CN113911124A

CN113911124A - Road surface adhesion coefficient estimation method based on Bayes and electronic equipment

Info

Publication number: CN113911124A
Application number: CN202111140257.3A
Authority: CN
Inventors: 钟毅
Original assignee: Wuhan University of Technology WUT
Current assignee: Wuhan University of Technology WUT
Priority date: 2021-09-28
Filing date: 2021-09-28
Publication date: 2022-01-11
Anticipated expiration: 2041-09-28
Also published as: CN113911124B

Abstract

The invention discloses a road surface adhesion coefficient estimation method based on Bayes and an electronic device. The method divides the integral data point set into a plurality of data subsets, the point sets of the road surfaces with different attachment coefficients in each data subset present the trend of normal distribution, the expectation and the variance of the normal distribution of the point sets of the road surfaces with different attachment coefficients are respectively different, and the road surfaces with different attachment coefficients can be distinguished by learning the difference; and then, by constructing a normal distribution probability function of the road surface model of each attachment coefficient in each data subset, when a real-time data point arrives, the attribution of the current road model can be obtained by combining the traction and slip rate of the real-time data point and the normal distribution probability function, and finally the attachment coefficient of the road surface model is obtained through a Kalman filter.

Description

Road surface adhesion coefficient estimation method based on Bayes and electronic equipment

Technical Field

The invention belongs to the technical field of road adhesion coefficients, and particularly relates to a road adhesion coefficient estimation method based on Bayes and an electronic device.

Background

At present, the number of automobiles is increasing dramatically, and the safety performance becomes the most important consideration of the automobiles. In the more and more mature field of automatic driving, there are not only great requirements for the safety performance of automobiles, but also very strict requirements for road conditions, such as the roughness of the road, the adhesion coefficient of the road, and other parameters, especially the adhesion coefficient. The braking distance of the automobile under different adhesion road surfaces is different, and the braking distance is longer under the road surface with smaller adhesion coefficient, so when the automobile runs on the road surface with different adhesion coefficients for switching, especially in the field of automatic driving, the faster identification switching and the identification of the road surface to which adhesion coefficient the automobile is switched are particularly important.

In the existing sensor-based road surface classification recognition and switching system, direct recognition equipment of a sensor is limited by conditions such as installation environment, use environment, high price and the like, the recognition and switching effect on the road surface is not obvious in some special cases (such as severe weather such as heavy fog and heavy snow) and the switching speed does not meet the actual requirement.

However, in a general algorithm, if the data characteristic judgment is performed on the sampling point in a window manner by the least square method, the real-time requirement of the high-speed driving of the automobile on the road surface switching identification is very high, and the actual requirement of the automobile during the driving can not be met. Other algorithms, such as fast filtering by a kalman filter, can meet the real-time requirement, and can obtain a result in the first time when a vehicle switches a road surface, but can only judge that the road surface has switched, but cannot calculate the adhesion coefficient of the switched road surface with the required precision, which cannot meet the actual requirement of the automatic driving field with strict requirements on the road surface condition.

Disclosure of Invention

Aiming at the problems in the prior art, the invention provides a road adhesion coefficient estimation method and electronic equipment based on Bayes, which are used for rapidly identifying the switching of the road and estimating the road adhesion coefficient.

In order to achieve the above object, the present invention provides a road adhesion coefficient estimation method based on Bayes, comprising the steps of:

obtaining a set of data points comprising slip rate and tractive effort;

dividing the tractive effort into a plurality of subintervals for dividing the set of data points into a plurality of data subsets; in the same data subset, the slip rates of the road surface models with different attachment coefficients are subjected to different normal distributions;

in each data subset, calculating expectation and variance of normal distribution of the road surface model of each attachment coefficient, and constructing a normal distribution probability function of the road surface model of each attachment coefficient according to the calculated expectation and variance;

collecting real-time data points, and selecting a corresponding data subset according to the traction force of the real-time data points;

calculating the probability of the real-time data points belonging to the pavement model of each adhesion coefficient according to the slip rate of the real-time data points and the normal distribution probability function of the pavement model of each adhesion coefficient in the selected data subset;

determining a road surface model to which the real-time data point belongs according to the maximum attribution probability;

and calculating the adhesion coefficient of the current road surface model through a Kalman filter.

In some alternative embodiments, the slip ratio is the slip ratio of the left or right side of the vehicle during travel of the vehicle.

In some alternative embodiments, the slip ratio calculation formula for the left or right side of the vehicle is as follows:

wherein S represents a slip ratio on one side of the vehicle, W₁And W₂Respectively representing the angular velocity, R, of the driving and driven wheels on one side of the vehicle₁And R₂Respectively, the tire radii of a driving wheel and a driven wheel on one side of the vehicle, wherein: one side is the left side or the right side.

In some alternative embodiments, the subintervals of tractive effort are the same length.

In some optional embodiments, after determining the road surface model to which the real-time data point belongs, determining whether switching of the road surface model occurs; and if so, calculating the road surface model to which each point belongs in the next collected multiple data points, counting the number of the data points corresponding to each road surface model, and taking the road surface model with the maximum number of the data points as the current road surface model.

In some alternative embodiments, the plurality of data points is 22.

An electronic device comprising one or more processors and memory;

one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the above-described road adhesion coefficient estimation method based on Bayes.

Compared with the prior art, the invention has the following advantages and beneficial effects:

the method determines the attribution of the road model through the collection of data points of the traction force of a certain side and the automobile slip rate when the automobile runs, and accurately calculates the adhesion coefficient of the road model. The invention does not take the integral data point set as the total sample set for learning, but divides the integral data point set into a plurality of data subsets, the point sets of the road surfaces with different attachment coefficients in each data subset present the trend of normal distribution, the expectation and the variance of the normal distribution of the point sets of the road surfaces with different attachment coefficients are respectively different, and the road surfaces with different attachment coefficients can be distinguished by learning the difference; by constructing a normal distribution probability function of the road surface model of each adhesion coefficient in each data subset, when a real-time data point arrives, the time for road surface switching can be calculated and the point set attribution result after road surface switching can be obtained by combining the traction force and the slip rate of the real-time data point and the normal distribution probability function, so that the attribution of the road model is determined, and the adhesion coefficient of the road surface model is obtained.

Drawings

Fig. 1 is a flowchart of a road surface adhesion coefficient estimation method based on Bayes according to an embodiment of the present invention.

FIG. 2 is a data point set plot of slip rate versus tractive effort provided by an embodiment of the present invention;

FIG. 3 is a time chart of the vehicle switching from a high adhesion coefficient road surface to a low adhesion coefficient road surface according to the embodiment of the present invention;

FIG. 4 is a normal distribution plot of slip rates for different subsets of data for a fixed tractive effort range provided by an embodiment of the present invention;

fig. 5 is a kalman calculation result diagram of a high adhesion coefficient road surface to a low adhesion coefficient road surface provided by the embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.

The invention mainly relates to how to process a sample set to obtain prior probability, how to quickly calculate the posterior probability and identify the switching of road surfaces by a Bayes classification method when a new data set arrives. How to calculate a road adhesion coefficient with a desired accuracy by controlling the convergence speed of the kalman filter when a road surface switch occurs.

The road surface adhesion coefficient estimation method based on Bayes in the embodiment of the invention, as shown in FIG. 1, comprises the following steps:

s1, acquiring a set of data points including slip ratio and traction.

The Bayes classification of the present invention is based on the fact that in a data subset, when the range of traction is fixed, the slip ratio of the data set is characterized differently on the road surface with different adhesion coefficients. The road surface slip rates of different adhesion coefficients can be seen from the point diagram of the data set, and the operation performance of the algorithm can be well shown due to the great difference between the high adhesion road surface and the low adhesion road surface.

A Bayes classifier based road adhesion coefficient estimation algorithm relies on a data set of slip ratio (S) and traction (F) of a vehicle on the left (XL) or right (XR) side of the vehicle during driving_f) The principle of the traction coefficient estimation is calculated from the difference in the speeds of the driven and non-driven wheels and the friction correlation in slip.

Counting with front wheel as driving wheelThe slip ratio on the left side is for example: wherein S_XLLeft side slip ratio, ω_FLω_RLRespectively the left front wheel angular velocity and the left rear wheel angular velocity, R_FLR_RLThe radius of the tire of the left front wheel and the left rear wheel, respectively, S_XLThe calculation formula of (2) is as follows:

the Bayes formula is applied to the data set of the invention and is a precondition of Bayes classification, in the Bayes prior probability calculation, the prior probability learning of the data set subset is based on the fact that the road surfaces with different attachment coefficients present normal distribution with different expectation and variance, the obtained expectation and variance of the normal distribution of each attachment coefficient of each subset are used as prior probability, and the new data point calculates the posterior probability according to the conditional probability, thereby obtaining the new road surface model attribution of the data point.

As shown in FIG. 2, the slip ratio and traction force point chart of the automobile running on the road surface with different adhesion coefficients is shown, wherein the vertical axis is the traction force (F)_f) The horizontal axis represents the slip ratio S. And the experimental conditions are a data point set from a dry asphalt pavement (high adhesion coefficient) to a gravel pavement (low adhesion coefficient) of an automobile, "· represents the dry asphalt pavement, and" × "represents the gravel pavement. It can be seen from the figure that the road surfaces with different traction coefficients show a significant difference in slip ratio for the same traction range. Due to the fact that the difference of the adhesion coefficients of the two road surfaces is large, classification switching can be achieved by adopting a Kalman filtering method. However, when the adhesion coefficients of two kinds of road surfaces are close to each other, such as a dry asphalt road surface and a ponding asphalt road surface, the effect of classification switching is poor only by adopting the Kalman filter, and the time for obtaining a data result is greatly increased, so that the requirements of a real system cannot be met.

According to the method, the road surfaces with different known adhesion coefficients are learned to obtain the prior probability, then the thought of a Bayes classifier is added to classify the subsequent data sets, the posterior probability is calculated through the prior probability to judge which road surface model the prior data sets belong to, and then the adhesion coefficients of the road surface model can be accurately calculated through a Kalman filter, so that the switching identification of the road surfaces in the driving process of the automobile and the calculation of the adhesion coefficients of the switched road surface model can be realized.

S2, dividing the traction force into a plurality of subintervals for dividing the data point set into a plurality of data subsets; in the same data subset, the slip rates of the road surface models with different adhesion coefficients are subjected to different normal distributions.

In the invention, all data sets are not taken as an integral learning sample, but the data sets of traction force and slip rate are divided into data subsets taking traction force as a certain range, each subset is taken as a learning sample, the expectation and variance of normal distribution of the data sets of the road surface with different attachment coefficients in each subset have obvious difference, and the prior probability of different data sets caused by the difference is learned.

In the experiment, the traction force range is set at 50 intervals, and the data sets of the road surfaces with different adhesion coefficients are learned in each interval. A normal expression histogram of different adhesion coefficient road surfaces in the same traction range is drawn through a large number of data experiments, and when the traction range is fixed, the slip rates of the different adhesion coefficient road surfaces are not only obvious in representation, but also show a normal distribution trend, as shown in FIG. 4. This is also the theoretical basis for road surface friction coefficient estimation algorithms based on Bayes classification.

S3, in each data subset, calculating expectation and variance of normal distribution of the road surface model of each attachment coefficient, and constructing a normal distribution probability function of the road surface model of each attachment coefficient according to the expectation and variance obtained through calculation;

as described above, the distribution of the road surface data sets of different adhesion coefficients in the same traction range is normally distributed, so the probability of each point in different road models is actually the probability of finding a normal distribution of the known expectation and variance.

Constructing a normal distribution probability function of the road surface model of each adhesion coefficient in each data subset:

in the formula, x is the slip rate s of the real-time data points, the mean value and the variance are different traction force ranges, the expectation and the variance of different road surfaces are different, and one road surface is provided for each adhesion coefficient under the same traction force.

And S4, collecting real-time data points, and selecting a corresponding data subset according to the traction force of the real-time data points.

And when a real-time data point arrives, selecting the normal distribution probability function of the road surface model with different attachment coefficients in the corresponding traction range according to the traction of the real-time data point, and further calculating the attribution probability.

As can be seen from fig. 4, the expectation and variance of the normal distribution of the road surfaces with different adhesion coefficients are significantly different, and by learning the normal distribution, a road surface model in a fixed traction range can be obtained, when a new data point arrives, a traction range section to which the new data point belongs is determined according to the traction, and the posterior probability of the new point set in each road model is calculated in the section. When the probability of belonging to a certain model is the maximum, the new point can be considered to belong to the road model. The calculation of the posterior probability has been described in the embodiments, and it is actually the probability of the normal distribution for each new data point, and the expectation of the normal distribution and the variance of the road surface with different adhesion coefficients are known, so the calculation of the posterior probability is also reliable.

S5, calculating the probability that the real-time data points belong to the pavement model with each adhesion coefficient according to the slip rate of the real-time data points and by combining the normal distribution probability function of the pavement model with each adhesion coefficient in the selected data subset;

the calculation of the prior probability and the posterior probability of Bayes probability is carried out by first dividing the learning set into a plurality of subsets of the total data set, which is the traction force of the vehicle (F), according to the learning set mentioned in the summary of the invention_f) To the left or right of the carI.e. the abscissa of each data point set is the slip rate and the ordinate is the traction.

The new point is denoted A and the different tractive effort ranges are denoted F₁、F₂、F₃.., the road surfaces with different adhesion coefficients in the same traction range are marked as B₁、B₂、B₃.., according to Bayes' equation, at F₁Respective probabilities under tractive effort range:

P₁＝P(A|B₁)

P₂＝P(A|B₂)

P₃＝P(A|B₃)

in other words, in the subset under the same traction force range, the probability of the new data point under the traction force range can be calculated through Bayes conditional probability, and the road model attribution of the new data point is C, then:

C＝max(P₁,P₂,P₃)

in this way, it is possible to determine which road model the new data point belongs to. In a specific probability calculation, the conditional probability can be obtained by Bayes formula as follows:

in the present invention, the probability of road appearance with different adhesion coefficients is independent of the new data point, so the formula can be modified as follows:

the normal distribution probabilities of the road surfaces with different attachment coefficients are different, so that the calculated attribution probabilities of the same data point under the road surfaces with different attachment coefficients are different.

S6, determining the road surface model to which the real-time data point belongs according to the maximum attribution probability;

in the present invention, it is assumed that the probability of occurrence in each tractive effort interval is equal when a new data point arrives. The maximum value of the calculated posterior probability exists in a road model in one traction area, and the data point at the moment is considered to be in the road model.

S7, judging whether the road surface model is switched; and if so, calculating the road surface model to which each point belongs in the next collected multiple data points, counting the number of the data points corresponding to each road surface model, and taking the road surface model with the maximum number of the data points as the current road surface model.

When determining which road model the new data point belongs to, it is far from enough to rely on only one point to decide whether the vehicle has switched to another road, and this way has a low fault tolerance. In actual automobile driving, there is a case that an adhesion coefficient is suddenly changed due to loss of the road surface or due to some foreign objects (for example, liquid splashed on the road) on a road surface with a relatively constant adhesion coefficient, and if a road surface model is judged by data points when an automobile passes through the place, erroneous judgment is generated, and a judgment result of the whole system is influenced. The solution to this phenomenon is to add a logic for determining the start of a road switch.

When the new data points are calculated through posterior probability, and the judgment result is that the point is not in the road model of the previous point, namely the road model is likely to be switched, the judgment logic of starting switching of the road surface is activated, and after the fact that the road surface is switched is judged through the new data points, the attachment coefficient of the switched road model can be quickly and accurately calculated through adjusting a Kalman filtering measurement noise covariance matrix.

The invention makes a strict test on the judgment logic for starting switching of the activated road surface, and finally takes 22 continuous new points after the activation judgment logic as the basis for judging attribution. The specific contents of this logic are: the probability of 22 continuous points after activating the judgment logic is obtained by Bayes calculation posterior probability and is recorded asOn the surface A of a road, there is A₁A, belonging to the road surface B with B₁A, belonging to the group C on the road surface C₁And (3 kinds of road surfaces are taken as an illustration here). Noting the probability of belonging to road surface A as P_AProbability of belonging to road surface B is P_BProbability of belonging to road surface C is P_CIt is possible to obtain:

A₁+B₁+C₁＝22

P＝max(P(A₁),P(B₁),P(C₁))

the maximum probability of 22 points of the road surface belonging to various adhesion coefficients in the traction range can be obtained, the automobile is considered to run on the road model at the moment, if the calculated road model is different from the previous road model at the moment, the road surface is considered to be switched, and then the adhesion coefficients of the switched road model can be rapidly and accurately calculated by adjusting a Kalman filtering measurement noise covariance matrix.

As shown in fig. 3, the time performance of the vehicle when switching from the high adhesion coefficient road to the low adhesion coefficient road is shown, 0 represents that the vehicle is on a dry asphalt road, 1 represents that the vehicle is on a gravel road, the vehicle starts to enter the gravel road when the 1670 th data point arrives, when the road switching is identified, the derived data point can be regarded as when the 1690 th data point arrives, the speed interval is 20 data points, the period of the system for collecting data points is 10ms, and the time for identifying the road switching is 0.2s, which can meet the time requirement of the actual system.

And S8, calculating the adhesion coefficient of the current road surface model through a Kalman filter.

The Kalman filtering based on the invention is an algorithm which utilizes a linear system state equation and carries out optimal estimation on the system state through inputting and outputting observation data of the system. The optimal estimation can also be seen as a filtering process, since the observed data includes the effects of noise and interference in the system.

The Kalman filtering algorithm of the invention is mainly divided into two steps: and (4) predicting and updating.

And (3) prediction: estimating the state of the current moment (k moment) according to the posterior estimation value of the last moment (k-1 moment) to obtain the prior estimation value of the k moment;

updating: and correcting the estimated value of the prediction stage by using the measured value of the current moment to obtain the posterior estimated value of the current moment.

The kalman filter can be divided into a time update equation and a measurement update equation. A time updating equation (namely a prediction stage) is used for calculating a state variable prior estimation value and an error covariance prior estimation value at the current moment according to the state estimation value at the previous moment; the measurement update equation (i.e., the update phase) is responsible for combining the a priori estimates and the new measured variables to construct an improved a posteriori estimate. The time update equation and the measurement update equation are also referred to as a prediction equation and a correction equation. The kalman algorithm is therefore a recursive prediction-correction method.

The kalman filter time update equation is as follows:

the kalman filter state update equation is as follows:

the meaning of each parameter in the kalman filter equation is as follows:

1、

and

the a posteriori state estimates, representing time k-1 and time k, respectively, are one of the results of the filtering, i.e. the updated result, also called the optimal estimate. In the present embodiment

And

respectively representing the optimal estimated value of the road adhesion coefficient at the previous moment and the optimal estimated value of the road adhesion coefficient at the current moment.

2、

The prior state estimate at time k is the intermediate result of the filtering, i.e., the predicted k-time based on the optimal estimate at the previous time (time k-1), and is the result of the prediction equation. In the present embodiment

Representing the optimal estimation of the road adhesion coefficient by the last instant

And obtaining the predicted value of the road adhesion coefficient at the current moment.

3、P_k-1And P_k: representing the posteriori estimated covariance at time k-1 and k, respectively (i.e., the k-time

And

represents the uncertainty of the state) is one of the results of the filtering. In this embodiment P_k-1And P_kRoad surface adhesion coefficient estimation value representing last time

Covariance of (2) and road surface adhesion coefficient estimation value at current time

The covariance of (a).

4、

A priori estimated covariance at time k: (

Covariance of (d) is the intermediate calculation result of the filtering. In the present embodiment

And a covariance showing the predicted value of the road adhesion coefficient at the current time.

5. H: the Kalman filter is a linear relation, is responsible for converting a measured value of m dimensions into n dimensions, conforms to the mathematical form of the state variable and is one of the preconditions of the filtering.

The traction and slip ratio in this embodiment approximately satisfy the following linear relationship:

where the subscripts l and r represent the left and right sides of the vehicle, s (t) represents the slip ratio, and u (t) represents the normalized tractive effort.

And

respectively representing the slopes, delta, of the left and right sides of the linear model_l(t) and δ_r(t) represents intercepts on the left and right sides of the linear model, respectively. The two equations above are written in vector form, with the following transformations:

h (t) can be obtained from the above formula:

6、z_k: the measured value (observed value), is the input to the filtering. The measured value is represented by a slip rate in the present embodiment.

7、K_k: the filter gain matrix is a filtered intermediate calculation result, a Kalman gain, or a Kalman coefficient. By measuring value z_k(slip ratio) pair

(road surface adhesion coefficient predicted value) is corrected.

8. A: the state transition matrix is actually a guessing model for the target state transition. In the present embodiment, it is assumed that the road surface adhesion coefficient is substantially constant on the same road surface, i.e., the road surface adhesion coefficient at the present time is equal to the road surface adhesion coefficient at the previous time.

9. Q: process excitation noise covariance (covariance of the systematic process). This parameter is used to represent the error between the state transition matrix and the actual process, i.e. the noise introduced by the prediction model itself.

10. R: the noise covariance is measured. When the filter is actually implemented, the measured noise covariance R is typically observed and is a known condition of the filter. In this embodiment, the convergence rate of the kalman filter is controlled by dynamically adjusting the size of R according to the Bayes classification result.

12、

The residuals of the actual observation and the predicted observation are corrected a priori (predicted) together with the kalman gain to obtain the posterior. The estimated value of the road adhesion coefficient is corrected by this residual error.

The Kalman filter can realize accurate calculation of the road adhesion coefficient.

As shown in fig. 5, this figure depicts a kalman calculated result graph of a high-adhesion-coefficient road surface, which is a dry asphalt road surface and has an actual adhesion coefficient of about 0.8, to a low-adhesion-coefficient road surface, which is a gravel road surface and has an actual adhesion coefficient of about 0.4. As can be seen from the figure, the road adhesion coefficient initially stabilizes at 200, and after the road surface switching occurs, the calculated adhesion coefficient of the new road surface model is 95, where the road adhesion coefficient is obtained by amplifying the road adhesion coefficient by a certain magnification in the algorithm.

The classification effect and the classification time of the road surface friction coefficient estimation algorithm based on Bayes classification are both dependent on the learning condition of the slip rate under the traction force in the range, namely, the prior probability is depended on, the thinner the road model is, the more data sets are, and the more accurate the posterior probability obtained through the prior probability is. The invention needs to rely on the prior learning of data sets of different adhesion coefficient road surfaces, and then calculates the prior probability in each data set for storage.

The invention realizes the switching identification of the road surface meeting the actual real-time requirement, and can have an identification attribution function with satisfactory precision for the switched road surface. The algorithm is not influenced by factors such as environment and the like, so that the method has good compatibility and workability. In addition, the Bayes classification based on the algorithm has the advantages that the calculation result of the posterior probability is related to the prior probability, the sample set of the prior probability is enough, the classification is accurate enough, the reliability of the result obtained by the Bayes formula is very high, and the method has great significance to the field of automobile driving, particularly the field of automatic driving.

Finally, the invention also provides an electronic device, which is characterized by comprising one or more processors and a memory; one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the above-described road adhesion coefficient estimation method based on Bayes.

It should be noted that, according to the implementation requirement, each step/component described in the present application can be divided into more steps/components, and two or more steps/components or partial operations of the steps/components can be combined into new steps/components to achieve the purpose of the present invention.

It will be understood by those skilled in the art that the foregoing is merely a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included within the scope of the present invention.

Claims

1. A road surface adhesion coefficient estimation method based on Bayes is characterized by comprising the following steps:

obtaining a set of data points comprising slip rate and tractive effort;

2. The Bayes-based road adhesion coefficient estimation method according to claim 1, wherein the slip ratio is a slip ratio of a left side or a right side of the automobile during running of the automobile.

3. The Bayes-based road adhesion coefficient estimation method according to claim 2, wherein a slip ratio calculation formula of a left side or a right side of the vehicle is as follows:

4. The Bayes-based road adhesion coefficient estimation method according to claim 1, wherein the subintervals of tractive effort are of the same length.

5. The Bayes-based road adhesion coefficient estimation method according to claim 1, wherein after determining a road surface model to which the real-time data points belong, it is determined whether the road surface model is switched; and if so, calculating the road surface model to which each point belongs in the next collected multiple data points, counting the number of the data points corresponding to each road surface model, and taking the road surface model with the maximum number of the data points as the current road surface model.

6. The Bayes-based road adhesion coefficient estimation method as recited in claim 5, wherein the plurality of data points is 22.

7. An electronic device comprising one or more processors and memory;

one or more programs stored in the memory and configured to be executed by the one or more processors, the one or more programs configured to perform the Bayes-based road adhesion coefficient estimation method of any of claims 1-6.