CN101275900A - Method for recognizing road surface types based on vehicle wheel vibration - Google Patents

Abstract

Disclosed is a road surface type identification method based on wheel vibration, acquiring a vibration signal of the present wheel in the vehicle driving process after the wheel vibration model is established to obtain a present wheel vibration high-frequency spectrum eigenvector which is then compared with the wheel vibration high-frequency spectrum eigenvector of the typical road surface, so as to identify the road surface type. The method is simple and practicable, of simple equipment and easy installation.

Description

Method for recognizing road surface types based on unsteadiness of wheels

Technical field

The invention belongs to the vehicle driving safety technical field, be specifically related in a kind of vehicle ' process, the method for road surface types identification.

Background technology

Necessary power that vehicle safety is smooth-going to travel (as ABS, ASR, EBD, ESP etc.) all derives from the surface of contact on tire and road surface, and this power depends on the skid resistance on road surface.Therefore, the pavement skid resistance evaluation is the necessary link of vehicle safety control.

Pavement skid resistance is by the road surface characteristics determined, and the road surface feature comprises macrostructure and micromechanism (Sha Qinglin, " asphalt pavement of highway early damage phenomenon and prevention ", People's Transportation Press, in January, 2005).The microtexture on road surface is meant the roughness of road surface surface of aggregate, i.e. surface of aggregate horizontal direction 0～0.5mm, and the microtexture of vertical direction 0～0.2mm, micromechanism road pavement cling property when low speed (below 30～50km/h) plays a decisive role.The macrostructure of road surfaces is meant space or the drainability of road surface between gathering materials, the structure that forms between gathering materials by the road surface, and promptly the road level direction is 0.5～50mm, and vertical direction is 0.2～10mm, and macrostructure mainly influences the anti-slide performance when running at high speed.

At present, common expressway face comprises cement concrete pavement and SMA bituminous pavement, these two kinds of road surface aggregate size differences, road surface table feature macrostructure difference.The cement concrete pavement surface gathers materials and is the medium sand between the modulus of fineness 2.3～3.2, medium coarse sand and thin partially coarse sand, and the cement concrete pavement surface characteristics is levied macrostructure and seen Fig. 1.SMA bituminous pavement skeleton is for greater than the coarse aggregate of 4.75mm (account for compound 70%), pitch, breeze, fine aggregate, fibrous matizhi inserts (account for compound 30%), and SMA bituminous pavement surface characteristics is levied macrostructure and is seen Fig. 2.

In sum, dissimilar road surfaces, surface characteristics macrostructure difference.Skid resistance when macrostructure influences galloping is the principal element that influences automotive safety.Therefore, can estimate the skid resistance on road surface by measuring road surface feature macrostructure, its essence is exactly road surface types identification.

The common method of measuring road surface feature macrostructure at present is as follows:

1, surface profile tire sensor identification road surface feature macrostructure (Han Jianbao, Zhang Lubin, Li Bangguo, " the real-time induction recognition system of tire coefficient of road adhesion ", vehicle and power technology, the second phase in 2005)

The surface profile tire sensor, see Fig. 3, its core component is the dull and stereotyped piezoelectric crystal of little shape, electromotive force by the piezoelectric crystal generation, the distortion of perception rubber for tire (essence is to obtain road surface feature macrostructure) identifies the size of wheel drive power, damping force, side force, wheel weight, tire pressure, tire print and the attachment coefficient between position, tire and the ground etc. in view of the above.The not charged source of tire sensor itself, by the excitation of external wireless electromagnetic wave, information is transmitted into the vehicle electronics receiving equipment by radio magnetic wave simultaneously.

There is following shortcoming in this method:

(1) accurately the location is relatively more difficult in rubber for tire for sensor;

(2) sensor wireless power supply and wireless messages transmission more complicated;

(3) because of sensor is very little, it is fewer to obtain information, can not reflect the road surface overall condition.

2, road surface feature (Pieter L.Swart is measured in sound scattering, BeatrysM.Lacquet, " An Acoustic Sensor System for Determination ofMacroscopic Surface Roughness ", IEEE Transactions onInstrumentation and Measurement, Vol.45, No.5, October 1996)

The signal of road surface feature macrostructure is measured in sound scattering, sees Fig. 4, and it is measured road surface characteristic principle and is, to road surface emission sound wave, the road surface feature is coarse more, then the sound wave that wavelength is more little be scattered many more.Emitter is launched sound wave with specific sound wave (frequency and power) and emission angle continuously to the road surface, and the sound wave receiving device is accepted the road reflection sound wave with specific scope.By the sound wave that analysis receives, the sound wave of certain frequency is reflected as can be known, and the sound wave of certain frequency is scattered.According to reflection, diffuse sound wave frequency and degree, estimate road surface feature macrostructure.

There is following shortcoming in this method:

(1) sound wave that receives is subjected to extraneous sound wave easily and disturbs;

(2) equipment is installed inconvenience, bumps closely easily from the road surface, and precision far away reduces from the road surface;

(3) emitter and receiving device are subjected to pollutions such as dust, rainwater easily.

Summary of the invention

The object of the present invention is to provide a kind of method for recognizing road surface types, to overcome the defective that said method exists based on unsteadiness of wheels.

The technical scheme that the present invention is based on the method for recognizing road surface types of unsteadiness of wheels is: it is after setting up the unsteadiness of wheels model, when vehicle ', gather vibration signal when front vehicle wheel, obtain current unsteadiness of wheels high frequency spectrum proper vector, compare with typical road surface unsteadiness of wheels high frequency spectrum proper vector, thus the identification road surface types.

At first, set up the unsteadiness of wheels model, prove that unsteadiness of wheels high frequency frequency equates with road surface macrofeature excitation frequency; Secondly, gather many groups typical road surface (as cement concrete pavement, SMA bituminous pavement etc.) unsteadiness of wheels signal,, obtain the typical road surface of many groups unsteadiness of wheels high frequency spectrum proper vector by wavelet analysis and Fast Fourier Transform (FFT); Utilize unsteadiness of wheels high frequency spectrum proper vector structure road surface, typical road surface RBF neural network classifier; At last, gather road surface to be identified unsteadiness of wheels signal, obtain its spectrum signature vector, input road surface RBF neural network classifier obtains the road surface types recognition result.

This method is simple, and device therefor is simple, and is easy for installation.

The present invention's main effect in vehicle safety control (as ABS, ASR, EBD, ESP etc.) is as follows:

(1) determines optimal slip ratio

Therefore dissimilar road surfaces optimal slip ratio difference can determine the vehicle optimal slip ratio according to road surface types, sees Fig. 5.

(2) determine maximum braking force and driving force

After determining road surface types and optimal slip ratio, just can determine the road surface maximum grip coefficient, see Fig. 5.Maximum grip coefficient multiply by car weight, is exactly the maximal friction that the road surface can provide.Determine maximum braking force of vehicle and driving force according to maximal friction.

Description of drawings

Fig. 1 cement concrete pavement surface characteristics is levied macrostructure.

Fig. 2 SMA bituminous pavement surface characteristics is levied macrostructure.

Fig. 3 surface profile sensor and match are relatively.

Fig. 4 measures sound scattering identification road surface, road surface.

μ-λ the curve on the different road surfaces of Fig. 5.

Fig. 6 is based on the road surface types identification process of unsteadiness of wheels.

Fig. 7 single-degree-of-freedom unsteadiness of wheels model.

Fig. 8 cement concrete pavement, asphalt surface unsteadiness of wheels signal.

The reconstruct of Fig. 9 list band improves algorithm.

Figure 10 cement concrete pavement, asphalt surface unsteadiness of wheels spectrum signature vector.

Figure 11 RBF neural network structure.

Figure 12 unsteadiness of wheels neural network road surface types identification equipment principle.

Embodiment

The road surface types identification process that the present invention is based on unsteadiness of wheels is seen Fig. 6.It consists of the following components: set up the unsteadiness of wheels model, gather the unsteadiness of wheels signal, obtain the typical road surface of many groups unsteadiness of wheels high frequency spectrum proper vector, structure road surface types identification neural network classifier, exploitation unsteadiness of wheels neural network road surface types identification equipment.

1.1 set up the unsteadiness of wheels model

This method research to as if unsteadiness of wheels, according to research needs, to the Vehicular vibration system suitably simplify (Jin Xiaoxiong, Zhang Lijun, Jiang Hao, " automobile vibration analysis ", publishing house of Tongji University, in May, 2002 first published).When vehicle mass partition factor ε=1, unsteadiness of wheels is not contact each other, and the unsteadiness of wheels model simplification is seen Fig. 7 for the single-degree-of-freedom unsteadiness of wheels model that wheel m1 forms.Unsteadiness of wheels comprises that road roughness causes that wheel low-frequency vibration and road surface feature macrostructure cause the unsteadiness of wheels high frequency, and dither is a research object of the present invention, and low-frequency vibration can be rejected by discriminated union.

Below proof unsteadiness of wheels high frequency spectrum is identical with road surface feature macrostructure excitation spectrum:

Simple in order to analyze, the excitation of road surface feature macrostructure is reduced to

x _s＝a?sin?ωt (1)

The unsteadiness of wheels model differential equation is

$m_1{x}_1+c_1{x}_1+k_1{x}_1=k_1{x}_s+c_1---\left(2\right)$

This oscillatory differential equation separate for

$x_1\left(t\right)=a\sqrt{\frac{1+{\left(2\mathrm{\ξ\λ}\right)}^2}{{(1-{\mathrm{\λ}}^2)}^2+{\left(2\mathrm{\ξ\λ}\right)}^2}}\sin (\mathrm{\ωt}-\mathrm{\ψ})---\left(3\right)$

Wherein, x _sBe the excitation of road surface feature, a is road surface feature excitation amplitude, and ω is a road surface feature macrostructure excitation frequency, m ₁Be tire quality, x ₁Be unsteadiness of wheels, c ₁Be tire damping, k ₁Be tire stiffness, ξ is a damping ratio, and λ is a frequency ratio, and ψ is a drag angle.

Separate (3) from the differential equation, prove that the unsteadiness of wheels high frequency spectrum is identical with road surface feature macrostructure excitation spectrum.

1.2 gather the unsteadiness of wheels signal

Near wheel position acceleration transducer is installed vertically in vehicle bridge, is measured the wheel vertical vibration.Vehicle ' is gathered many groups road surface unsteadiness of wheels signal on typical high speed road surface, wherein one group of cement concrete pavement, SMA bituminous pavement unsteadiness of wheels signal are seen Fig. 8.

1.3 obtain the typical road surface of many groups unsteadiness of wheels high frequency spectrum proper vector

(1.3.1) wavelet analysis unsteadiness of wheels signal

Select suitable wavelet function and yardstick, adopt the reconstruct of list band improve algorithm (Yang Jianguo, " wavelet analysis and engineering thereof are used ", China Machine Press, in July, 2005 first published), obtain unsteadiness of wheels high-frequency sub-band signal.The reconstruct of list band improves algorithm and sees Fig. 9.

(1.3.2) obtain the unsteadiness of wheels high frequency spectrum

Unsteadiness of wheels high-frequency sub-band signal is carried out fast fourier transform, obtain the unsteadiness of wheels high frequency spectrum.

(1.3.3) obtain unsteadiness of wheels high frequency spectrum proper vector

To disturb and reduce data volume in order to reduce, with branch unsteadiness of wheels high frequency spectrums such as fixed frequency bands, and with spectral magnitude mean value as this band spectrum amplitude, frequency spectrum is carried out normalization, obtain unsteadiness of wheels high frequency spectrum proper vector.

$f\left(x\right)=\left\{\begin{array}{cc}{F}_1& {\mathrm{<figure-callout\; id="0"\; label="f"\; filenames="A20081004761200132.png,A20081004761200141.png"\; state="\{\{state\}\}">f</figure-callout>}}_0<x<f_1\\ F_2& {\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="A20081004761200132.png,A20081004761200133.png"\; state="\{\{state\}\}">f</figure-callout>}}_1<x<{\mathrm{<figure-callout\; id="2"\; label="f"\; filenames="A20081004761200142.png"\; state="\{\{state\}\}">f</figure-callout>}}_2\\ & \mathrm{\&Lambda;}\\ F_n& f_{n-1}<x<f_n\end{array}\right.---\left(4\right)$

On typical road surface, gather all kinds of road surfaces and organize the unsteadiness of wheels signal more, and obtain all kinds of road surfaces and organize unsteadiness of wheels spectrum signature vector more.Wherein, one group of cement concrete pavement, SMA bituminous pavement spectrum signature vector are seen Figure 10 (last figure is a cement concrete pavement spectrum signature vector, and figure below is a SMA bituminous pavement spectrum signature vector).

1.4 structure road surface types identification neural network classifier

(1.4.1) set up road surface types identification RBF neural network

The RBF neural network is made up of two-layer, comprises hidden layer and output layer, and hidden layer has several neurons, and what the node function was the most frequently used is Gaussian function, and output layer has several neurons, and the node function is simple linear function.If network input X is a n-dimensional vector, output Y is the L dimensional vector, and its structure is seen Figure 11.

(1.4.2) training road surface types identification RBF neural network

The learning process of RBF network is divided into two stages, and the phase one is the teacherless learning, and subordinate phase is that teacher learning is arranged.

Teacherless learning's stage

Cluster (Li Guoyong, " Based Intelligent Control and MATLAB thereof realize ", Electronic Industry Press, in May, 2005) is carried out in the input of all samples, try to achieve the center vector c of each hidden node _iBelow be the K-means clustering algorithm, algorithm steps is as follows:

The initial center vector c of given each hidden node _i(0) and judge the ε stop to calculate;

(i) computed range (Euclidean distance) and obtain the minor increment node;

$\left\{\begin{array}{c}{d}_i\left(k\right)=\left|\right|x\left(k\right)-c_i(k-1)\left|\right|,1\&le;i\&le;q\\ d_{\min }\left(k\right)=\min d_i\left(k\right)=d_r\left(k\right)\end{array}\right.---\left(5\right)$

I=1 wherein, 2,3...q, q are the number of hidden nodes, c _iThe center vector of the gaussian kernel function of i hidden node.

(ii) adjust the center

$\left\{\begin{array}{c}{c}_i\left(k\right)=c_i(k-1),1\&le;i\&le;q,i\&le;r\\ c_r\left(k\right)=c_r(k-1)+\mathrm{\&beta;}\left(k\right)[x\left(t\right)-c_r(k-1)]\end{array}\right.---\left(6\right)$

In the formula, β (k) is a learning rate, β (k)=β (k-1)/1+int (k/q) ^1/2, learning rate reduces gradually.

Judge the cluster quality

Whole sample k are carried out above i, ii step repeatedly, and up to meeting the following conditions, then cluster finishes.

$J_e=\underset{i=1}{\overset{q}{\mathrm{\&Sigma;}}}{\left|\right|\stackrel{\&OverBar;}{x}\left(k\right)-{\stackrel{\&OverBar;}{c}}_i\left(k\right)\left|\right|}^2\&le;\mathrm{\&epsiv;}---\left(7\right)$

Teacher learning is arranged

Work as c _iAfter determining, train by hidden layer to the weights between the output layer.It is a system of linear equations, then asks weights with regard to becoming the linear optimization problem, can obtain overall smallest point certainly.Ask the weights W of hidden layer and output layer _iLearning algorithm is

$w_{\mathrm{ki}}(k+1)=w_{\mathrm{ki}}\left(k\right)+\mathrm{\&eta;}(t_k-y_k)u_i\left(\stackrel{\&OverBar;}{x}\right)/{\stackrel{\&OverBar;}{u}}^T\stackrel{\&OverBar;}{u}---\left(8\right)$

In the formula, u _i(k+1) be Gaussian function, η is a learning rate.

Organize unsteadiness of wheels spectrum signature vector training RBF neural network with all kinds of road surfaces more, finally obtain road surface types identification neural network classifier.

1.5 exploitation unsteadiness of wheels neural network road surface types identification equipment

Unsteadiness of wheels neural network road surface types identification equipment principle is seen Figure 12, sensor is used to gather road surface to be identified unsteadiness of wheels signal, data acquisition system (DAS) is finished unsteadiness of wheels signal from analog amount to the conversion of digital quantity, DSP finishes and 1.3 obtains unsteadiness of wheels high frequency spectrum proper vector, the road surface RBF neural network classifications of 1.4 structures of packing into and with vehicle safety control ECU data communication.

In the present invention's experiment, identifying object is cement concrete pavement and SMA bituminous pavement, and the speed of a motor vehicle is 40Km/h.In the implementation process, at first gather the unsteadiness of wheels signal, obtain unsteadiness of wheels high frequency spectrum proper vector, structure road surface types identification neural network classifier with Matlab emulation with the high speed dynamic collecting instrument; Next develops the road surface types identification equipment, and comprising with the single-chip microcomputer is the unsteadiness of wheels data acquisition system (DAS) of core, is signal analysis, the identification of RBF neural network road surface types and the communication system of core with DSP.

2.1 gather the unsteadiness of wheels signal

Automobile is selected Shanghai GM Buick Sail SL1.6 car for use; The high speed dynamic collecting instrument is selected wavebook516E and expansion module wbk18 thereof for use, and it is 16KHz that sample frequency is set, and it is 5KHz that the frequency overlapped-resistable filter frequency is set; Acceleration transducer is selected the CA-YD-181 type for use, and its effective vibration frequency is 1～10KHZ, and it is vertically mounted on back axle apart from wheel 30mm place.

Respectively gather 50 groups of unsteadiness of wheels signals at cement concrete pavement, SMA bituminous pavement, be to guarantee real-time and accuracy, gather that to count be 4096, wherein one group of cement concrete pavement and SMA bituminous pavement unsteadiness of wheels signal are seen Fig. 3.

2.2 obtain the typical road surface of many groups unsteadiness of wheels high frequency spectrum proper vector

(2.2.1) wavelet analysis unsteadiness of wheels signal

Adopting the reconstruct of small echo list band to improve algorithm decomposes and reconstruct unsteadiness of wheels signal.Wavelet function is selected db10 for use, carries out 2 yardstick list band reconstruct, obtains wheel 2K～4KHz dither subband signal.The biorthogonal wavelet bank of filters function that wave filter directly adopts Matlab to provide:

[H_D，G_D，h_R，g_R]＝wfilters(′db10′)

(2.2.2) obtain the unsteadiness of wheels high frequency spectrum

The high-frequency sub-band signal that obtains is carried out fast fourier transform, obtain the high-frequency sub-band frequency spectrum.Fast fourier transform is directly called Matlab the ground function is provided:

X2＝fft(d2，4096)

D2 wheel 2K～4KHz dither subband signal wherein, X2 are frequency spectrum in vibration signal 2K～4KHz frequency band.

(2.2.3) obtain unsteadiness of wheels high frequency spectrum proper vector

Each rumble spectrum is divided into 20 frequency ranges, and with every frequency range average amplitude as this band spectrum amplitude, again frequency spectrum is carried out normalization, obtain unsteadiness of wheels spectrum signature vector.Wherein, one group of cement concrete pavement and SMA bituminous pavement spectrum signature vector are seen Fig. 6.

50 groups of cement concrete pavements and SMA bituminous pavement unsteadiness of wheels signal are analyzed by (2.2.1), (2.2.2), (2.2.3) step, obtained 50 groups of spectrum signature vectors, each vector 20 dimension.Wherein 40 groups of vector training road surface types are discerned neural network classifiers, and other 10 groups are used to check neural network classifier.

2.3 structure road surface neural network classifier

(2.3.1) set up road surface types identification RBF neural network, the network input vector is 20 dimensions, and the hidden layer neuron number is 20, and the network output vector is 1 dimension;

(2.3.2) training road surface types identification RBF neural network at first with 40 cement concrete pavement unsteadiness of wheels high frequency spectrum proper vector input RBF neural networks, makes network be output as 0; Secondly with 40 asphalt surface unsteadiness of wheels high frequency spectrum proper vector input RBF neural networks, make network be output as 1.0.By training RBF neural network, obtain the road surface neural network classifier.

2.4 neural network road surface types identification checking

With spectrum signature vector input road surface, other 10 groups of road surfaces neural network classifier, network output the results are shown in Table 1.Network classifier output near 0 be cement concrete pavement, network classifier export the result near 1.0 be the SMA bituminous pavement.From road surface types identification experimental result as can be seen, for cement concrete pavement and SMA bituminous pavement, recognition accuracy can reach 100%.

The neural network classifier output of table 1 road surface

Concrete road surface face SMA bituminous pavement

1 -0.15743 6 -0.16787 1 0.84977 6 0.82057

2 -0.16787 7 -0.16053 2 0.97070 7 0.79684

3 0.028965 8 -0.15645 3 0.75505 8 0.59715

4 -0.17188 9 -0.19633 4 0.79872 9 0.87721

5 -0.09944 10 -0.16654 5 0.79484 10 0.98537

2.5 exploitation unsteadiness of wheels neural network road surface types identification equipment

Unsteadiness of wheels neural network road surface types identification equipment principle is seen Figure 12, and this equipment comprises:

(2.5.1) wheel vibration sensor is used to gather road surface to be identified unsteadiness of wheels signal, and its model is selected CA-YD-181 for use.

Be the unsteadiness of wheels data acquisition system (DAS) of core with the single-chip microcomputer (2.5.2), finish the conversion of unsteadiness of wheels signal A/D, single-chip microcomputer is selected AT89C2051 for use, and A/D converter is selected AD7574 for use, and data-carrier store is selected dual port RAM IDT7134 for use.

(2.5.3) with DSP be signal analysis, neural network road surface types recognition system and the communication system of core, inside comprises the data analysis algorithm and passes through the road surface neural network classifiers of 2.3 structures, finish signal spectral analysis, road surface classification, data send the automotive safety electronic control unit to by bus at last.DSP selects TMS320LF2407 for use.

Claims (7)

1, a kind of method for recognizing road surface types based on unsteadiness of wheels, it is after setting up the unsteadiness of wheels model, when vehicle ', gather vibration signal when front vehicle wheel, obtain current unsteadiness of wheels high frequency spectrum proper vector, compare with typical road surface unsteadiness of wheels high frequency spectrum proper vector, thus the identification road surface types.

2, according to claim 1 based on the method for recognizing road surface types of unsteadiness of wheels, it is characterized in that the described method of obtaining unsteadiness of wheels high frequency spectrum proper vector is: adopt the reconstruct of list band to improve algorithm and decompose and reconstruct unsteadiness of wheels signal, obtain unsteadiness of wheels high-frequency sub-band signal; Unsteadiness of wheels high-frequency sub-band signal is carried out fast fourier transform, obtain the unsteadiness of wheels high frequency spectrum; With branch unsteadiness of wheels high frequency spectrums such as fixed frequency bands, and with spectral magnitude mean value as this band spectrum amplitude, frequency spectrum is carried out normalization, obtain unsteadiness of wheels high frequency spectrum proper vector.

3, according to claim 1 based on the method for recognizing road surface types of unsteadiness of wheels, it is characterized in that current unsteadiness of wheels high frequency spectrum proper vector and each quasi-representative road surface unsteadiness of wheels high frequency spectrum proper vector are compared, concrete grammar is: structure road surface types identification neural network classifier, utilize the neural network classifier road pavement to discern.

4, as described in the claim 3 based on the method for recognizing road surface types of unsteadiness of wheels, it is characterized in that the method for constructing road surface types identification neural network classifier is: set up road surface types identification RBF neural network, the RBF neural network is made up of two-layer, comprises hidden layer and output layer; Utilize the unsteadiness of wheels spectrum signature vector training RBF neural network of unsteadiness of wheels model, realize road surface types identification thereby finally obtain road surface types identification neural network classifier.

5, as described in the claim 4 based on the method for recognizing road surface types of unsteadiness of wheels, it is characterized in that described training RBF neural network comprises that the phase one is the teacherless learning, subordinate phase is that teacher learning is arranged; Described teacherless learning carries out cluster to the input of all samples, tries to achieve the center vector c of each hidden node _iDescribed teacher learning is to work as c _iAfter determining, train by hidden layer to the weight w between the output layer _i

6, as described in the claim 2 based on the method for recognizing road surface types of unsteadiness of wheels, it is characterized in that and will obtain the method dsp program of unsteadiness of wheels high frequency spectrum proper vector, and download in the dsp chip.

7, as described in the claim 3 based on the method for recognizing road surface types of unsteadiness of wheels, it is characterized in that the RBF neural network classifier dsp program that will obtain, and in the download dsp chip.

Images