CN105774776A - Automobile and pedestrian anti-collision intelligent control system and method based on pedestrian and automobile cooperation - Google Patents

Abstract

The invention provides an automobile and pedestrian anti-collision intelligent control system and method based on pedestrian and automobile cooperation. The control system is provided with an automobile-mounted sensor, a road information acquisition module, a microprocessor, an early warning device, a dangerous zone difference module, an expected acceleration generating module and a braking control module. The control method includes the steps that information of an automobile and surroundings of the automobile is acquired, and information characteristics of pedestrians in front of the automobile are extracted; a safe distance model of the automobile and the pedestrians is established, dangerous zone judging and differentiating criteria are designed, an early warning is given in an early warning zone to remind a driver, and auxiliary braking control is conducted in a dangerous zone; based on the longitudinal motion characteristics of the automobile and the pedestrians, a pedestrian and automobile interval longitudinal kinetic model and an automobile acceleration response model are established, and a pedestrian and automobile coupling kinetic model is comprehensively generated; the expected acceleration needed for avoiding pedestrian and automobile collisions is dynamically planned in real time; and a nerve fuzzy sliding formwork braking controller is designed, commands for braking pressure are adjusted to complete tracking of the expected acceleration, and active protection and anti-collision conducted by the automobile to the pedestrians are achieved.

Description

A kind of automobile pedestrian anticollision intelligence control system collaborative based on people's car and method

Technical field

The invention belongs to automobile active safety and auxiliary driving field, particularly relate to a kind of automobile pedestrian anticollision intelligence control system collaborative based on people's car and method.

Background technology

Vehicle-pedestrian impact accident is one of main Types of road traffic accident, is also the principal element causing pedestrian disadvantaged group to be injured.How to develop and design and actively protect pedestrian's CAS, avoid automobile and pedestrian collision's accident, it is achieved the active of running car front pedestrian is protected as far as possible, improve road safety, there is actual application value widely.

Document 1 (BoLi; etc.PedestrianDetectionBasedonClusteredPoseletModelsandH ierarchicalAnd-OrGrammar [J] .IEEETransactiononVehicularTechnology; 2015,64 (4): 1435-1444.) pedestrian detection based on stratificational grammar model and guard method are proposed.Document 2 (WadaTomotaka, etc.PedestrianOrientedVehicularCollisionAvoidanceSupport System:P-VCASS [J] .IeiceTransactionsonFundamentalsofElectronicsCommunicati ons&ComputerSciences, devise the crashproof support system that based on pedestrian feature and minimum pedestrian terminal consume 2010,93-A (4): 679-688.).But, intercouple between deathtrap one skilled in the art and vehicle, influence each other, and the non-linear design difficulty considerably increasing the crashproof active control system of pedestrian of the stochastic uncertainty of pedestrian movement's feature and longitudinal direction of car vehicle dynamics.Therefore difficult point and focus that the pedestrian's anti-collision control system being devoted to overcome above-mentioned characteristic is the research of future automobile intelligent and safe control loop how are designed.

Summary of the invention

It is an object of the invention to solve the above-mentioned problems in the prior art; offer can effectively overcome the stochastic uncertainty of pedestrian movement's feature and the non-linear of vehicle vehicle dynamics; quickly realize automobile Braking mode under dangerous working condition to control, actively protect a kind of automobile pedestrian anticollision intelligence control system collaborative based on people's car of pedestrains safety.

Another object of the present invention is to provide and a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car.

Described a kind of automobile pedestrian anticollision intelligence control system collaborative based on people's car is provided with onboard sensor, road information acquisition module, microprocessor, precaution device, deathtrap difference module, expectation acceleration generation module, brake control module；

Described onboard sensor is located on automobile, onboard sensor connects the input of road information acquisition module, the input of the output termination microprocessor of road information acquisition module, pedestrian's feature analysis of microprocessor processes signal output part and connects the input of precaution device and the input of deathtrap difference module respectively, when detection pedestrian is when pre-police region, the danger signal outfan of precaution device sends early warning signal prompting driver, when detection pedestrian is when hazardous area, the input of the danger signal output termination expectation acceleration generation module of deathtrap difference module, expect the input of the output termination brake control module of acceleration generation module, the outfan of brake control module is the tracking control signal of automobile output expectation acceleration, complete actively to protect pedestrian.

Described a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car, comprises the following steps:

1) gather automobile and ambient condition information thereof, extract running car front pedestrian information feature；

2) set up automobile and pedestrains safety distance model, design deathtrap criterion, carry out early warning driver in pre-police region, in hazardous area, then carry out auxiliary braking control；

3) based on automobile and pedestrian's lengthwise movement feature, people's following distance Longitudinal Dynamic Model and pickup response model are set up, comprehensive generation people's car Coupling Dynamic Model；

4) adopting prediction optimization method, Real-time and Dynamic cooks up the expectation acceleration avoided needed for automobile and pedestrian collision；

5) design fuzzy neuron sliding formwork brake monitor, the instruction being regulated brake pressure by this controller completes the tracking to expectation acceleration, it is achieved the active of pedestrian is protected and crashproof by automobile.

In step 1) in, described collection automobile and ambient condition information thereof, the concrete grammar extracting running car front pedestrian information feature can be:

(1) fore-and-aft distance information and pedestrian's velocity information, the Negotiation speed encoder collection vehicle travel speed of automobile and pedestrian is obtained by vehicle-mounted millimeter wave radar sensor；

(2) utilize monocular ccd video camera to gather environment surrounding automobile information, by vehicle-mounted microprocessor, the image gathered processed and feature extraction, detect in real time and know vehicle front reliably, pedestrian's characteristic information accurately.

In step 2) in, described automobile and the pedestrains safety distance model set up, design deathtrap criterion, carry out early warning driver in pre-police region, the concrete grammar then carrying out auxiliary braking control in hazardous area can be:

(1) foundation comprises driver's subjectivity Safety distance model that after driver identifies traffic conditions make a response required distance and parking of automobile, between the desired Pedestrians and vehicles of driver, static distance forms；

(2) according to the automobile obtained and the actual fore-and-aft distance information of pedestrian, design deathtrap criterion, the road area of vehicle front is divided into prewarning area and deathtrap, in pre-police region, carries out early warning driver, in hazardous area, then carry out auxiliary braking control.

In step 3) in, described based on automobile and pedestrian's lengthwise movement feature, set up people's following distance Longitudinal Dynamic Model and pickup response model, comprehensive generate people's car Coupling Dynamic Model concrete grammar can be:

(1) the relative distance deviation delta d with automobile Yu front pedestrian is set up, relative speed Δ v is people's car Longitudinal Dynamic Model of state variable, adopt the dynamic response model of least squares identification automobile longitudinal acceleration, comprehensive structure people's car Coupling Dynamic Model；

(2) by sampling period 0.15s, people's car Coupling Dynamic Model is carried out sliding-model control, the state equation of discrete linear systems can be obtained.

In step 4) in, described employing prediction optimization method, Real-time and Dynamic is cooked up the concrete grammar of the expectation acceleration needed for avoiding automobile and pedestrian collision and can is:

(1) design safety performance target function: crashproof in order to realize quick, safe pedestrian, using two norms of people car longitudinal pitch error delta d and people car relative velocity error delta v as safe performance indexes L, as follows:

L=ω₁Δd²+ω₂Δv²

In formula, ω₁And ω₂Represent the weight coefficient of range error and velocity error；

(2) the prediction form of performance indications discretization in prediction time domain is set up；

(3) set up the quadratic programming type of prediction optimization, adopt active set m ethod to solve this prediction optimization problem, calculate the expectation acceleration required for pedestrian's avoidance under deathtrap in real time.

In step 5) in, described design fuzzy neuron sliding formwork brake monitor, the instruction being regulated brake pressure by this controller completes the tracking to expectation acceleration, it is achieved automobile is protected by the active of pedestrian and crashproof concrete grammar can be:

(1) for the non-linear of dynamics of vehicle and parameter uncertainty, adopting neural network sliding mode control method, the brake pressure neural network sliding mode control that goes that design is made up of equivalent control and variable-structure control is restrained, it is achieved the tracing control to expectation acceleration.

(2) the neural network sliding mode control rule obtained for previous step, adopts quasisliding mode method, introduces saturation function, the fuzzy control logic that plan boundary layer " broadens in real time and narrows ", effectively weakens the buffeting that neural Sliding mode variable structure control causes.

The present invention first passes through onboard sensor and microprocessor gathers automobile itself and ambient condition information, extract front pedestrian's feature, and analyze and process, signal transmission to deathtrap after processing is differentiated and in warning module, pedestrian is in pre-police region in detection, then carry out early warning driver, pedestrian is in hazardous area in detection, then by danger signal transmission to expectation acceleration generation module, pass through prediction optimization, dynamic programming goes out the expectation acceleration of automobile and pedestrian's collision free, the tracing control of expectation acceleration is realized finally by brake control module, complete actively to protect pedestrian.

The present invention takes into full account the coupled characteristic between pedestrian and vehicle; differentiate by pedestrian's feature extraction, deathtrap, expect acceleration dynamic programming and fuzzy neuron sliding formwork control for brake; effectively overcome the stochastic uncertainty of pedestrian movement's feature and the non-linear of vehicle vehicle dynamics; quickly realize automobile Braking mode under dangerous working condition to control, actively protect pedestrains safety.

Accompanying drawing explanation

Fig. 1 is the structural representation of the automobile pedestrian anticollision intelligence control system embodiment collaborative based on people's car of the present invention.

Fig. 2 is automobile and the longitudinally opposed position view of pedestrian of the present invention.

Detailed description of the invention

Following example will the present invention is further illustrated in conjunction with accompanying drawing.

As it is shown in figure 1, a kind of automobile pedestrian anticollision intelligence control system embodiment collaborative based on people's car of the present invention is provided with onboard sensor 2, road information acquisition module 3, microprocessor 4, precaution device 5, deathtrap difference module 6, expectation acceleration generation module 7, brake control module 8.

Described onboard sensor 2 is located on automobile 1, onboard sensor 2 connects the input of road information acquisition module 3, the input of the output termination microprocessor 4 of road information acquisition module 3, pedestrian's feature analysis of microprocessor 4 processes signal output part and connects the input of precaution device 5 and the input of deathtrap difference module 6 respectively, when detection pedestrian is when pre-police region, the danger signal outfan of precaution device 5 sends early warning signal prompting driver, when detection pedestrian is when hazardous area, the input of the danger signal output termination expectation acceleration generation module 7 of deathtrap difference module 6, expect the input of the output termination brake control module 8 of acceleration generation module 7, the outfan of brake control module 8 is the tracking control signal of automobile 1 output expectation acceleration, complete actively to protect pedestrian.

Described a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car, comprises the following steps:

Step 1: detection running car front pedestrian, obtains automobile oneself state, automobile and pedestrian's fore-and-aft distance information.

The first step: gather running car front information by vehicle-mounted monocular ccd video camera, utilizes vehicle-mounted microprocessor that the image gathered is processed and feature extraction, detect in real time and know vehicle front reliably, pedestrian's characteristic information accurately.

Second step: obtained fore-and-aft distance information and pedestrian's velocity information of automobile and pedestrian by millimetre-wave radar sensor.By vehicular speeds encoder collection vehicle longitudinal driving speed.

Step 2: according to automobile and the longitudinally opposed positional information of pedestrian, set up Safety distance model, carry out deathtrap differentiation, design anticollision early warning mechanism.

The first step: the automobile obtained according to step 1 and pedestrian's relative distance positional information, is divided into prewarning area and deathtrap by the road area of vehicle front.

Second step: set up the minimum safetyspacing model describing driver intention by statistical analysis method, identifies after traffic conditions make a response required distance and parking of automobile static distance between the desired Pedestrians and vehicles of driver including driver, as follows:

d_des=τ_hv_f+d₀

In formula, d_desRepresent minimum safetyspacing model, τ_hRepresent people Che Shi from, value is 1.2～2.0s, d₀Represent static distance between the desired pedestrian of driver and vehicle, d₀Value 2～5m, v_fFor vehicular longitudinal velocity.

3rd step: the minimum safetyspacing model obtained by above-mentioned steps, is differentiated the safe condition of pedestrian and vehicle, it determines according to being expressed as

$\left\{\begin{array}{c}d>d_a\\ d_{des}<d<d_a\\ d\&le;d_{des}\end{array}\right.$

In formula, d represents the actual range between automobile and pedestrian, d_aRepresent maximum safe distance.

If the actual range d between automobile and pedestrian is more than the maximum safe distance d set_a, then car is in a safe condition.

If the actual range d between automobile and pedestrian is more than minimum safe distance d_desWith less than maximum safe distance d_a, then people's car is in prewarning area, and human pilot and pedestrian need to be sent alarm signal by system.

If the actual range d between automobile and pedestrian is less than minimum safe distance d_des, pedestrian and automobile are in deathtrap, it is necessary to carry out intelligence auxiliary braking and control.

Step 3: setting up pedestrian/longitudinal direction of car Coupling Dynamic Model, pedestrian/vehicle Coupling Dynamic Model depends on people's following distance kinetic model and pickup response model.

The first step: as in figure 2 it is shown, x and x_pRepresenting vehicle and front pedestrian's lengthwise position coordinate respectively, definition Δ d is the automobile relative distance deviation relative to front pedestrian, and Δ v is automobile vehicle speed deviation relative to front pedestrian, and people's following distance kinetic model is:

Δ d=d_des-d

Δ v=v_p-v_f

Wherein, v_pFor front pedestrian's speed, v_fFor vehicular longitudinal velocity.

Second step: adopt the one order inertia transfer function model of least squares identification automobile longitudinal acceleration dynamic response, as follows:

$a_f=\frac{K}{Ts+1}{a}_{fdes}$

In formula, K represents that gain coefficient, T represent and represents time delay, s variable, a_fAutomobile longitudinal acceleration, a_fdesRepresent automobile expectation acceleration.

3rd step: comprehensive people-following distance kinetic model and pickup dynamic response model, obtains people's car Longitudinal data kinetic model, is expressed as state-space expression equation form:

$\stackrel{\&CenterDot;}{x}=A_{1}x+B_{1}u+G_{1}v$

Y=C₁x+w

Wherein, x=[Δ d Δ va_f]^TFor system mode vector, y is system output vector, u=a_desFor controlling input vector, v=a_pFor input nonlinearities vector, w is output interference vector, A₁、B₁And G₁For input coefficient matrix, C₁For output factor matrix.

$A_1=\left[\begin{array}{ccc}0& 1& -{\mathrm{\&tau;}}_h\\ 0& 0& -1\\ 0& 0& -1/T\end{array}\right];B_1=\left[\begin{array}{c}0\\ 0\\ K/T\end{array}\right]:C_1=\left[\begin{array}{ccc}1& 0& 0\\ 0& 1& 0\\ 0& 0& 1\end{array}\right]$

G₁=[010]^T

4th step: the continuous time system model set up by the 3rd step is carried out sliding-model control, by sampling period 0.15s, people's car Coupling Dynamic Model is carried out sliding-model control, the state equation of discretization linear system can be obtained:

X (k+1)=Ax (k)+Bu (k)+Gv (k)

Y (k)=Cx (k)+w (k)

Wherein

A=T_sA₁+I；B=T_sB₁；G=T_sG₁

In formula, A, B and G is the coefficient matrix of discrete state equations, and C is the output factor matrix of discrete state equations, T_sRepresent the sampling period.

Step 4: couple Longitudinal Dynamic Model for people's car, adopts prediction optimization control method, and Real-time and Dynamic is cooked up automobile in deathtrap and avoided the expectation acceleration with pedestrian collision.

The first step: design pedestrian anti-collision safe performance indexes function, crashproof in order to realize pedestrian quickly and safely, using two norms of people's following distance error and people's vehicle speed error as safe performance indexes function L:

L=ω₁Δd²+ω₂Δv²

In formula, ω₁And ω₂Represent the weight coefficient of range error and velocity error, adopt experience standardizition to obtain.

Second step: set up the prediction form of safe performance indexes function L:

$L=\underset{i=1}{\overset{P}{\&Sigma;}}\left|\right|\mathrm{\&Delta;}d(k+i+1|k)|{|}_{{\mathrm{\&omega;}}_1}^2+\underset{i=1}{\overset{P}{\&Sigma;}}|\left|\mathrm{\&Delta;}v(k+i+1|k)\right|{|}_{{\mathrm{\&omega;}}_2}^2$

Wherein, k represents current time, and (k+i+1 | k) represents and utilize k time information that k+i+1 state is predicted, and P is prediction time domain length.

3rd step, sets up the quadratic programming type of prediction optimization, adopts active set m ethod to solve prediction optimization problem, calculates the expectation acceleration required for pedestrian's avoidance under deathtrap in real time.Prediction optimization problem is as follows:

$\underset{i=0:P-1}{min}L=\underset{i=1}{\overset{P}{\&Sigma;}}\left|\right|\mathrm{\&Delta;}d(k+i+1|k)|{|}_{{\mathrm{\&omega;}}_1}^2+\underset{i=1}{\overset{P}{\&Sigma;}}|\left|\mathrm{\&Delta;}v(k+i+1|k)\right|{|}_{{\mathrm{\&omega;}}_2}^2$

$\begin{array}{cc}s.t.& \left\{\begin{array}{c}x(k+1)=Ax\left(k\right)+Bu\left(k\right)+Gv\left(k\right)\\ y\left(k\right)=Cx\left(k\right)+w\left(k\right)\\ u_{\min }\&le;u(k+i|k)\&le;u_{\max }\\ {\mathrm{\&Delta;u}}_{\min }\&le;\mathrm{\&Delta;}u(k+i|k)\&le;{\mathrm{\&Delta;u}}_{max}\\ i=0:P-1\end{array}\right.\end{array}$

In formula, u_minControl input lower bound, u_maxControl the input upper bound, Δ u_minFor controlling increment lower bound, Δ u_maxFor the controlling increment upper bound.

Step 5: design fuzzy neuron sliding formwork brake monitor, the instruction being regulated brake pressure of automobile by this controller realizes the tracing control to expectation acceleration, completes automobile the active of pedestrian is crashproof.

Step 5.1: characteristic when travelling according to longitudinal direction of car, sets up the vehicle overall design model describing Railway Cars under Braking Working Conditions transmission, running gear and car load motor system model:

$\left\{\begin{array}{c}\stackrel{\&CenterDot;}{v}=\frac{1}{J}(\frac{T_{e\min }}{r}-\frac{T_b}{r}-Mgf\cos \mathrm{\&theta;}-C_d{A}_a{v}^2-Mg\sin \mathrm{\&theta;})+\mathrm{\&Delta;}E\\ {\mathrm{\&tau;}}_b{\stackrel{\&CenterDot;}{T}}_b+T_b=K_p\&CenterDot;P_b\end{array}\right.$

In formula, T_eminRepresenting accelerator open degree is the minimum output moment of torsion of electromotor when zero, T_bRepresent braking moment,Representing road grade, r represents radius of wheel, and J is equivalent moment of inertia, and f represents that coefficient of rolling resistance, M represent car mass, and g is acceleration of gravity, C_dRepresent air resistance coefficient, A_aRepresent equivalence front face area, P_bFor brake pressure, K_pFor brake pressure proportionality coefficient, τ_bBrake system response lag time, Δ E is bounded indeterminate.

Step 5.2: have the characteristics such as parameter uncertainty, non-linear and external disturbance for vehicle overall design system, adopts Equivalent control law and the variable-structure control rule of neural network sliding mode control method design braking moment.

Step 5.2.1: acceleration maker output expectation acceleration is a_des, vehicular speeds sensor measurement actual acceleration is a, then expectation acceleration a_desWith actual acceleration deviation e it is:

E=a_des-a

Step 5.2.2: first determine slip curved surface S:

$S=e+\mathrm{\&lambda;}{\&Integral;}_0^{t}edt$

In formula, λ represents sliding-mode surface coefficient.

If reaching desirable Variable Structure Control, need to meet:

$\frac{dS}{dt}={\stackrel{\&CenterDot;}{a}}_{des}-\stackrel{\&CenterDot;}{a}+\mathrm{\&lambda;}(a_{des}-a)=0$

When sliding mode motion reaches desirable terminal, meet S=0 andThe equivalent control of system braking moment in Fault slip rate can be obtained by formula.

Step 5.2.4: adopt neural network sliding mode control method, Longitudinal Dynamic Model is brought into expectation sliding mode:

$\stackrel{\&CenterDot;}{e}+{\mathrm{\&lambda;a}}_{des}-\frac{\mathrm{\&lambda;}}{Jr}{T}_{e,min}+\frac{\mathrm{\&lambda;}}{Jr}{T}_b+\frac{\mathrm{\&lambda;}}{J}\mathrm{\&xi;}=0$

Wherein, ξ represents time-varying Uncertain nonlinear function.

Obtain braking moment Equivalent control law:

In formula, T_b,eqRepresent equivalent braking force square,Represent nonlinear function estimated value, ε₁For the approximate error of network, W₁ ^TFor the weight vector of neutral net, h₁(x₁)=[h_1j]^TGaussian bases for network is vectorial, is expressed as follows:

$h_{1j}=\exp \left(\frac{\left|\right|x_1-c_{1j}|{|}^2}{2b_{1j}^2}\right)$

In formula, x₁For the input signal of network, j is network hidden layer jth node, c_1jFor the center vector of jth node, b_1jNeuron width for jth node.

Step 5.2.4: definition sliding mode curves S₁=S, design variable-structure control rule is as follows:

T_b,vs=K₁sgn(S₁)

Wherein T_b,vsRepresent and restrained the braking moment obtained, K by variable-structure control₁Represent variable-structure control gain coefficient, andIts dr₁The upper bound for Δ E.

Step 5.2.5: comprehensive Equivalent control law and variable-structure control rule, obtaining desired braking Torque Control rule is:

T_b,des=T_b,eq+T_b,vs

In formula, T_b,desFor expectation braking moment.

Step 5.3: for the dynamic response characteristic of brake pressure of automobile, according to the desired braking moment that step 5.2 is obtained, derives the desired braking pressure needed for automobile.

Step 5.3.1: for brake pressure response characteristic, the deviation e ' of defining ideal braking moment and actual braking force square:

E '=T_b,des-T_b

The desired braking moment defined based on above formula and the control deviation of actual moment, define sliding mode curves S₂:

S₂=e '

Step 5.3.2: for the sliding mode curves S of step 5.3.1 definition₂, adopt neural network sliding mode control method, trying to achieve the braking pressure control rule being made up of equivalent control and variable-structure control is:

$P_p=\frac{1}{K_p}({\mathrm{\&tau;}}_b{\stackrel{\&CenterDot;}{T}}_{b,des}+T_b+{\mathrm{\&tau;}}_b{K}_2\mathrm{sgn}(S_2\left)\right)$

In formula,For expectation braking moment variation rate, K₂Represent variable-structure control gain coefficient, W₂ ^TFor the weight vector of neutral net, h₂(x₂)=[h_2j]^TGaussian bases output vector for network:

$h_{2j}=\exp \left(\frac{\left|\right|x_2-c_{2j}|{|}^2}{2b_{2j}^2}\right)$

In formula, x₂For the input signal of network, c_2jFor the center vector of jth node, b_2jNeuron width for node.

Step 5.4: adopt and weaken, based on the adaptive boundary coating control method of fuzzy logic adjustment, the automobile longitudinal neural network sliding mode control that step 4 derives and restrain the buffeting problem caused.

Step 5.4.1: for weakening the buffeting of sliding moding structure, adopt quasisliding mode method, introduces saturation function sat (S in the neural network sliding mode control of braking moment and brake pressure is restrained_i/Δ_i) replace switching function sgn (S_i), i=1,2.

Step 5.4.2: according to sliding-mode surface S_iSize, following boundary region width Delta_iTend under the principle being gradually reduced in steady-state process at system motion, the fuzzy control actuator that plan boundary layer " broadens in real time and narrows ", the width Delta of sliding formwork boundary region_iAccording to sliding-mode surface state S_iSize dynamically regulate.

The present invention includes the collaborative expectation acceleration dynamic programming process of people's car and auxiliary braking control design case process.First, gather automobile self and ambient condition information by onboard sensor and microprocessor, extract front pedestrian's feature, and analyze and process.Secondly, signal transmission to the deathtrap after process is differentiated and in warning module, it is judged that pedestrian in pre-police region, then carries out early warning driver, it is judged that pedestrian is in hazardous area, then by danger signal transmission extremely expectation acceleration dynamic programming module.Then, by prediction optimization, Real-time and Dynamic cooks up the automobile expectation acceleration avoiding Vehicle-pedestrian impact.Finally, the tracing control to expectation acceleration is completed by fuzzy neuron sliding formwork brake control module, it is achieved actively pedestrian protecting.

Above content is the further description present invention done in conjunction with optimal technical scheme.

Claims (7)

1. the automobile pedestrian anticollision intelligence control system worked in coordination with based on people's car, it is characterised in that be provided with onboard sensor, road information acquisition module, microprocessor, precaution device, deathtrap difference module, expectation acceleration generation module, brake control module；

Described onboard sensor is located on automobile, onboard sensor connects the input of road information acquisition module, the input of the output termination microprocessor of road information acquisition module, pedestrian's feature analysis of microprocessor processes signal output part and connects the input of precaution device and the input of deathtrap difference module respectively, when detection pedestrian is when pre-police region, the danger signal outfan of precaution device sends early warning signal prompting driver, when detection pedestrian is when hazardous area, the input of the danger signal output termination expectation acceleration generation module of deathtrap difference module, expect the input of the output termination brake control module of acceleration generation module, the outfan of brake control module is the tracking control signal of automobile output expectation acceleration, complete actively to protect pedestrian.

2. the automobile pedestrian anticollision intelligent control method worked in coordination with based on people's car, it is characterised in that comprise the following steps:

1) gather automobile and ambient condition information thereof, extract running car front pedestrian information feature；

2) set up automobile and pedestrains safety distance model, design deathtrap criterion, carry out early warning driver in pre-police region, in hazardous area, then carry out auxiliary braking control；

3) based on automobile and pedestrian's lengthwise movement feature, people's following distance Longitudinal Dynamic Model and pickup response model are set up, comprehensive generation people's car Coupling Dynamic Model；

4) adopting prediction optimization method, Real-time and Dynamic cooks up the expectation acceleration avoided needed for automobile and pedestrian collision；

5) design fuzzy neuron sliding formwork brake monitor, the instruction being regulated brake pressure by this controller completes the tracking to expectation acceleration, it is achieved the active of pedestrian is protected and crashproof by automobile.

3. a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car as claimed in claim 2, it is characterised in that in step 1) in, described collection automobile and ambient condition information thereof, extract running car front pedestrian information feature method particularly includes:

(1) fore-and-aft distance information and pedestrian's velocity information, the Negotiation speed encoder collection vehicle travel speed of automobile and pedestrian is obtained by vehicle-mounted millimeter wave radar sensor；

(2) utilize monocular ccd video camera to gather environment surrounding automobile information, by vehicle-mounted microprocessor, the image gathered processed and feature extraction, detect in real time and know vehicle front reliably, pedestrian's characteristic information accurately.

4. a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car as claimed in claim 2, it is characterized in that in step 2) in, described automobile and the pedestrains safety distance model set up, design deathtrap criterion, carry out early warning driver in pre-police region, in hazardous area, then carry out auxiliary braking control method particularly includes:

(1) foundation comprises driver's subjectivity Safety distance model that after driver identifies traffic conditions make a response required distance and parking of automobile, between the desired Pedestrians and vehicles of driver, static distance forms；

(2) according to the automobile obtained and the actual fore-and-aft distance information of pedestrian, design deathtrap criterion, the road area of vehicle front is divided into prewarning area and deathtrap, in pre-police region, carries out early warning driver, in hazardous area, then carry out auxiliary braking control.

5. a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car as claimed in claim 2, it is characterized in that in step 3) in, described based on automobile and pedestrian's lengthwise movement feature, set up people's following distance Longitudinal Dynamic Model and pickup response model, comprehensive generation people's car Coupling Dynamic Model method particularly includes:

(1) the relative distance deviation delta d with automobile Yu front pedestrian is set up, relative speed Δ v is people's car Longitudinal Dynamic Model of state variable, adopt the dynamic response model of least squares identification automobile longitudinal acceleration, comprehensive structure people's car Coupling Dynamic Model；

(2) by sampling period 0.15s, people's car Coupling Dynamic Model is carried out sliding-model control, obtain the state equation of discrete linear systems.

6. a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car as claimed in claim 2, it is characterized in that in step 4) in, described employing prediction optimization method, Real-time and Dynamic is cooked up and is avoided expectation acceleration needed for automobile and pedestrian collision method particularly includes:

(1) design safety performance target function: crashproof in order to realize quick, safe pedestrian, using two norms of people car longitudinal pitch error delta d and people car relative velocity error delta v as safe performance indexes L, as follows:

L=ω₁Δd²+ω₂Δv²

In formula, ω₁And ω₂Represent the weight coefficient of range error and velocity error；

(2) the prediction form of performance indications discretization in prediction time domain is set up；

(3) set up the quadratic programming type of prediction optimization, adopt active set m ethod to solve this prediction optimization problem, calculate the expectation acceleration required for pedestrian's avoidance under deathtrap in real time.

7. a kind of automobile pedestrian anticollision intelligent control method collaborative based on people's car as claimed in claim 2; it is characterized in that in step 5) in; described design fuzzy neuron sliding formwork brake monitor; the instruction being regulated brake pressure by this controller completes the tracking to expectation acceleration, it is achieved the active of pedestrian is protected and crashproof by automobile method particularly includes:

(1) for the non-linear of dynamics of vehicle and parameter uncertainty, adopting neural network sliding mode control method, the brake pressure neural network sliding mode control that goes that design is made up of equivalent control and variable-structure control is restrained, it is achieved the tracing control to expectation acceleration；

(2) the neural network sliding mode control rule obtained for previous step, adopts quasisliding mode method, introduces saturation function, the fuzzy control logic that plan boundary layer " broadens in real time and narrows ", effectively weakens the buffeting that neural Sliding mode variable structure control causes.