CN113619587B - Road adhesion coefficient estimation method based on Bayes classifier - Google Patents
Road adhesion coefficient estimation method based on Bayes classifier Download PDFInfo
- Publication number
- CN113619587B CN113619587B CN202110205041.4A CN202110205041A CN113619587B CN 113619587 B CN113619587 B CN 113619587B CN 202110205041 A CN202110205041 A CN 202110205041A CN 113619587 B CN113619587 B CN 113619587B
- Authority
- CN
- China
- Prior art keywords
- road
- data
- road surface
- traction
- adhesion coefficient
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
- B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
- B60W40/02—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models related to ambient conditions
- B60W40/06—Road conditions
- B60W40/064—Degree of grip
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Automation & Control Theory (AREA)
- Mathematical Physics (AREA)
- Transportation (AREA)
- Mechanical Engineering (AREA)
- Traffic Control Systems (AREA)
- Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
Abstract
The invention discloses a road surface adhesion coefficient estimation method based on a Bayes classifier, which is based on the theory of the Bayes classifier, determines the attribution of a road model through a data point set of traction force and automobile slip rate of a certain side when an automobile runs, divides the collected integral data point set into a plurality of data subsets according to the traction force range, analyzes newly collected data points, calculates the posterior probability after obtaining the prior probability, determines the attribution of the road model, and finally accurately calculates the adhesion coefficient of the road model according to the new road model. Through the method, the Bayes subset prior probability is learned, the posterior probability of each traction subset interval is calculated when a real-time data point arrives, the time of road surface switching can be calculated through the method, the point set attribution result after the road surface switching can be obtained, the attribution of a road model is determined, and the adhesion coefficient of the road model is obtained.
Description
Technical Field
The invention relates to the field of road adhesion coefficient calculation, in particular to a road adhesion coefficient estimation method based on a Bayes classifier.
Background
At present, the number of automobiles in China is increased greatly, and the safety performance becomes the most important consideration of the automobiles. In the more and more mature field of autonomous driving, the identification of the road adhesion coefficient is increasingly important. The braking distance of the automobile under different adhesion road surfaces is different, the braking distance under the road surface with smaller adhesion coefficient is lengthened, and the stability of the automobile is not easy to control, so that when the adhesion coefficient of the road surface is switched, the change of the adhesion coefficient is recognized more quickly, which is particularly important.
In the existing sensor-based road surface classification identification and switching system, since complicated sensor equipment (a camera, a strain sensor, an acoustic sensor and the like) is limited by conditions such as installation environment, use environment, price and the like, and in some special cases (for example, severe weather such as heavy fog and heavy snow and the like), the identification switching effect on the road surface is not obvious, and the switching speed does not meet the actual requirement.
In the prior art, various algorithms are adopted to judge the road model, and a general algorithm, such as a least square method, is used for judging data characteristics of a sampling point in a window mode, and because the real-time requirement of high-speed running of an automobile on road surface switching identification is very high, the actual requirement of the automobile in running can not be met; other algorithms, such as fast filtering by a kalman filter, can meet the real-time requirement, can identify the adhesion coefficient transition when the vehicle switches the road surface, but the convergence speed is greatly influenced by the road surface condition, and cannot meet the actual requirements in the field of automatic driving with very strict requirements on the road surface condition.
Therefore, it is necessary to design a road adhesion coefficient estimation method based on a Bayes classifier, which satisfies the switching recognition of the road surface required for real-time performance and can make the road adhesion coefficient estimation meet the design accuracy requirement by adjusting the parameters.
Disclosure of Invention
In order to solve the problems that the real-time performance of the identification of the existing road surface switching identification system is poor, and the convergence speed of the algorithm is greatly influenced by the road surface condition to cause inaccurate calculation data, the invention provides a road surface adhesion coefficient estimation method based on a Bayes classifier.
In order to achieve the purpose, the invention adopts the technical scheme that:
a road adhesion coefficient estimation method based on a Bayes classifier comprises the following steps:
s1, collecting traction force data point sets of automobiles running on road models with different attachment coefficients, and dividing the data point sets into a plurality of data subsets according to the traction force corresponding to the data point sets;
s2, calculating expectation and variance of a data set obtained by the data subset under the same traction force on road models with different attachment coefficients, and further verifying data characteristics of the data set under the road models with different attachment coefficients;
s3, analyzing and processing the different data subsets to obtain the prior probability of the data set of the data subsets in each traction range on the road surface models with different adhesion coefficients;
s4, collecting new data points under the road surface, judging the traction range to which the new data points belong, and then calculating the posterior probability of the data points on the road surface models with different attachment coefficients according to the prior probability of the data subsets corresponding to the traction range under the road models with different attachment coefficients;
s5, sorting the posterior probabilities of the data points under all the road models, and comparing the posterior probabilities, wherein the road model to which the maximum value in the posterior probabilities belongs is the source road surface of the data points;
s6, judging whether a road surface switching judgment logic is activated or not, and judging that the automobile runs on the road surface under the road model at the moment when data points belonging to the same road model among a plurality of continuously collected data points reach a preset threshold value;
and S7, obtaining the attachment coefficient of the switched road model by adopting an extended Kalman filtering algorithm according to the determined road model.
Further, S1 specifically is: the method comprises the steps of collecting traction force data point sets of automobiles running on road models with different attachment coefficients, dividing the obtained traction force and slip rate data point sets into data subsets with traction force as a certain range, and taking each data subset as an analysis object.
Further, the step S2 is specifically to judge whether the data set exhibits normal distribution according to expectation and variance of the data set obtained by the data subset under the same traction force on the road model with different attachment coefficients, so as to be suitable for the extended kalman filter algorithm.
Further, the data set is obtained by analyzing and calculating data subsets of the slip ratio S of the left side XL or the right side XR of the automobile and the traction force F of the automobile during the driving process of the automobile, and the corresponding slip ratio is calculated according to the speed difference of the driving wheel and the non-driving wheel and the friction correlation in the slip:
in the above formula, S X Left side slip ratio, ω F ，ω R Respectively the left front wheel angular velocity and the left rear wheel angular velocity, R F ，R R The tire radii of the left front wheel and the left rear wheel are respectively.
Further, the S3 specifically is: and analyzing the prior probability of the data set on the basis of normal distributions with different expectations and variances of the data set on the road surface with different attachment coefficients, and taking the expectation and the variance of the obtained normal distribution under each attachment coefficient of each data subset as the prior probability.
Further, S4 specifically is: the prior probability belongs to a plurality of intervals which are divided according to the range of the traction force, data sets of different intervals are analyzed, the posterior probability is calculated in a corresponding model in a table look-up mode according to the obtained prior probability, A represents a newly collected data point, B represents a road surface model, and then the posterior probability P (B | A) is obtained.
Further, for the newly acquired data point set A, the different tractive effort ranges are denoted as F 1 、F 2 、F 3 .., marking the road surface with different adhesion coefficient under the same traction force range as B 1 、B 2 、B 3 .., according to Bayes' formula, at F 1 The respective probabilities in the traction range are:
P 1 ＝P(A|B 1 )
P 2 ＝P(A|B 2 )
P 3 ＝P(A|B 3 )
further, it is preset that when a newly acquired data point arrives, the probability of occurrence in each traction interval is equal, and then the P is compared 1 、P 2 、P 3 And (3) judging that the data point at the moment is positioned in the road model corresponding to the traction force section according to the maximum value of the calculated posterior probability.
Further, S6 specifically is: the preset judgment condition for starting switching of the road surface is as follows: and the road model of the posterior probability obtained by calculating the data points acquired in real time in Bayes classification is changed compared with the road model of the data points acquired last time, and at the moment, the road surface switching judgment logic is activated, and the change of the road model to which the road surface belongs is confirmed.
Further, S7 is specifically to implement fast convergence of the kalman filter and accurate calculation of the road adhesion coefficient after switching by dynamically adjusting the measurement noise covariance matrix of the extended kalman filter algorithm, and define a nonlinear system equation of the traction force and the slip ratio as:
wherein, U is traction force, S is slip ratio, and (a, b and c) are waiting coefficients; the above formula is substituted into the extended Kalman filtering algorithm, and a system model of the nonlinear system can be established:
X(k+1)＝f(x(k))+e(k)
Z(k)＝h(x(k))+U(k)
wherein k is the number of iterations, k +1 is the number of next iterations,
wherein X (k) = [ a, b, c ]; z (k) is equal to U, X (k) represents an estimated value of the parameters a, b and c in the current iteration period, X (k + 1) represents that the k +1 th iteration is an estimated value of the parameters a, b and c, f (X (k)) represents the relative conversion relation of the parameters in the adjacent period, e (k) is the system error of the parameters, Z (k) is an estimated value of the k-th period to the traction force U, U (k) represents a measured value of the traction force in the k-th period, h (X (k)) represents the relative conversion relation between the parameters a, b and c and the traction force, namely:
where S is a measurement of the kth cycle slip rate.
Compared with the prior art, the invention has the beneficial effects that:
1. the invention discloses a road adhesion coefficient estimation method based on a Bayes classifier, which is characterized in that an integral data point set is not taken as a total learning sample set when the prior probability is calculated, but the integral data point set is divided into a plurality of data subsets, the classification condition of the data subsets is a traction force range of a longitudinal axis of the data points, and the estimation value of the road adhesion coefficient is obtained through the correlation characteristic of the traction force and the slip rate of the data points.
2. According to the pavement adhesion coefficient estimation method based on the Bayes classifier, the switching identification of the pavement meeting the actual real-time requirement is realized through the Bayes classifier-based theory, the pavement adhesion coefficient estimation can meet the design precision requirement through parameter adjustment, and the algorithm is not influenced by factors such as the environment, so that the method has good smoothness and high online time; in addition, based on the Bayes classification algorithm, because 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, and the result obtained by the Bayes formula has high reliability.
3. According to the road adhesion coefficient estimation method based on the Bayes classifier, the extended Kalman filtering algorithm is adopted, and when the road surface is switched, the fast convergence of a Kalman filter and the accurate calculation of the switched road adhesion coefficient can be realized by dynamically adjusting the measurement noise covariance matrix of the extended Kalman filtering algorithm.
Drawings
FIG. 1 is a flow chart of the method of the present invention;
FIG. 2 is a graph illustrating slip rate versus tractive effort curves in accordance with the present invention;
FIG. 3 is a diagram of an extended Kalman filter simulation result of the present invention;
FIG. 4 is a schematic diagram of a set of switching actual data points of the left slip ratio and the traction of the vehicle on two road surfaces with different road surface adhesion coefficients (the vertical axis is the traction, and the horizontal axis is the slip ratio);
FIG. 5 is a schematic view showing the result of the road surface switching when the automobile of the present invention is driven from a high adhesion coefficient road surface to a low adhesion coefficient road surface;
FIG. 6 is a schematic diagram showing the time variation of the automobile when switching from a high adhesion coefficient road surface to a low adhesion coefficient road surface;
FIG. 7 is a normal distribution plot of data set slip rate for a fixed tractive effort range of the present invention;
FIG. 8 is a graph of the Kalman calculation results for an automobile from a high adhesion coefficient road surface to a low adhesion coefficient road surface in accordance with the present invention;
FIG. 9 is a schematic diagram illustrating the adhesion coefficient variation during the road surface switching process of the automobile in actual driving;
Detailed Description
The following detailed description of the preferred embodiments of the present invention, taken in conjunction with the accompanying drawings, will make the advantages and features of the present invention more comprehensible to those skilled in the art, and will thus provide a clear and concise definition of the scope of the present invention. It should be apparent that the described embodiments are only some of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without any inventive step, are within the scope of the present invention.
Example 1
As shown in fig. 1, a flow chart of a road adhesion coefficient estimation method based on a Bayes classifier comprises the following steps:
s1, collecting traction force data point sets of automobiles running on road models with different attachment coefficients, and dividing the data point sets into a plurality of data subsets according to the traction force corresponding to the data point sets;
s2, calculating expectation and variance of a data set obtained by the data subset under the same traction force on road models with different attachment coefficients, and further verifying data characteristics of the data set under the road models with different attachment coefficients;
s3, analyzing and processing the different data subsets to obtain the prior probability of the data set of the data subsets in each traction range on the road surface models with different adhesion coefficients;
s4, collecting new data points under the road surface, judging the traction range to which the new data points belong, and then calculating the posterior probability of the data points on the road surface models with different attachment coefficients according to the prior probability of the data subsets corresponding to the traction range under the road models with different attachment coefficients;
s5, sorting the posterior probabilities of the data points under all the road models, and comparing the posterior probabilities, wherein the road model to which the maximum value in the posterior probabilities belongs is the source road surface of the data points;
s6, judging whether a road surface switching judgment logic is activated or not, and judging that the automobile runs on the road surface under the road model at the moment when data points belonging to the same road model among a plurality of continuously collected data points reach a preset threshold value;
and S7, obtaining the attachment coefficient of the switched road model by adopting an extended Kalman filtering algorithm according to the determined road model.
In the embodiment, the invention does not take the whole data point set as the total sample set of learning when calculating the prior probability, but divides the whole data point set into a plurality of data subsets, the classification condition of the data subsets is the traction force range of the longitudinal axis of the data points, and the estimated value of the road adhesion coefficient is obtained through the correlation characteristic of the traction force and the slip ratio of the data points, because the characteristic of the traction force is not obvious in the road relation performance, the data subsets obtained by limiting the traction force range exclude the influence of the traction force, and concentrate more on the distribution of the slip ratio on the distribution of the data point set.
The road adhesion coefficient estimation method based on the Bayes classifier relies on a data set which is a data subset of slip rate (S) and traction force (F) of the left side (XL) or the right side (XR) of an automobile in the driving process of the automobile, and corresponding slip rates are calculated according to the difference of speeds of a driving wheel and a non-driving wheel and the friction correlation in slip:
in the above formula, S X Left side slip ratio, ω F ，ω R Respectively, the angular velocity of the left front wheel and the angular velocity of the left rear wheel, R F ，R R The tire radii of the left front wheel and the left rear wheel are respectively.
In the embodiment, a Bayes formula is applied to the data point set of the invention, which is a precondition of Bayes classification, in the calculation of Bayes prior probability, the analysis of the prior probability of the data set subset is to use the expectation and the variance of the normal distribution of each attachment coefficient of each obtained data subset as prior probability on the basis that the roads with different attachment coefficients present normal distribution with expectation and variance difference, and the newly collected data points calculate posterior probability according to the conditional probability, thereby obtaining the attribute of the road model of the new data points.
As shown in fig. 4, the graph is a data point set from a dry asphalt pavement (with a high adhesion coefficient) to a gravel pavement (with a low adhesion coefficient) of an automobile, a circle graph represents the dry asphalt pavement, and a dot graph represents the gravel pavement, and it can be seen from the graph that under the same traction force range, the slippage rates of the pavements with different adhesion coefficients are obviously different.
In this embodiment, the prior probability and the posterior probability of Bayes probability are calculated by first dividing the sample set into a plurality of data subsets of the total data point set according to the invention, wherein the total data set is the vehicle traction force (F) f ) A point set plane diagram of the slip rate (S) of the left side or the right side of the automobile, namely the abscissa of each data point set is the slip rate, and the ordinate is the traction force;
wherein, the new point set is recorded asA, the different traction ranges are denoted F 1 、F 2 、F 3 .., the road surfaces with different adhesion coefficients in the same traction range are marked as B 1 、B 2 、B 3 .., according to Bayes' formula, at F 1 The respective probabilities in the traction range are:
P 1 ＝P(A|B 1 )
P 2 ＝P(A|B 2 )
P 3 ＝P(A|B 3 )
in the experiment, the traction range is set at 500 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, when the traction range is fixed, the slip rate of the road surfaces with different adhesion coefficients is obvious in representation and presents a normal distribution trend, as shown in FIG. 7, and the normal expression histogram is also a theoretical basis of a road surface friction coefficient estimation algorithm based on Bayes classification.
As described above, the expectation and variance of normal distribution of road surfaces with different adhesion coefficients are obviously different, a road surface model in a fixed traction range can be obtained through analysis of the normal distribution, when a new data point is collected, a traction range interval to which the data point belongs is determined according to the traction of the data point, and the posterior probability of a newly collected data point set under each road model is calculated in the interval; when the probability of belonging to a certain model is the highest, the newly acquired point can be considered to belong to the road model.
In the above description, it has been described in the embodiments that the probability of each new data point is actually the probability of the normal distribution, and the expectation and variance of the normal distribution of the road surfaces with different adhesion coefficients are known, so that the calculation of the posterior probability is also reliable.
FIG. 5 is a schematic diagram of switching between pavements of different adhesion coefficients plotted according to the data of FIG. 4, where in the lower diagram of FIG. 5, 0 represents the vehicle is on a dry asphalt pavement and 1 represents the vehicle is on a gravel pavement; the front section data is dry asphalt pavement, the last section data is gravel pavement, different representations of the two pavements can be seen from the performance of the slip rate, in the switching schematic diagram of the pavement, the adhesion coefficients of the two pavements have large difference, the adhesion coefficient of the dry asphalt pavement can reach 0.8, the adhesion coefficient of the gravel pavement is generally about 0.4, and the system can accurately work when the pavement with large difference of the adhesion coefficients is switched.
In the embodiment, by the method, when a new data point is collected, the probability of the road surface with different adhesion coefficients in the data subset in the same traction range is calculated, and in the first 30s of automobile driving, the probability of the data point on the asphalt road surface can reach 80% or even 90% through the result obtained by the algorithm, and the probability of the data point on the wet and slippery glass road surface is below 20%, so that the data point belongs to the asphalt road surface in the probability calculation of a plurality of continuous points, and the road surface on which the automobile runs is considered to be the dry asphalt road surface at this time; when the automobile runs for about 30s, the probability of the new data point on the wet glass road surface begins to rise through probability calculation, the probability exceeds 50%, and the data point at the moment is considered to be collected on the wet glass road surface in the aspect of probability; after determining which road surface model the new data point is in, it is far from insufficient to decide whether the vehicle has switched to another road surface by only one point, and this way has low fault tolerance, and in the actual vehicle running, there is a case that the road surface with a relatively constant adhesion coefficient has a sudden change of the adhesion coefficient due to the loss of the road surface or due to some foreign objects (such as liquid splashed on the road), if the road surface model is judged by the data point when the vehicle passes through this place, a false judgment is generated, which affects the judgment result of the whole system, and the logic of considering the road surface switching is that when the new data point does not belong to the original road model (in fig. 5, the probability of the data point on the wet and slippery glass is higher than that of the dry asphalt road surface when the vehicle runs for about 30 s), then, of the 22 continuous data points, more than 16 data points belong to the wet and slippery glass road surface through the probability calculation, at this time, the road surface has switched; on the contrary, when the number of the data points is less than 16, the road surface on which the automobile runs at the moment is still in the original road model.
Further explanation is made on a road surface estimation calculation method and a road surface switching result through specific test results: as shown in fig. 9, in order to show the process of the experimental automobile driving from the dry asphalt pavement to the wet glass pavement in the test field, the vertical axis in the figure is the pavement adhesion coefficient, when the experimental automobile is in the dry asphalt pavement, the adhesion coefficient is maintained in a certain higher range, which is consistent with the higher adhesion coefficient of the asphalt pavement; when the experimental automobile runs to a wet and slippery glass road surface, the road surface switching logic starts to work, the magnitude of the adhesion coefficient drops steeply, and the switching time is about 0.3s as can be seen from the figure.
As shown in fig. 6, when the automobile completely runs on a dry asphalt pavement, the pavement is switched, and the data point derived when the pavement is switched can be considered as entering a gravel pavement when 129910 data points arrive, while when the pavement switching is completed, the data point derived can be considered as 130050 data points arrive, the speed interval is 60 data points, the period for collecting the data points according to the system is 10ms, and the time is 0.6s when the pavement is switched, so that the time requirement of the actual system can be met.
As shown in fig. 7, the graph depicts normal distributions of data points for road surfaces with different adhesion coefficients in the same traction interval, i.e. in the same data subset, where the two normal distributions are for road surfaces with two adhesion coefficients that are relatively close to each other; it can be seen that even if two kinds of road surfaces with similar adhesion coefficients exist, the expectation and variance of normal distribution are obviously different, in the prior probability learning process of the system, probability learning of the two kinds of road surfaces is obviously distinguished, for the road surface with larger adhesion coefficient difference, the expectation and variance of normal distribution are more different, and for the prior probability, the calculation of posterior probability is more accurate.
Through the analysis and verification, the road model determined by the calculation method is accurate and reliable in switching.
In this embodiment, the road model in which the data points are located is defined such that the probability of occurrence in each traction segment is equal. If the maximum value of the calculated posterior probability exists in a road model under one traction interval, the data point at the moment is considered to be positioned in the road model;
when a specific implementation mode is adopted, a strict test is carried out on the judgment logic, the number of data points for activating the road surface switching is defined to be 22, the preset threshold value is 16, namely when more than 16 data points in the continuously collected 22 data points belong to the same road model, the automobile is judged to be driven on the road surface under the road model at the moment; and then, the attachment coefficient of the road model after switching can be quickly and accurately calculated by adjusting the noise covariance matrix of the Kalman filtering algorithm.
In this embodiment, the algorithm for calculating the attachment coefficient adopted by the invention is an extended kalman filtering algorithm, and since the traditional linear kalman filtering utilizes the state equation of a linear system and outputs observation data through the input and output of the system, the optimal estimation of the system state is realized, and when the state equation of the system is not a linear gaussian model, the traditional linear kalman filtering algorithm cannot be well applied, so that the method performs the optimal estimation on the relationship between the slip ratio and the traction force based on the extended kalman filtering algorithm.
As shown in fig. 8, which depicts a map of the kalman calculation results from a high-adhesion-coefficient road surface to a low-adhesion-coefficient road surface, it can be seen from the map that the road adhesion coefficient is initially stabilized at 0.8, and after the road surface switching occurs, the adhesion coefficient of the new road surface model calculated is 0.4.
In fig. 8, the vertical axis is an estimated value of the road adhesion coefficient reflected by the traction force and the automobile slip ratio, and the adhesion coefficient reduction caused by the road surface switching is filtered by the extended kalman filter to obtain a more accurate and stable value.
In this embodiment, fig. 2 is a distribution relationship diagram of the slip ratio and the traction force of the vehicle, and it can be seen from the diagram that the relationship does not present a linear distribution, and the observation result is optimized and estimated by establishing a model through an extended kalman filter algorithm:
the nonlinear system equation of the traction force and the slip ratio is as follows:
wherein U is traction force, S is slip ratio, and (a, b, c) are coefficients to be determined. In the extended kalman filter algorithm, the prediction of the state and the prediction of the observed value are calculated from this nonlinear function.
The system model of the nonlinear system is as follows:
X(k+1)＝f(X(k))+e(k) (2)
Z(k)＝h(X(k))+U(k) (3)
wherein X (k) = [ a, b, c]；Z(k)∈U，
Wherein k is the number of iterations, k +1 is the number of next iterations,
wherein X (k) = [ a, b, c ]; z (k) is equal to U, X (k) represents an estimated value of the parameters a, b and c in the current iteration period, X (k + 1) represents that the k +1 th iteration is an estimated value of the parameters a, b and c, f (X (k)) represents the relative conversion relation of the parameters in the adjacent period, e (k) is the system error of the parameters, Z (k) is an estimated value of the k-th period to the traction force U, U (k) represents a measured value of the traction force in the k-th period, h (X (k)) represents the relative conversion relation between the parameters a, b and c and the traction force, namely:
where S is a measurement of the kth cycle slip rate.
The state matrix a is calculated as follows:
in the system:
the calculation of the transition matrix H from equation (3) is as follows:
the optimal estimation result of the extended Kalman filtering on the observation result can be obtained through the state matrix A and the transition matrix H, and the following results are obtained:
the result of using the extended kalman filter algorithm can be obtained by equation (8), equation (9), equation (10), and equation (11):
as shown in fig. 3, the feasibility of using the extended kalman filter algorithm is verified according to the graph, in the graph, the slope distribution of the data point set can be obtained through the extended kalman filter algorithm, and by comparing the maximum value with the slope after filtering, the slope of the data point set is well estimated from the result after processing by the extended kalman filter algorithm, and the influence of noise is also eliminated to a small extent.
Therefore, it can be considered that the extended kalman filter algorithm adopted by the method can realize accurate calculation of the road adhesion coefficient.
According to the mode, firstly, the pavement with different known adhesion coefficients is analyzed to obtain the prior probability, then, the subsequently acquired data set is classified to obtain different data subsets by adopting the Bayes classifier principle, the posterior probability is calculated according to the prior probability of the data subsets to judge which pavement model the newly acquired data set belongs to, then, the adhesion coefficients of the pavement model can be accurately calculated through the extended Kalman filtering algorithm, and therefore switching identification of the pavement in the driving process of an automobile and the adhesion coefficient calculation of the switched pavement model can be achieved; it is worth noting that the classification effect and the classification time of the road surface friction coefficient estimation method based on Bayes classification are both dependent on the analysis condition of the slip rate under the traction force in the range, namely dependent on the prior probability, the thinner the road model is, the more data sets are, and the more accurate the posterior probability obtained through the prior probability is, therefore, the method needs to rely on the prior inductive analysis of the data sets of the road surfaces with different attachment coefficients, and calculate the prior probability in each data set for storage.
The above description is only for the purpose of illustrating the technical solutions of the present invention and is not intended to limit the same; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some or all of the technical features may be equivalently replaced; all the equivalent structures or equivalent processes performed by using the contents of the specification and the drawings of the invention, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.
Claims (10)
1. A road adhesion coefficient estimation method based on a Bayes classifier is characterized by comprising the following steps:
s1, collecting traction data point sets of automobiles running on road models with different attachment coefficients, and dividing the data point sets into a plurality of data subsets according to the traction values corresponding to the data point sets;
s2, calculating expectation and variance of a data set obtained by the data subset under the same traction force on road models with different attachment coefficients, and further verifying data characteristics of the data set under the road models with different attachment coefficients;
s3, analyzing and processing the different data subsets to obtain the prior probability of the data set of the data subsets in each traction range on the road surface models with different adhesion coefficients;
s4, collecting new data points under the road surface, judging the traction range to which the new data points belong, and then calculating the posterior probability of the data points on the road surface models with different attachment coefficients according to the prior probability of the data subsets corresponding to the traction range under the road models with different attachment coefficients;
s5, sorting the posterior probabilities of the data points under all road models, and comparing the posterior probabilities, wherein the road model to which the maximum value in the posterior probabilities belongs is the source road surface of the data points;
s6, judging whether a road surface switching judgment logic is activated or not, and judging that the automobile runs on the road surface under the road model at the moment when data points belonging to the same road model among a plurality of continuously collected data points reach a preset threshold value;
and S7, obtaining the attachment coefficient of the switched road model by adopting an extended Kalman filtering algorithm according to the determined road model.
2. The road surface adhesion coefficient estimation method based on the Bayes classifier as claimed in claim 1, wherein said S1 is specifically: the method comprises the steps of collecting traction force data point sets of automobiles running on road models with different attachment coefficients, dividing the obtained traction force and slip rate data point sets into data subsets with traction force as a certain range, and taking each data subset as an analysis object.
3. The Bayes classifier based road adhesion coefficient estimation method according to claim 1, wherein S2 is specifically configured to determine whether the data set exhibits normal distribution according to expectation and variance of the data set obtained by the data subset under the same traction force on road models with different adhesion coefficients, so as to be suitable for an extended Kalman filter algorithm.
4. The Bayes classifier based road adhesion coefficient estimation method of claim 3, wherein the data set is obtained by analyzing and calculating data subsets of slip ratio S and traction F of left XL or right XR of the vehicle during driving, and calculating corresponding slip ratios according to the difference between the speeds of the driving wheel and the non-driving wheel and the friction correlation in slip:
in the above formula, S X Left side slip ratio, ω F ，ω R Respectively the left front wheel angular velocity and the left rear wheel angular velocity, R F ，R R The tire radii of the left front wheel and the left rear wheel are respectively.
5. The Bayes classifier based road adhesion coefficient estimation method according to claim 1, wherein S3 specifically is: and analyzing the prior probability of the data set based on normal distribution with different expectations and variances of the data set on the road surfaces with different attachment coefficients, and taking the expectation and the variance of the normal distribution under each attachment coefficient of each data subset as the prior probability.
6. The Bayes classifier based road adhesion coefficient estimation method according to claim 1, wherein the S4 specifically is: the prior probability belongs to a plurality of intervals which are divided according to the range of the traction force, data sets of different intervals are analyzed, the posterior probability is calculated in a corresponding model in a table look-up mode according to the obtained prior probability, A represents a newly collected data point, B represents a road surface model, and then the posterior probability P (B | A) is obtained.
7. The Bayes classifier based road adhesion coefficient estimation method as claimed in claim 6, wherein for a newly collected data point set A, a different traction force range is recorded as F 1 、F 2 、F 3 .., the road surfaces with different traction coefficients in the same traction range are denoted as B 1 、B 2 、B 3 .., in accordance with Bayes' formula, at F 1 The respective probabilities in the traction range are:
P 1 ＝P(A|B 1 )
P 2 ＝P(A|B 2 )
P 3 ＝P(A|B 3 ) 。
8. the Bayes classifier based road adhesion coefficient estimation method as claimed in claim 7, wherein it is preset that when a newly collected data point arrives, in each traction areaEqual in probability of occurrence therebetween, and then comparing said P 1 、P 2 、P 3 And (3) judging that the data point at the moment is positioned in the road model corresponding to the traction force section according to the maximum value of the calculated posterior probability.
9. The road surface adhesion coefficient estimation method based on the Bayes classifier as claimed in claim 1, wherein said S6 is specifically: the preset judgment condition for starting switching of the road surface is as follows: the posterior probability obtained through calculation of data points collected in real time is changed in Bayes classification compared with the road model of the data points collected last time, and at the moment, the road surface switching judgment logic is activated, and the change of the road model to which the road surface belongs is confirmed.
10. The Bayes classifier based road adhesion coefficient estimation method according to claim 1, wherein S7 is specifically implemented by dynamically adjusting a measurement noise covariance matrix of an extended Kalman filter algorithm, so as to realize fast convergence of a Kalman filter and accurate calculation of a road adhesion coefficient after switching, and a nonlinear system equation defining traction force and slip ratio is as follows:
wherein, U is traction force, S is slip ratio, and (a, b and c) are waiting coefficients; the above formula is substituted into the extended Kalman filtering algorithm, and a system model of the nonlinear system can be established:
X(k+1)＝f(X(k))+e(k)
Z(k)＝h(X(k))+U(k)
wherein k is the number of iterations, k +1 is the number of next iterations,
wherein X (k) = [ a, b, c ]; z (k) is equal to U, X (k) represents an estimated value of the parameters a, b and c in the current iteration period, X (k + 1) represents that the k +1 th iteration is an estimated value of the parameters a, b and c, f (X (k)) represents the relative conversion relation of the parameters in the adjacent period, e (k) is the system error of the parameters, Z (k) is an estimated value of the k-th period to the traction force U, U (k) represents a measured value of the traction force in the k-th period, h (X (k)) represents the relative conversion relation between the parameters a, b and c and the traction force, namely:
where S is a measurement of the kth cycle slip rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110205041.4A CN113619587B (en) | 2021-02-24 | 2021-02-24 | Road adhesion coefficient estimation method based on Bayes classifier |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110205041.4A CN113619587B (en) | 2021-02-24 | 2021-02-24 | Road adhesion coefficient estimation method based on Bayes classifier |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113619587A CN113619587A (en) | 2021-11-09 |
| CN113619587B true CN113619587B (en) | 2022-11-04 |
Family
ID=78377854
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110205041.4A Active CN113619587B (en) | 2021-02-24 | 2021-02-24 | Road adhesion coefficient estimation method based on Bayes classifier |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113619587B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101844561A (en) * | 2009-03-24 | 2010-09-29 | 通用汽车环球科技运作公司 | Road surface condition identification based on statistical model identification |
| CN103434511A (en) * | 2013-09-17 | 2013-12-11 | 东南大学 | Joint estimation method of travel speed and road attachment coefficient |
| CN104325980A (en) * | 2014-10-16 | 2015-02-04 | 北京汽车股份有限公司 | Attachment coefficient estimation method and device |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4046059B2 (en) * | 2002-11-08 | 2008-02-13 | 株式会社豊田中央研究所 | Road surface condition estimation device |
-
2021
- 2021-02-24 CN CN202110205041.4A patent/CN113619587B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101844561A (en) * | 2009-03-24 | 2010-09-29 | 通用汽车环球科技运作公司 | Road surface condition identification based on statistical model identification |
| CN103434511A (en) * | 2013-09-17 | 2013-12-11 | 东南大学 | Joint estimation method of travel speed and road attachment coefficient |
| CN104325980A (en) * | 2014-10-16 | 2015-02-04 | 北京汽车股份有限公司 | Attachment coefficient estimation method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113619587A (en) | 2021-11-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111104969B (en) | Collision possibility pre-judging method for unmanned vehicle and surrounding vehicles | |
| CN111079590B (en) | Peripheral vehicle behavior pre-judging method of unmanned vehicle | |
| CN110095635B (en) | Longitudinal speed estimation method of all-wheel drive vehicle | |
| CN108995653A (en) | A kind of driver's driving style recognition methods and system | |
| CN103839264B (en) | A kind of detection method of lane line | |
| CN104554271B (en) | A kind of road gradient and vehicle condition parametric joint method of estimation based on parameter estimating error | |
| CN109910897B (en) | Safe distance estimation method based on front road surface peak value adhesion coefficient | |
| CN113386778B (en) | Method for judging rapid deceleration driving behavior based on vehicle driving track data | |
| CN111459995B (en) | Multi-mode vehicle speed prediction method based on driving data | |
| CN110386144B (en) | GHMM/GGAP-RBF hybrid model for identifying driver braking intention and identification method | |
| CN111688706A (en) | Road adhesion coefficient interactive estimation method based on vision and dynamics | |
| CN111605559B (en) | Vehicle mass estimation method, torque control method and device | |
| CN110450794A (en) | A kind of optimized adhesion control method based on optimal creep speed search and tracking | |
| WO2021115063A1 (en) | Data-driven method and apparatus for identifying adhesion stability of electric-driven vehicle | |
| CN111688707A (en) | Vision and dynamics fused road adhesion coefficient estimation method | |
| CN107274668A (en) | A kind of congestion in road modeling method based on vehicle detection | |
| CN111994129A (en) | Wheel-rail train antiskid control method and system | |
| CN112464773A (en) | Road type identification method, device and system | |
| CN112677982A (en) | Vehicle longitudinal speed planning method based on driver characteristics | |
| CN110502971A (en) | Road vehicle recognition methods and system based on monocular vision | |
| CN113619587B (en) | Road adhesion coefficient estimation method based on Bayes classifier | |
| CN115476861A (en) | Safety evaluation system of intelligent networked automobile | |
| CN112328651B (en) | Traffic target identification method based on millimeter wave radar data statistical characteristics | |
| Jin et al. | Road friction estimation method based on fusion of machine vision and vehicle dynamics | |
| CN109190153A (en) | A kind of Calculation Method of Energy Consumption and its system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |