to model in physical models under different road surface types, tire configurations, and environmental conditions, thereby ensuring the smoothness and stability of the estimation results and improving the accuracy and robustness of the estimation method.

[0077] Meanwhile, this embodiment employs an asymmetric hybrid design, combining a physical model with a data-driven fusion architecture. The physical model serves as a deterministic foundation, outputting the target peak adhesion coefficient and providing a safety lower bound with a clear physical mechanism that is interpretable and verifiable. The residual correction part learns the residual error of the physical model under complex operating conditions, rather than replacing the physical model for end-to-end black-box mapping. This avoids unpredictable outputs caused by blind extrapolation from a purely data-driven model under unknown conditions. Through this asymmetric hybrid architecture design, the estimation accuracy of the peak adhesion coefficient is significantly improved while strictly ensuring the interpretability and reliability of the system.

[0078] Figure 4 The flowchart for Embodiment 4 of the method for estimating the road surface adhesion coefficient provided in this application is shown below. Please refer to... Figure 4 Based on the above embodiments, the current peak adhesion coefficient is determined according to the target peak adhesion coefficient and the residual correction value, including:

[0079] Step S401: Dimensionality reduction is performed on the residual correlation features to obtain dimensionality-reduced features.

[0080] In practice, dimensionality reduction can be achieved in various ways. For example, principal component analysis (PCA) and linear discriminant analysis (LDA) can be used.

[0081] The dimensionality reduction process described above effectively reduces the dimensionality of distance calculations in subsequent steps while preserving the main feature information, thus improving the real-time performance and efficiency of the method. The feature dimension after dimensionality reduction can be set according to actual application requirements to balance computational efficiency and information preservation. For example, PCA can be used to reduce the dimensionality to an 8-dimensional embedding vector.

[0082] Step S402: Obtain the training sample set of the residual correction model, and for each training sample in the training sample set, determine the distance between the dimensionality reduction feature and the training sample to obtain the distance corresponding to the training sample.

[0083] In this step, all training samples are extracted from the training sample set of the residual correction model, and the distance between the dimensionality-reduced feature at the current time step and each training sample is calculated. The training sample set originates from the sample data used during the training of the residual correction model, and each training sample contains the residual-related feature vector at the corresponding time step.

[0084] It should be noted that the above distance calculation can be achieved in various ways. For example, Euclidean distance, Mahalanobis distance, etc. can be used.

[0085] Step S403: Determine the nearest neighbor distance and distance threshold based on the distances corresponding to each training sample.

[0086] Based on all calculated distances, the nearest neighbor distance and distance threshold are determined. The nearest neighbor distance is used to quantify the proximity of the current dimensionality-reduced feature to the most similar sample in the training sample set, and the distance threshold is used to determine whether this proximity is sufficient to consider the current feature to be within the distribution coverage of the training samples.

[0087] In one possible implementation, the nearest neighbor distance and distance threshold are determined based on the distances corresponding to each training sample, including: determining the minimum distance among the distances corresponding to each training sample as the nearest neighbor distance; sorting the distances corresponding to each training sample in ascending order to obtain a sorting result, and determining the distance located at a preset quantile position in the sorting result as the distance threshold.

[0088] In practice, the minimum value among the distances corresponding to each training sample is determined as the nearest neighbor distance, so as to quantify the proximity between the current dimensionality reduction feature and the most similar sample in the training sample set.

[0089] Furthermore, the calculated distances corresponding to all training samples are sorted in ascending order to obtain the sorted distance sequence {d1, d2, ..., d...}. N}, where N is the total number of training samples. The distance value located at a preset quantile position in the sorted sequence is determined as the distance threshold. This preset quantile can be experimentally calibrated by those skilled in the art based on real vehicle data, and typical values ​​can be 0.95, 0.99, etc. For example, if the preset quantile is p (a value between 0 and 1, such as p=0.99), then the index position of the distance threshold is... (i.e., rounded up), the distance threshold is the distance corresponding to the index position.

[0090] Furthermore, the mean and standard deviation of the distances corresponding to each training sample can be calculated, and the sum of the mean and the standard deviation of a preset multiple can be used as the distance threshold. Alternatively, the product of the maximum distance among the distances corresponding to each training sample and a preset scaling factor can be used as the distance threshold. This embodiment does not specifically limit the method for determining the distance threshold; those skilled in the art can flexibly select the appropriate method according to actual application needs.

[0091] Step S404: When the nearest neighbor distance is greater than the distance threshold, the target peak adhesion coefficient is determined as the current peak adhesion coefficient; when the nearest neighbor distance is less than or equal to the distance threshold, the sum of the target peak adhesion coefficient and the residual correction amount is determined as the current peak adhesion coefficient.

[0092] Specifically, when the nearest neighbor distance is greater than the distance threshold, it indicates that the dimensionality reduction feature at the current moment is far from any sample in the training sample set, and the current condition deviates from the training sample distribution range of the residual correction model, making its output unreliable. In this case, the target peak attachment coefficient is directly used as the current peak attachment coefficient, i.e., no residual correction value is used for correction, thereby ensuring the reliability of the output result.

[0093] When the nearest neighbor distance is less than or equal to the distance threshold, it indicates that the current dimensionality reduction feature is sufficiently close to at least one sample in the training sample set, the current working condition is within or close to the distribution coverage of the training samples, and the output of the residual correction model has high confidence. At this point, the target peak adhesion coefficient is added to the residual correction value to obtain the current peak adhesion coefficient. The specific calculation formula is as follows:

[0094] (4)

[0095] In equation (4), This represents the current peak adhesion coefficient. The target peak adhesion coefficient; This is the residual correction value.

[0096] For example, the current vehicle is driving on a dry asphalt road surface, and the target peak adhesion coefficient is... At this point, the residual correction model receives the current multidimensional residual correlation features and outputs the residual correction value. This indicates a slight difference between the vehicle's tire characteristics and the calibration parameters, requiring fine-tuning of the physical model output. The residual correction value is within the preset limit range [-0.15, 0.15] and does not require truncation. Simultaneously, the nearest neighbor distance between the current dimensionality-reduced feature and the training sample set is calculated to be 2.3, and the distance threshold is 7.8, meaning the current operating condition is within the training sample coverage range of the residual correction model (in-domain samples), indicating that the residual correction value is reliable. Based on this, the current peak adhesion coefficient is output. .

[0097] For example, when a vehicle is currently driving on an icy or snowy road, the target peak coefficient of friction is... Since icy and snowy road surfaces are rarely covered in the training sample set of the residual correction model, the residual correction model outputs the residual correction value after receiving the current multidimensional residual correlation features. The residual correction value exceeds the preset limit range [-0.15, 0.15], and is truncated to 0.15. Simultaneously, the nearest neighbor distance between the current dimensionality reduction feature and the training sample set is calculated to be 12.5, while the distance threshold is 7.8. This indicates that the current operating condition significantly deviates from the training sample coverage range of the residual correction model (out-of-domain samples), triggering an out-of-domain alarm and determining that the residual correction value is unreliable. Based on this, the current peak adhesion coefficient is output. .

[0098] It should be noted that, besides using nearest neighbor distance as the out-of-domain discrimination criterion, other implementations can also use divergence (e.g., KL divergence) for discrimination. Specifically, the divergence between the feature distribution corresponding to the dimensionality reduction feature and the feature distribution corresponding to the training sample is calculated. When the divergence is greater than a preset divergence threshold, the target peak attachment coefficient is directly used as the current peak attachment coefficient; when the divergence is less than or equal to the preset divergence threshold, the target peak attachment coefficient is added to the residual correction value to obtain the current peak attachment coefficient.

[0099] The method provided in this embodiment performs dimensionality reduction on residual-related features, calculates the distance between the current dimensionality-reduced feature and each training sample in the training sample set, determines the nearest neighbor distance and distance threshold based on the distances corresponding to each training sample, and then determines whether the current working condition is within the training sample coverage range of the residual correction model based on the nearest neighbor distance and distance threshold. In turn, it can select an appropriate strategy to determine the current peak attachment coefficient based on the judgment result, so as to ensure that it does not rely on unreliable residual correction output under unfamiliar working conditions. The physical model always provides reliable estimation results as a safety lower bound, effectively improving the accuracy and robustness of the estimation method.

[0100] Figure 5 The flowchart for Embodiment 5 of the method for estimating the adhesion coefficient of the road surface provided in this application is shown below. Please refer to... Figure 5 Based on the above embodiments, according to the target data sequence, the slope of the estimated adhesion coefficient relative to the tire slip ratio is determined to obtain the target slope, including:

[0101] Step S501: Extract multiple target data points from the target data sequence and arrange the multiple target data points in chronological order to obtain a sub-data sequence.

[0102] The target data sequence contains a large number of data points, some of which may correspond to tire slip ratios in the nonlinear region or experience abnormal fluctuations due to sensor noise, transient operating conditions, or other interference. Therefore, this embodiment selects data points from the target data sequence that meet preset conditions as target data points.

[0103] For example, data points with tire slip ratios within a preset slip ratio range can be extracted as target data points. For instance, the preset slip ratio range might be 0.5% to 3%, meaning only data points within small slip zones are retained. The upper and lower limits of this preset slip ratio range can be set according to actual application requirements.

[0104] For example, in one embodiment, each data point in the target data sequence is traversed to determine whether each data point meets a preset condition; wherein, the preset condition is that the tire slip ratio in the target data point is within a preset slip ratio range, the absolute value of the rate of change of the tire slip ratio in the target data point relative to the tire slip ratio in the previous data point is less than a first preset threshold, and the absolute value of the longitudinal force of the tire corresponding to the target data point is greater than a second preset threshold; multiple data points that meet the preset conditions are determined as multiple target data points.

[0105] In specific implementation, each data point in the target data sequence is traversed, and each data point is judged to meet the preset conditions in turn. The preset conditions include: (1) The tire slip ratio in the data point is within the preset slip ratio range. For example, the preset slip ratio range can be 0.5% to 3%, which is within the small slip zone, and can ensure that the adhesion coefficient and slip ratio meet the approximately linear relationship. The upper and lower limits of the preset slip ratio range can be set according to the actual application requirements. (2) The absolute value of the rate of change of the tire slip ratio in the data point relative to the tire slip ratio in the previous data point (i.e., the change in tire slip ratio per unit time) is less than the first preset threshold. This condition is used to exclude data points at the moment of violent fluctuation or transient jump in slip ratio, so as to ensure that the current moment is in a quasi-steady state. For example, the first preset threshold can be set to 1% / s, that is, the change in slip ratio per second is required not to exceed 1%, so as to ensure that the contact state between the tire and the road surface is relatively stable during the data acquisition process. (3) The absolute value of the longitudinal force of the tire corresponding to the data point is greater than the second preset threshold. This condition is used to exclude situations where signal noise causes unreliable data when the slip ratio is extremely low, ensuring that the signal-to-noise ratio of the data points meets the requirements. The tire longitudinal force can be estimated from the vehicle's driving force or braking force. When the absolute value of the longitudinal force is small, even a small amount of sensor noise can cause a large relative error in the adhesion coefficient estimate. By setting a reasonable lower limit for the longitudinal force (for example, the second preset threshold can be set to 200N), this problem can be effectively avoided, ensuring the reliability and stability of the adhesion coefficient estimate.

[0106] When a data point in the target data sequence simultaneously meets all three of the above conditions, that data point is designated as a target data point. In this way, the selected target data points are all within the small slip zone, quasi-steady-state condition, and effective load range, which can truly reflect the adhesion characteristics of the current road surface.

[0107] For example, data points whose estimated adhesion coefficient is within a preset adhesion coefficient range can be extracted as target data points to exclude invalid data caused by sensor saturation or signal abnormalities.

[0108] It should be noted that the above filtering methods can be used individually or in combination according to actual needs.

[0109] Furthermore, after extracting multiple target data points, these data points are arranged in chronological order (i.e., the order in which they were collected) to obtain a sub-data sequence.

[0110] Step S502: Determine the target slope based on the sub-data sequence.

[0111] In this embodiment, the method for determining the target slope includes, but is not limited to: performing linear fitting on the data points in the sub-data sequence and determining the slope obtained by fitting as the target slope; and using a random sampling consensus algorithm to fit the data points in the sub-data sequence, solving for the slope of the adhesion coefficient estimate relative to the tire slip ratio, and obtaining the target slope.

[0112] The method provided in this embodiment extracts target data points that meet preset conditions from the target data sequence and arranges them into sub-data sequences, thereby effectively filtering the data involved in slope estimation. This enables the filtered target data points to truly and stably reflect the adhesion characteristics of the current road surface, which helps to improve the accuracy of subsequent slope estimation.

[0113] Figure 6 The flowchart for Embodiment Six of the method for estimating the road surface adhesion coefficient provided in this application is shown below. Please refer to... Figure 6 Based on the above embodiments, the target slope is determined according to the sub-data sequence, including:

[0114] Step S601: According to the time sequence, the target data point at the first moment in the sub-data sequence is taken as the data point to be processed, the tire slip ratio in the data point to be processed is taken as the observed slip ratio, and the estimated value of the adhesion coefficient in the data point to be processed is taken as the observed adhesion coefficient.

[0115] In this step, the initialization operation of the recursive estimation process is performed. Following the temporal sequence, the target data point at the first time step is extracted from the sub-data sequence as the starting point for subsequent recursive processing. The tire slip ratio at this data point is taken as the observed slip ratio at the current time step. The estimated adhesion coefficient at this data point is used as the observed adhesion coefficient at the current time. .

[0116] It should be noted that, during the recursive estimation process, the observed adhesion coefficient and tire slip ratio satisfy the following observation equation:

[0117] (5)

[0118] In equation (5), The observed adhesion coefficient at the current moment; The observed slip rate at the current moment; Let be the slope at the current moment to be estimated; For example, to observe noise. ,Right now The noise follows a Gaussian white noise distribution with a mean of 0 and a covariance of R, where R is the preset observation noise covariance.

[0119] In practice, for the first time step of the recursive estimation, since there is no posterior estimate from the previous time step, an initial state needs to be pre-set. This involves setting the initial posterior slope estimate. (For example, it can be set to the slope value of 12 corresponding to dry asphalt pavement), and the initial posterior covariance P(0) (for example, it can be set to 4 to characterize the uncertainty of the initial estimate).

[0120] Optionally, the initial posterior slope estimate can be set based on the default road surface type when the vehicle starts (e.g., the default assumption is dry asphalt) or the stored value at the end of the previous driving cycle to speed up convergence.

[0121] Step S602: Based on the posterior slope estimate of the previous time step, determine the prior slope estimate of the current time step, and based on the preset forgetting factor and the posterior covariance of the previous time step, determine the prior covariance of the current time step.

[0122] In this step, the state equation is used to make a priori prediction of the state at the current moment. That is, without relying on the current observation, the priori estimate of the current moment is calculated based on the posterior estimate of the previous moment.

[0123] Specifically, based on the posterior slope estimate from the previous time step... Determine the prior slope estimate at the current moment. Taking a random walk model of the state equation as an example, the slope is considered constant in the short term but allowed to vary slowly. Therefore, the prior slope estimate is equal to the posterior slope estimate of the previous time step, i.e. .

[0124] At the same time, based on the posterior covariance of the previous time step Preset forgetting factor (The range of values ​​is) Determine the prior covariance at the current moment. .For example, Q represents the preset process noise covariance. The preset forgetting factor is used to dilate the posterior covariance of the previous time step, so that the influence of old data on the current estimate decays exponentially over time; the preset process noise covariance characterizes the time-varying rate allowed by the slope itself.

[0125] Step S603: Based on the prior slope estimate, prior covariance, observed slip rate, and observed adhesion coefficient, determine the posterior slope estimate and the posterior covariance at the current time.

[0126] Specifically, the prior prediction results are corrected using the observations at the current time to obtain the posterior slope estimate and posterior covariance at the current time.

[0127] Taking unscented Kalman filtering as an example, based on the prior slope estimate, prior covariance, observed slip rate, and observed adhesion coefficient, the posterior slope estimate and the posterior covariance at the current time are determined, including:

[0128] (1) Generate multiple sampling points based on the prior slope estimate and prior covariance, and determine multiple predicted adhesion coefficients based on the multiple sampling points and observed slip rate.

[0129] Multiple sampling points (Sigma points) are generated based on the prior slope estimate and the prior covariance. In a one-dimensional state, typically three sampling points are generated: a center point centered on the prior slope estimate, and two sampling points symmetrically distributed on either side of the center point, their distribution range determined by the prior covariance. Each sampling point represents a possible assumption about the slope value. For example, sampling points... , , The calculation formula is as follows: , , .

[0130] Multiply each sampling point by the observed slip rate at the current time to obtain the predicted adhesion coefficient for each sampling point. The calculation formula is as follows:

[0131] (6)

[0132] In equation (6), To predict the adhesion coefficient; Let be the slope value of the i-th sampling point; To observe the slip ratio. Therefore, the sampling points in the state space are mapped to the observation space, resulting in multiple predicted observations (i.e., ...). ).

[0133] (2) Determine the Kalman gain based on the prior slope estimate, multiple sampling points and multiple predicted adhesion coefficients.

[0134] Specifically, the weighted covariance (i.e., state-observation cross-covariance) is calculated between the deviation of the sampling points from the prior slope estimate and the deviation of the predicted adhesion coefficient from the mean of the predicted adhesion coefficient. ), and the weighted variance of the predicted adhesion coefficient relative to its mean (i.e., the observed-predicted covariance). Then the Kalman gain .

[0135] (3) Determine the mean of the predicted adhesion coefficients based on multiple predicted adhesion coefficients, and correct the prior slope estimate based on the Kalman gain and the residual between the observed adhesion coefficients and the mean of the predicted adhesion coefficients to obtain the posterior slope estimate at the current time.

[0136] The mean of the predicted adhesion coefficients is determined based on multiple predicted adhesion coefficients, and the residual between the observed adhesion coefficient and the mean of the predicted adhesion coefficients at the current time is calculated. Based on this residual and the Kalman gain, the prior slope estimate is corrected to obtain the posterior slope estimate at the current time, calculated using the following formula:

[0137] (7)

[0138] In equation (7), This is the posterior slope estimate at the current time. This is the prior slope estimate at the current moment; Kalman gain; To observe the adhesion coefficient; To predict the mean adhesion coefficient.

[0139] (4) Determine the observation prediction covariance based on multiple prediction adhesion coefficients and the mean of prediction adhesion coefficients, and correct the prior covariance based on the Kalman gain and the observation prediction covariance to obtain the posterior covariance at the current time.

[0140] In practice, the formula for calculating the posterior covariance at the current moment is as follows:

[0141] (8)

[0142] In equation (8), The posterior covariance at the current moment; The prior covariance at the current moment; Kalman gain; To observe and predict covariance; This is the transpose of the Kalman gain.

[0143] Step S604: Determine whether the preset convergence condition is met.

[0144] In this step, it is determined whether the recursive estimation has reached convergence. Convergence conditions can be implemented in various ways, including but not limited to: when the number of data points processed reaches a preset number (e.g., 200 data points, corresponding to 2 seconds), the estimation is determined to have converged; when the absolute value of the rate of change of the posterior slope estimate at adjacent time points is less than a preset rate of change threshold, the estimation is determined to have converged.

[0145] Optionally, the above convergence conditions can be used in combination to ensure the reliability of the convergence judgment.

[0146] Step S605: If not, take the target data point at the next moment after the data point to be processed as the data point to be processed, and execute the step of taking the tire slip ratio of the data point to be processed as the observed slip ratio again.

[0147] When it is determined that the preset convergence condition has not been met, the data point to be processed is updated to the target data point at the next time step after the current data point in the sub-data sequence, and then the step of using the tire slip ratio of the data point to be processed as the observed slip ratio is executed, and the recursive estimation process is repeated.

[0148] Therefore, each target data point in the sub-data sequence is processed one by one. In each iteration, the slope estimate is recursively updated using the observation information at the current time until the convergence condition is met or all data points in the sub-data sequence have been traversed.

[0149] Step S606: If yes, determine the posterior slope estimate at the current time as the target slope.

[0150] When the preset convergence condition is met, the posterior slope estimate at the current moment is determined as the final target slope.

[0151] The method provided in this embodiment determines the target slope through recursive estimation. It can process each target data point in the sub-data sequence one by one, updating the slope estimate online as soon as a new data point arrives, without waiting for the complete data sequence to be collected, thus fully meeting the real-time output requirements of the estimation results during vehicle operation. Simultaneously, by employing a recursive estimation framework based on unscented Kalman filtering with a forgetting factor, the algorithm uses the forgetting factor to exponentially decay the weights of historical data, enabling it to adaptively track the time-varying characteristics of road surface adhesion conditions, ensuring estimation stability while also considering the response speed to sudden changes in road surface conditions. Furthermore, by using unscented transformation to pass sampling points to the observation equation, it maintains good estimation capability even when the tire exhibits nonlinear characteristics, thereby ensuring estimation accuracy under multiple operating conditions.

[0152] The present embodiment will now be described and illustrated through preferred embodiments.

[0153] Figure 7 The flowchart of Embodiment Seven of the method for estimating the adhesion coefficient of the road surface provided in this application is as follows: Figure 7 As shown, the estimation method includes the following steps:

[0154] Step S701: Obtain the target data sequence of the vehicle; wherein, the target data sequence includes data points at different times, and the data points include the estimated value of the adhesion coefficient and the tire slip ratio;

[0155] Step S702: Based on the target data sequence, determine the slope of the estimated adhesion coefficient relative to the tire slip ratio to obtain the target slope;

[0156] Step S703: Determine the target mapping coefficient corresponding to the target slope based on the target slope and the preset correspondence between the slope and the mapping coefficient; wherein, the preset correspondence between the slope and the mapping coefficient is used to record the correspondence between the slope of the adhesion coefficient and the tire slip ratio and the mapping coefficient, and the mapping coefficient is the proportional coefficient between the slope and the peak adhesion coefficient corresponding to the slope.

[0157] Step S704: Determine the target peak adhesion coefficient corresponding to the target slope based on the target slope and the target mapping coefficient;

[0158] Step S705: Obtain the current residual correlation features of the vehicle, and input the residual correlation features into the pre-trained residual correction model for processing, so that the residual correction model outputs the residual correction value;

[0159] Step S706: Perform dimensionality reduction on the residual correlation features to obtain dimensionality-reduced features; obtain the training sample set of the residual correction model, and for each training sample in the training sample set, determine the distance between the dimensionality-reduced features and the training sample to obtain the distance corresponding to the training sample.

[0160] Step S707: Determine the nearest neighbor distance and distance threshold based on the distances corresponding to each training sample; when the nearest neighbor distance is greater than the distance threshold, determine the target peak attachment coefficient as the current peak attachment coefficient; and when the nearest neighbor distance is less than or equal to the distance threshold, determine the sum of the target peak attachment coefficient and the residual correction amount as the current peak attachment coefficient.

[0161] It should be noted that the steps shown in the above process or in the flowchart of the accompanying figures can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0162] This embodiment also provides a device for estimating the adhesion coefficient of a road surface, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. The terms "module," "unit," "subunit," etc., used below refer to combinations of software and / or hardware that perform a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0163] Figure 8 This is a structural block diagram of the road surface adhesion coefficient estimation device in this embodiment, as shown below. Figure 8As shown, the estimation device includes:

[0164] The acquisition module 10 is used to acquire the target data sequence of the vehicle; wherein, the target data sequence includes data points at different times, and the data points include the estimated value of the adhesion coefficient and the tire slip ratio;

[0165] Analysis module 20 is used to determine the slope of the estimated adhesion coefficient relative to the tire slip ratio based on the target data sequence, and obtain the target slope;

[0166] The processing module 30 is used to determine the current peak adhesion coefficient of the road surface on which the vehicle is currently traveling, based on the target slope.

[0167] Optionally, in one possible implementation, processing module 30 is specifically used for:

[0168] Based on the target slope and the preset correspondence between the slope and the mapping coefficient, the target mapping coefficient corresponding to the target slope is determined; wherein, the preset correspondence between the slope and the mapping coefficient is used to record the correspondence between the slope of the adhesion coefficient relative to the tire slip ratio and the mapping coefficient, and the mapping coefficient is the proportional coefficient between the slope and the peak adhesion coefficient corresponding to the slope.

[0169] Based on the target slope and the target mapping coefficient, determine the target peak adhesion coefficient corresponding to the target slope;

[0170] The current peak adhesion coefficient is determined based on the target peak adhesion coefficient.

[0171] Optionally, in one possible implementation, processing module 30 is specifically used for:

[0172] Obtain the current residual correlation features of the vehicle, and input the residual correlation features into the pre-trained residual correction model for processing, so that the residual correction model outputs the residual correction value;

[0173] The current peak adhesion coefficient is determined based on the target peak adhesion coefficient and the residual correction value.

[0174] Optionally, in one possible implementation, processing module 30 is specifically used for:

[0175] The residual correlation features are reduced in dimensionality to obtain the dimensionality-reduced features.

[0176] Obtain the training sample set of the residual correction model, and for each training sample in the training sample set, determine the distance between the dimensionality reduction feature and the training sample to obtain the distance corresponding to the training sample.

[0177] Based on the distances corresponding to each training sample, determine the nearest neighbor distance and the distance threshold;

[0178] When the nearest neighbor distance is greater than the distance threshold, the target peak adhesion coefficient is determined as the current peak adhesion coefficient. When the nearest neighbor distance is less than or equal to the distance threshold, the sum of the target peak adhesion coefficient and the residual correction is determined as the current peak adhesion coefficient.

[0179] Optionally, in one possible implementation, processing module 30 is specifically used for:

[0180] The minimum distance among the distances corresponding to each training sample is determined as the nearest neighbor distance;

[0181] The distances corresponding to each training sample are sorted in ascending order to obtain the sorting result, and the distances located at the preset quantile positions in the sorting result are determined as the distance thresholds.

[0182] Optionally, in one possible implementation, the analysis module 20 is specifically used for:

[0183] Multiple target data points are extracted from the target data sequence and arranged in temporal order to obtain a sub-data sequence;

[0184] Determine the target slope based on the sub-data sequence.

[0185] Optionally, in one possible implementation, the analysis module 20 is specifically used for:

[0186] Traverse each data point in the target data sequence and determine whether each data point meets the preset conditions. The preset conditions are that the tire slip ratio in the target data point is within the preset slip ratio range, the absolute value of the rate of change of the tire slip ratio in the target data point relative to the tire slip ratio in the previous data point is less than the first preset threshold, and the absolute value of the longitudinal force of the tire corresponding to the target data point is greater than the second preset threshold.

[0187] Multiple data points that meet the preset conditions are identified as multiple target data points.

[0188] Optionally, in one possible implementation, the analysis module 20 is specifically used for:

[0189] According to the time sequence, the target data point at the first time point in the sub-data sequence is taken as the data point to be processed, the tire slip ratio in the data point to be processed is taken as the observed slip ratio, and the estimated value of the adhesion coefficient in the data point to be processed is taken as the observed adhesion coefficient.

[0190] Based on the posterior slope estimate of the previous time step, determine the prior slope estimate of the current time step, and based on the preset forgetting factor and the posterior covariance of the previous time step, determine the prior covariance of the current time step.

[0191] Based on the prior slope estimate, prior covariance, observed slip rate, and observed adhesion coefficient, determine the posterior slope estimate and the posterior covariance at the current time.

[0192] Determine whether the preset convergence conditions are met;

[0193] If not, take the target data point at the next moment after the data point to be processed as the data point to be processed, and execute the step of taking the tire slip ratio of the data point to be processed as the observed slip ratio again;

[0194] If so, the posterior slope estimate at the current moment is determined as the target slope.

[0195] Optionally, in one possible implementation, the analysis module 20 is specifically used for:

[0196] Multiple sampling points are generated based on the prior slope estimate and prior covariance, and multiple predicted adhesion coefficients are determined based on the multiple sampling points and the observed slip rate.

[0197] The Kalman gain is determined based on the prior slope estimate, multiple sampling points, and multiple predicted adhesion coefficients.

[0198] The mean of the predicted adhesion coefficients is determined based on multiple predicted adhesion coefficients, and the prior slope estimate is corrected based on the Kalman gain and the residual between the observed adhesion coefficients and the mean of the predicted adhesion coefficients to obtain the posterior slope estimate at the current time.

[0199] The observed prediction covariance is determined based on multiple predicted adhesion coefficients and the mean of the predicted adhesion coefficients. The prior covariance is then corrected based on the Kalman gain and the observed prediction covariance to obtain the posterior covariance at the current time.

[0200] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can reside in the same processor; or the above modules can be located in different processors in any combination.

[0201] This embodiment also provides a vehicle including a control device, which includes a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to perform the steps in any of the above method embodiments.

[0202] It should be noted that the specific examples in this embodiment can refer to the examples described in the above embodiments and optional implementations, and will not be repeated in this embodiment.

[0203] Furthermore, in conjunction with the road surface adhesion coefficient estimation method provided in the above embodiments, this embodiment can also provide a storage medium for implementation. This storage medium stores a computer program; when executed by a processor, the computer program implements any of the road surface adhesion coefficient estimation methods described in the above embodiments.

[0204] It should be understood that the specific embodiments described herein are merely illustrative of the application and not intended to limit it. All other embodiments derived by those skilled in the art based on the embodiments provided in this application without inventive effort are within the scope of protection of this application.

[0205] Obviously, the accompanying drawings are merely some examples or embodiments of this application. Those skilled in the art can apply this application to other similar situations based on these drawings without any creative effort. Furthermore, it is understood that although the work done in this development process may be complex and lengthy, for those skilled in the art, certain design, manufacturing, or production modifications made based on the technical content disclosed in this application are merely conventional technical means and should not be considered as insufficient disclosure of this application.

[0206] The term "embodiment" in this application refers to a specific feature, structure, or characteristic described in connection with an embodiment that may be included in at least one embodiment of this application. The appearance of this phrase in various places in the specification does not necessarily imply the same embodiment, nor does it imply that it is mutually exclusive with or independent of other embodiments. It will be clearly or implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments without conflict.

[0207] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of patent protection. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the appended claims.

Claims

1. A method for estimating the adhesion coefficient of a road surface, characterized in that, The estimation method includes: Obtain a target data sequence for the vehicle; wherein the target data sequence includes data points at different times, and the data points include an estimated coefficient of adhesion and a tire slip ratio; Based on the target data sequence, the slope of the estimated adhesion coefficient relative to the tire slip ratio is determined to obtain the target slope; The current peak adhesion coefficient of the road surface on which the vehicle is currently traveling is determined based on the target slope.

2. The method for estimating the adhesion coefficient of a road surface according to claim 1, characterized in that, Determining the current peak adhesion coefficient of the road surface on which the vehicle is currently traveling based on the target slope includes: Based on the target slope and the preset correspondence between slope and mapping coefficient, the target mapping coefficient corresponding to the target slope is determined; wherein, the preset correspondence between slope and mapping coefficient is used to record the correspondence between the slope of the adhesion coefficient relative to the tire slip ratio and the mapping coefficient, and the mapping coefficient is the proportional coefficient between the slope and the peak adhesion coefficient corresponding to the slope. Based on the target slope and the target mapping coefficient, determine the target peak adhesion coefficient corresponding to the target slope; The current peak adhesion coefficient is determined based on the target peak adhesion coefficient.

3. The method for estimating the adhesion coefficient of a road surface according to claim 2, characterized in that, Determining the current peak adhesion coefficient based on the target peak adhesion coefficient includes: Obtain the current residual correlation features of the vehicle, and input the residual correlation features into a pre-trained residual correction model for processing, so that the residual correction model outputs the residual correction value; The current peak adhesion coefficient is determined based on the target peak adhesion coefficient and the residual correction value.

4. The method for estimating the adhesion coefficient of a road surface according to claim 3, characterized in that, Determining the current peak adhesion coefficient based on the target peak adhesion coefficient and the residual correction value includes: The residual correlation features are subjected to dimensionality reduction processing to obtain dimensionality-reduced features; Obtain the training sample set of the residual correction model, and for each training sample in the training sample set, determine the distance between the dimensionality reduction feature and the training sample to obtain the distance corresponding to the training sample; Based on the distances corresponding to each training sample, determine the nearest neighbor distance and the distance threshold; When the nearest neighbor distance is greater than the distance threshold, the target peak adhesion coefficient is determined as the current peak adhesion coefficient; and when the nearest neighbor distance is less than or equal to the distance threshold, the sum of the target peak adhesion coefficient and the residual correction is determined as the current peak adhesion coefficient.

5. The method for estimating the adhesion coefficient of a road surface according to claim 4, characterized in that, The step of determining the nearest neighbor distance and distance threshold based on the distances corresponding to each of the training samples includes: The minimum distance among the distances corresponding to each of the training samples is determined as the nearest neighbor distance; The distances corresponding to each training sample are sorted in ascending order to obtain a sorting result, and the distance located at a preset quantile position in the sorting result is determined as the distance threshold.

6. The method for estimating the adhesion coefficient of a road surface according to claim 1, characterized in that, The step of determining the slope of the estimated adhesion coefficient relative to the tire slip ratio based on the target data sequence, to obtain the target slope, includes: Multiple target data points are extracted from the target data sequence, and the multiple target data points are arranged in chronological order to obtain a sub-data sequence; The target slope is determined based on the sub-data sequence.

7. The method for estimating the adhesion coefficient of a road surface according to claim 6, characterized in that, The step of extracting multiple target data points from the target data sequence includes: Traverse each data point in the target data sequence and determine whether each data point meets a preset condition; wherein, the preset condition is that the tire slip ratio in the target data point is within a preset slip ratio range, the absolute value of the rate of change of the tire slip ratio in the target data point relative to the tire slip ratio in the previous data point is less than a first preset threshold, and the absolute value of the longitudinal force of the tire corresponding to the target data point is greater than a second preset threshold. The data points that meet the preset conditions are identified as the multiple target data points.

8. The method for estimating the adhesion coefficient of a road surface according to claim 6, characterized in that, Determining the target slope based on the sub-data sequence includes: According to the time sequence, the target data point at the first moment in the sub-data sequence is taken as the data point to be processed, the tire slip ratio in the data point to be processed is taken as the observed slip ratio, and the estimated value of the adhesion coefficient in the data point to be processed is taken as the observed adhesion coefficient. Based on the posterior slope estimate of the previous time step, determine the prior slope estimate of the current time step, and based on the preset forgetting factor and the posterior covariance of the previous time step, determine the prior covariance of the current time step. Based on the prior slope estimate, the prior covariance, the observed slip rate, and the observed adhesion coefficient, determine the posterior slope estimate and the posterior covariance at the current time. Determine whether the preset convergence conditions are met; If not, take the target data point at the next moment after the data point to be processed as the data point to be processed, and execute the step of taking the tire slip ratio of the data point to be processed as the observed slip ratio again; If so, the posterior slope estimate at the current moment is determined as the target slope.

9. The method for estimating the adhesion coefficient of a road surface according to claim 8, characterized in that, The step of determining the posterior slope estimate and the posterior covariance at the current time based on the prior slope estimate, the prior covariance, the observed slip rate, and the observed adhesion coefficient includes: Multiple sampling points are generated based on the prior slope estimate and the prior covariance, and multiple predicted adhesion coefficients are determined based on the multiple sampling points and the observed slip rate. The Kalman gain is determined based on the prior slope estimate, the plurality of sampling points, and the plurality of predicted adhesion coefficients; The mean of the predicted adhesion coefficients is determined based on the plurality of predicted adhesion coefficients, and the prior slope estimate is corrected based on the Kalman gain and the residual between the observed adhesion coefficients and the mean of the predicted adhesion coefficients to obtain the posterior slope estimate at the current time. The observation prediction covariance is determined based on the plurality of predicted adhesion coefficients and the mean of the predicted adhesion coefficients, and the prior covariance is corrected based on the Kalman gain and the observation prediction covariance to obtain the posterior covariance at the current time.

10. A device for estimating the coefficient of adhesion of a road surface, characterized in that, The estimation device includes: An acquisition module is used to acquire a target data sequence of the vehicle; wherein, the target data sequence includes data points at different times, and the data points include an estimated value of the adhesion coefficient and a tire slip ratio; The analysis module is used to determine the slope of the estimated adhesion coefficient relative to the tire slip ratio based on the target data sequence, and obtain the target slope; The processing module is used to determine the current peak adhesion coefficient of the road surface on which the vehicle is currently traveling, based on the target slope.

11. A vehicle, characterized in that, include: A control device, the control device including a memory, a processor, and a computer program; The computer program is stored in the memory and configured to be executed by the processor to implement the method for estimating the adhesion coefficient of the road surface as described in any one of claims 1 to 9.

Images