CN101311047A - Vehicle anti-lock brake control method based on least squares support vector machine - Google Patents

Abstract

The invention discloses a method for controlling an anti-lock braking system of automobiles, which bases on least squares support vector machines (LS-SVM) and relates to the technical filed of the electronic control of the automobiles. The controlling method estimates automobile body deceleration according to the wheel velocity peak values of four wheels; the automobile body reference speed is calculated according to the automobile body deceleration and then the reference slip rate of the wheels is obtained by calculation; the angular deceleration of the wheels is calculated according to the angular speed of the wheels; the angular speed and reference slip rate of every wheel are respectively taken as input to construct a controller of the least squares support vector machines; the expected brake pressure of every wheel is calculated; according to the expected brake pressure, a pulse width modulation method is adopted to control the on/off time of an electromagnetic valve, thus controlling the brake pressure and realizing safety braking. The method is applicable to controlling the anti-lock braking system (ABS) of automobiles.

Description

Vehicle anti-lock brake control method based on least square method supporting vector machine

Technical field

The present invention relates to the auto electronic control technology field, be specifically related to a kind of control method of automobile anti-lock braking braking control system.

Background technology

Automobile anti-lock braking braking control system (ABS) is the important electronic system that guarantees automobile not locking of wheel in the emergency braking process, for guaranteeing not locking of wheel, can brake with fast speeds again simultaneously, ABS must adopt different control policies at different road surface situations, therefore the ABS system must real-time identification road surface situation, and then takes different control policies that brake system is controlled according to the road surface situation.The ABS control method of existing maturation mainly comprises based on wheel acceleration-deceleration thresholding auxiliary reference slip rate method, fuzzy control method, based on neural network method, based on method of Based Intelligent Control etc.Yet being control system, the ubiquitous problem of these control methods needs the real-time identification information of road surface, need the controlled variable of coupling more, need the very long experiment coupling cycle so that each parameter in the control algorithm is in optimum state, the assurance automobile is braked under various pavement conditions all stronger comformability.College journal in June, 2003 northeastward such as Chen Jun for this reason, " based on the ABS road surface identification of competition neural network " that the 24th the 6th phase of volume delivered adopts nerual network technique to carry out the road identification technology, agricultural mechanical journal, " the discerning the ABS Study on Fuzzy Controlling System automatically based on road " of delivering September calendar year 2001 adopted the method identification road surface of theoretical deceleration/decel of contrast wheel and wheel actual deceleration degree, Jin-Oh Hahn etc. adopt method identification tire and ground-surface friction coefficient (the Jin-Oh Hahn based on GPS, Rajesh Rajamani, and LeeAlexander.GPS-Based Real-Time Identification of Tire-Road FrictionCoefficient, IEEE TRANSACTIONS ON CONTROL SYSTEMS TECHNOLOGY, VOL.10, NO.3).Yet make a general survey of above these methods, need a large amount of sampled datas that network is trained based on neural network method, the method of theoretical deceleration/decel of contrast wheel and wheel actual deceleration degree need be known the quality of automobile, and the ABS system and do not know any quality of lorry loading after total mass, so this method can not realize the automatic identification of ground-surface in the ABS of bogie, and need automobile to be equipped with global positioning system based on the method for GPS, so cost is higher.Tang Dongsheng etc. are in " automotive engineering " in May, 2003, adopt genetic algorithm in " abs controller optimum Design of Parameters " that the 25th the 4th phase of volume delivered, on the basis of emulation experiment, adopt genetic algorithm that the controlled variable of ABS system is optimized, yet based on the problem that the parameter optimization of emulation exists is that realistic model is not considered the wheel speed signal fluctuation in the Motor Vehicle Braking Procedure and the time-delay of brake system, therefore have bigger difference between analogue system and the real system, precision is relatively poor.The optimization parameter that obtains on the basis of emulation probably is not the optimized parameter of braking control system, even, if can make the automobile braking procedure of causing danger as the controlled variable of ABS control system with this parameter.

Summary of the invention

Technical matters to be solved by this invention is, at needing real-time identification road surface situation in the prior art vehicle anti-lock brake control method, and need the very long road examination cycle with the controlled variable in the matching algorithm, controlled variable is inaccurate, defectives such as cost of development height.Propose a kind of vehicle anti-lock brake control method based on least square method supporting vector machine, this method need not carried out road identification, only need utilize support vector controller to be trained each parameter that gets final product controlled device.

The present invention solves the problems of the technologies described above the technical scheme that is adopted: propose a kind of vehicle anti-lock brake control method based on least square method supporting vector machine, this method at first utilizes ripe ECU to carry out ABS (Anti-lock Braking System) control experiment under various road surfaces, when automobile carries out emergency braking, the pressure signal of wheel speed and pressure sensor automobile wheel speed signal and brake wheel cylinder in the line drawing braking procedure, as support vector controller is trained the regression parameter a and the b of controlled device, utilize method of finite difference to obtain the angular deceleration ω of wheel; Slope according to two adjacent wheel speed peak of curve lines obtains vehicle body reference deceleration degree v, and obtains the vehicle body reference speed thus; Calculate the reference slip rate of each wheel according to vehicle body reference velocity and wheel angular deceleration; Wheel angular deceleration and wheel reference slip rate as input, are made up least square method supporting vector machine LS-SVM controller, calculate the desired braking pressure of each wheel; According to the desired braking pressure of each wheel, adopt the make-and-break time of the method control electromagnetic valve of pulse duration modulation, thereby the brake-pressure of controlling each wheel reaches desired braking pressure, realize braking fully.According to the positive and negative relationship change of wheel angular deceleration, determine the wheel speed peak of curve; Controller algorithm module invokes formula $S=\frac{v-\mathrm{r\ω}}{v}\×100\%$ Calculate the reference slip rate of each wheel of automobile.Described least square method supporting vector machine LS-SVM controller is the three layer model structure, comprise, input layer, hidden layer, output layer, described input layer is an incoming signal with wheel reference slip rate S and wheel angular deceleration ω ', make up the input function relation of x=(s, ω '); Hidden layer utilizes kernel function K (x, x _i)=exp (|| x-x _i|| ²/ σ ²) two-dimentional input vector and support vector are carried out kernel operation; Output layer is according to formula $f\left(x\right)=\underset{i=1}{\overset{g}{\mathrm{\Σ}}}{a}_{i}K(x,x_i)+b=P$ Make up the desired braking pressure P of wheel and the relation of kernel function.

The present invention has designed a kind of with low cost, method of calculating is simple, in the emergency braking process, with the reference slip rate of wheel angular deceleration and wheel as input, according to the angular deceleration of wheel with reference to slip rate, the brake-pressure of direct calculation expectation, and then, realize safety arrestment by pulse duration modulation method control brake pressure.Need not carry out road identification, only need utilize support vector that controller is trained each parameter that gets final product controlled device, time-to-market is short, and cost of development is low, and the controlled variable that obtains accurately and reliably.

Description of drawings

Fig. 1 is based on the abs controller structure of LS-SVM

Fig. 2 is to the training scheme drawing based on the abs controller of least square method supporting vector machine

Fig. 3 is based on the control process scheme drawing of the abs controller of LS-SVM

The specific embodiment

The present invention has designed a kind of vehicle anti-lock brake control method based on least square method supporting vector machine (LS-SVM), this method utilizes ripe automobile ABS ECU as controller, under various pavement conditions, carry out ABS (Anti-lock Braking System) control experiment, gather the pressure signal of wheel wheel speed and brake wheel cylinder in the experimentation by wheel speed sensors and pressure sensor, according to the wheel speed and the sampling period of wheel, utilize method of finite difference to obtain the angular deceleration of wheel, can the size that concern of continuous three or above wheel angular deceleration be compared, positive and negative relationship change according to the wheel angular deceleration, determine the peak value of wheel speed curve, slope according to two adjacent wheel speed peak of curve lines obtains vehicle body reference deceleration degree, vehicle body reference speed according to vehicle body reference deceleration degree and initial speed of a motor vehicle calculating automobile, calculate the reference slip rate of each wheel of automobile then, obtain by the wheel reference slip rate, the support vector that wheel angular deceleration and wheel cylinder brake-pressure are formed; Utilize above-mentioned support vector that controller is trained, the regression parameter a of controlled device and b; Structure has least square method supporting vector machine (LS-SVM) controller of three-decker, the three-decker of this controller comprises, input layer, hidden layer, output layer, described input layer is an incoming signal with wheel reference slip rate S and the wheel angular deceleration ω ' that automobile ABS ECU calculates, according to formula x=(x ¹, x ²)=(s, ω ') structure input function relation; Hidden layer utilizes kernel function to call formula: K (x, x _i)=exp (|| x-x _i|| ²/ σ ²) two-dimentional input vector and support vector are carried out kernel operation; Output layer calls formula $f\left(x\right)=\underset{i=1}{\overset{g}{\mathrm{\Σ}}}{a}_{i}K(x,x_i)+b=P$ Calculate the desired braking pressure P of wheel.With wheel angular deceleration and wheel reference slip rate is incoming signal based on the least square method supporting vector machine controller, through the computing of SVMs, obtains the desired braking pressure P of wheel at output layer.According to the desired braking pressure of the wheel of output layer output, adopt the make-and-break time of pulse duration modulation method control electromagnetic valve, and then the Control of Automobile brake-pressure, thereby be controlled at suitable scope with the angular deceleration of wheel with reference to slip rate, realize safety arrestment.

Be illustrated in figure 1 as training scheme drawing based on the Automobile ABS Controller of LS-SVM.At first use ripe automobile ABS ECU as control unit, do roadway experiment under different pavement conditions, wheel speed sensors extracts wheel velocity signal with certain sampling period, obtains wheel velocity ω, and obtains the pressure value P of brake wheel cylinder with pressure sensor.(ω P), regularly obtaining on the basis of wheel velocity according to the predetermined sampling period, utilizes method of finite difference to calculate wheel angular deceleration ω ' all can to obtain one group of bivector for each wheel like this; Controller judges, if the angular deceleration of wheel is greater than the angular deceleration of last sampling instant and back one sampling instant constantly for certain, the wheel speed of then judging this moment is the peak value wheel speed, obtains the peak value of wheel speed curve thus.The slope that calculates adjacent two wheel speed peak value lines obtains vehicle body reference deceleration degree, begin the rate of onset of the maximum wheel speed of glancing impact with automobile as vehicle body, and, calculate the reference slip rate of wheel at each sampling point by vehicle body reference speed and wheel velocity according to vehicle body reference deceleration degree and sampling period calculating vehicle body reference velocity.With the wheel angular deceleration, send into the LS-SVM controller model with reference to the pressure of slip rate, brake wheel cylinder as support vector and train, obtain regression parameter a, b.

Figure 2 shows that the abs controller model structure figure based on LS-SVM, controller model has the three-layer network structure, comprises input layer, hidden layer, output layer.

1) input layer, wheel reference slip rate S that calculates with automobile ABS ECU and wheel angular deceleration ω ' are as the incoming signal of input layer, according to formula x=(x ¹, x ²)=(s, ω ') make up the input function relation, constitute the two-dimentional incoming signal of SVMs.

2) hidden layer carries out kernel operation to two-dimentional input vector and support vector.Hidden layer utilizes kernel function to call formula: K (x, x _i)=exp (|| x-x _i|| ²/ σ ²) two-dimentional input vector and support vector are carried out kernel operation, K (x, x wherein _i) be kernel function, x is an input vector, x _iBe i support vector in the support vector, determine by the wheel reference slip rate and the wheel angular deceleration of i sampling instant, i.e. x _i=(si, ω ' _i), $\left|\right|x-x_i\left|\right|=\sqrt{\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{(x^k-x_i^k)}^2},$ σ is the nuclear width, needs through repeatedly experiment is definite.By kernel operation, data are mapped to high-dimensional feature space.

3) output layer

Output layer calls formula $f\left(x\right)=\underset{i=1}{\overset{g}{\mathrm{\Σ}}}\mathrm{aK}(x,x_i)+b=P$ Carry out linear regression at high-dimensional feature space, according to the kernel operation result of hidden layer output, regression parameter a in the controller and b, the desired braking pressure P of acquisition brake wheel cylinder.Wherein g is the number of support vector.

Figure 3 shows that control process scheme drawing based on the abs controller of LS-SVM.

1, the wheel speed sensors of the ECU of automobile ABS wheel speed signal in the line drawing Motor Vehicle Braking Procedure carries out speed of a motor vehicle estimation by sending into input layer behind the sqignal conditioning filter, calculates the automotive wheel angular velocity omega.The sampling period of wheel speed sensors is Δ T, according to the angular speed of wheel that two neighbouring samples are gathered constantly, utilizes method of finite difference to calculate wheel angular acceleration ω.If gather the angular speed of wheel of automobile constantly at n

, n-1 gathers the angular speed of wheel of automobile constantly

, algoritic module calls formula in the abs controller

Calculate n wheel angular acceleration constantly

2, calculate the vehicle body reference velocity according to the wheel speed peak signal.Constantly wheel speed is compared, if the wheel speed of certain sampling instant greater than the wheel speed of last sampling instant and the wheel speed of back one sampling instant, determines that then the wheel speed in this moment is the peak value wheel speed.The slope that calculates adjacent two wheel speed peak value lines obtains vehicle body reference deceleration degree, if n-1 and n wheel speed peak value constantly is respectively

With , Δ T _nIt is the wheel speed peak value

Apart from the wheel speed peak value Between time, then n and n+1 constantly between the reference deceleration degree of automobile $v^{\′}=\frac{v_{\mathrm{wn}}-v_{\mathrm{wn}-1}}{\mathrm{\Δ}{T}_n},$ Then at the vehicle body reference speed in next sampling period $v_{w_{n+1}}=v_{w_n}+v^{\′}\mathrm{\ΔT}$ 。

According to reference velocity ν, automotive wheel angular deceleration ω, controller algorithm module invokes slip rate formula $S=\frac{v-\mathrm{r\ω}}{v}\×100\%$ Calculate the reference slip rate of 4 wheels of automobile.Wherein r is a radius of wheel.

Wheel angular deceleration and wheel reference slip rate as the input based on the LS-SVM controller, are calculated the expectation control presssure (being the pressure of brake wheel cylinder) of wheel.According to control presssure, adopt electromagnetic valve make-and-break time in the pulse width modulation controlled Automobile ABS Controller, make the pressure of brake wheel cylinder reach desired pressure value, can be by different systems be carried out related experiment, according to the pressure of wheel cylinder and the functional relation of time, by the control electromagnetic valve make-and-break time, determine the desired pressure value of brake wheel cylinder.And then be controlled near the expectation value proper range with the angular deceleration of wheel with reference to slip rate, realize braking fully.

The inventive method is simple, calculates rapidly, only need utilize the road examination measures under the various road surfaces wheel angular deceleration, with reference to slip rate and brake-pressure as support vector, controller is trained each parameter of controlled device.In emergency braking with the wheel deceleration/decel of each wheel with reference to the input of slip rate as each wheel controller, can calculate the brake-pressure of each wheel expectation by controller, size according to brake-pressure, adopt the make-and-break time of pulse duration modulation method control electromagnetic valve, and then control brake pressure, realize safety arrestment.This method has overcome traditional control method needs experience to carry out parameter matching and the long shortcoming of time-to-market mostly, and this method does not need the identification information of road surface in control process simultaneously, has avoided the misoperation of control algorithm.

Claims (3)

1, a kind of vehicle anti-lock brake control method based on least square method supporting vector machine is characterized in that, comprises the steps:

(1) pressure signal of sensor automobile wheel speed signal and brake wheel cylinder in the line drawing braking procedure is trained the regression parameter a and the b of controlled device as support vector to controller, utilizes method of finite difference to obtain the angular deceleration ω of wheel;

(2) slope according to two adjacent wheel speed peak of curve lines obtains vehicle body reference deceleration degree V, and obtains the vehicle body reference speed thus;

(3) calculate the reference slip rate of each wheel according to vehicle body reference velocity and wheel angular deceleration;

(4) with wheel angular deceleration and wheel reference slip rate as input, make up least square method supporting vector machine LS-SVM controller, calculate the desired braking pressure of each wheel;

(5) according to the desired braking pressure of each wheel, adopt the make-and-break time of the method control electromagnetic valve of pulse duration modulation, thereby the brake-pressure of controlling each wheel reaches desired braking pressure, realize braking fully.

2, method according to claim 1 is characterized in that, according to the positive and negative relationship change of wheel angular deceleration, determines the wheel speed peak of curve; Controller algorithm module invokes formula $S=\frac{v-\mathrm{r\ω}}{v}\×100\%$ Calculate the reference slip rate of each wheel of automobile; If n-1 and n wheel speed peak value constantly is respectively

With

The time is Δ T between the two wheel speed peak values _n, then according to formula $v^{\′}=\frac{v_{\mathrm{wn}}-v_{\mathrm{wn}-1}}{\mathrm{\Δ}{T}_n}$ Calculate the reference deceleration degree of n and n+1 automobile between the moment.

3, method according to claim 1, it is characterized in that, described least square method supporting vector machine LS-SVM controller is the three layer model structure, comprise, input layer, hidden layer, output layer, described input layer is an incoming signal with wheel reference slip rate s and wheel angular deceleration ω ', makes up the input function relation of x=(s, ω '); Hidden layer utilizes kernel function K (x, x _i)=exp (|| x-x _i|| ²/ σ ²) two-dimentional input vector and support vector are carried out kernel operation; Output layer is according to formula $f\left(x\right)=\underset{i=1}{\overset{g}{\mathrm{\Σ}}}{a}_{i}K(x,x_i)+b=P$ Make up the desired braking pressure P of wheel and the relation of kernel function.

Images