CN114371691B - Tracking control method for auxiliary driving curve track - Google Patents

Abstract

The invention discloses a tracking control method for assisting in driving curve tracks, which comprises the following steps: the environment sensing unit identifies a lane track line; the ECU calculates the speeds of different positions of the curved path in real time according to an algorithm; the tracking controller tracks the planning speed and the road track; the speed planning method comprises the following steps: the method comprises the steps of adopting a fuzzy control method, taking lane line curvature, road attachment coefficient and road track tracking deviation as fuzzy input, and outputting speed curves of different points of a curve; the tracking control method comprises the following steps: by simulating the brain emotion memory learning process, a brain emotion tracking controller tracking speed curve and a track curve are designed. The invention improves the tracking precision of the curve track of the vehicle and the robustness of the tracking process, carries out speed planning on the curve running in advance, can effectively reduce the unstable condition of the vehicle, ensures the curve running process to be safer and more stable, and applies double insurance to the curve running safety.

Description

Tracking control method for auxiliary driving curve track

Technical Field

The invention relates to the technical field of auxiliary driving, in particular to an auxiliary driving curve track tracking control method.

Background

With the rapid development of the automobile industry, communication, computer and other technologies, intelligent driving automobiles are no longer far from reach, and have been taken into people's lives. The problem of intelligent driving vehicle safety is always described, and intelligent vehicle active safety technology gradually becomes a hot spot of current research.

The core of the unmanned system can be roughly divided into three parts of perception recognition, decision planning and behavior control. As a final step in achieving autopilot, a behavior control module is critical. The core of the behavior control module is the design of the controller, which can be seen as the ability of the system to track the desired trajectory, the tracking controller needs to have a certain anti-interference ability in view of the strong nonlinearity of the vehicle system and the interference situation during operation. The control layer tracking controller has various designs and different application algorithms, and adopts classical control theory such as PID control, modern control theory such as linear quadratic control, synovial membrane control, robust control, MPC and the like, and research on more intelligent control theory such as fuzzy control, neural network control, pretightening following theory and control strategies combining different algorithms.

The effective auxiliary driving track tracking control system needs to have reasonable sensor arrangement and a tracking control strategy which takes stability and accuracy into consideration. At present, most of methods used in researches on track tracking control are combined with an electronic stabilizing system of a vehicle body, and the coupling factor of transverse and longitudinal control during track tracking is the running speed of the vehicle; in the track tracking process, when the vehicle keeps the original speed to enable the vehicle to be critically unstable, the vehicle body electronic stability system is used for interventional control of the vehicle body to be stable, but the instability conditions such as sideslip, rollover and the like of the vehicle still occur when the speed is too high; therefore, the speed of the track tracking process is required to be planned in advance, the track tracking running efficiency of the curve is improved, and the safety is ensured.

Disclosure of Invention

The invention aims to: the invention aims to solve the technical problem of providing an auxiliary driving curve track tracking control method aiming at the defects in the background technology.

The technical scheme is as follows:

a method for tracking and controlling the track of an assisted driving curve comprises the following steps:

step 1), a camera collects lane line information, obstacle information and lane speed limit information of a lane at the current moment k, and the curvature kappa of a lane line of a lane where a recognized own vehicle is located _k =1/R, road track tracking deviation e _k ＝d _{real k} -d _{desire k} Speed limit V of lane _{lim k} Transmitting to CAN bus for VCU to fetch for calculation and judgment; l and R are the length and the radius of the lane line acquired by the camera at the moment k; d, d _{real k} For the true deviation of the longitudinal axis of the vehicle from the lane line at time k, d _{desire k} The expected deviation of the longitudinal axis of the vehicle from the lane line at time k;

step 2), the speed u of the automobile at the moment k is collected by a speed sensor, an acceleration sensor and a front wheel steering angle sensor respectively _{v k} Wheel speed u _{w k} Acceleration alpha _k Front wheel angle delta _k The collected data is input to a CAN bus for VCU calculation;

step 3), calculating the longitudinal force F of the tire by adopting a magic formula tire model _x Calculating road surface braking coefficient mu according to a vehicle dynamics formula _b Estimating the road adhesion coefficient mu at the moment k according to a fuzzy estimation algorithm _k ；

Step 4), VCU according to the received lane line curvature kappa _k Tracking deviation e of road track _k Road adhesion coefficient mu _k Calculating the k-moment planning speed u of the curve at the current k moment by using a fuzzy control algorithm _{plan k} The method comprises the steps of carrying out a first treatment on the surface of the Combining the planning speed at each moment before the moment k to obtain a speed planning curve;

step 5), the camera detects road traffic signs in real time, and when the road speed limit is detected to be V _{lim k} When VCU uses Kalman filtering to smooth speed planning curve and is smaller than road speed limit V _{lim k} If no road speed limit exists at the moment k, no speed limit processing is performed to obtain a smooth and reasonable speed curve；

Step 6), VCU calculates the planning speed u at k moment _{pla n} And the real speed u _{rea l} Deviation e of (2) _{v k} ＝u _plank -u _rea Calculating expected acceleration alpha at k moment through brain emotion learning loop controller _{des k} Judging that the vehicle needs accelerating, decelerating or idling according to a braking driving switching strategy, and finally calculating the automatic pedal pressure and the throttle opening according to an inverse engine model and an inverse brake model respectively so as to control the acceleration and deceleration of the vehicle;

step 7), the camera detects the position of the lane line of the road in real time, and the real deviation d of the longitudinal axis of the vehicle at the moment k and the lane line is obtained _{real k} Transmitting the deviation to a CAN bus, and calculating the expected deviation d of the longitudinal axis of the vehicle and the lane line at the moment k by the VCU in real time _{plan k} From true deviation d _{real k} E of the difference e of (2) _k ＝d _{real k} -d _{plan k} And the deviation value e _k The front wheel angle delta required by the vehicle to track the lane line at the current moment is calculated by inputting the front wheel angle delta into a brain emotion learning loop controller _k Finally, delta is transmitted through a CAN bus _k The input vehicle steer-by-wire unit controls the vehicle to track the lane line in real time;

and 8), repeating the steps 1) -7) in sequence, and calculating the acceleration and the front wheel rotation angle at the moment k+1.

Further, the step 3) specifically includes:

step 3.1) calculating the tire longitudinal force F at the current moment k according to the magic formula tire model _{x k} ：

F _x ＝D _x sin{C _x arctan[B _x λ-E _x (B _x λ-arctan B _x λ)]}

C _x ＝1.62

D _x ＝a ₁ F _z ² +a ₂ F _z

B _x C _x D _x ＝a ₃ sin[a ₄ arctan(a ₅ F _z )](1-a ₁₂ |λ|)

B _x C _x D _x ＝B _x C _x D _x /C _x D _x

E _x ＝a ₆ F _z ² +a ₇ F _z +a ₈

Wherein a is ₁ 、a ₂ 、a ₃ 、a ₄ 、a ₅ 、a ₆ 、a ₇ 、a ₈ 、a ₁₂ Fitting parameters of the magic formula tire, F _z For vertical load of the tyre, x represents longitudinal direction of the vehicle, lambda is slip ratio, D _x Is the peak factor, B _x As a rigidity factor, C _x Is a curve shape factor, E _x Is a curve curvature factor;

step 3.2) fitting six road surfaces through a large amount of experimental data: ice, snow, wet cobble, wet asphalt, dry cement, dry asphalt semi-empirical tire-road mathematical model:

wherein C is ₁ 、C ₂ 、C ₃ Three parameters for the pavement model; then drawing lambda-mu curves of six road surfaces according to the corresponding parameters; λ represents the slip ratio, μ is the road surface peak adhesion coefficient;

step 3.3) calculating the brake force coefficient μ _b And slip ratio λ:

λ＝(u _w -u _v )/u _w

wherein u is _v Indicating the speed of the automobile, u _w Representing the wheel speed of an automobile;

step 3.4) setting a fuzzy controller, wherein the input is mu _b And lambda, output as the similarity s of the current road surface and six road surfaces ₁ 、s ₂ 、s ₃ 、s ₄ 、s ₅ 、s ₆ The method comprises the steps of carrying out a first treatment on the surface of the Then the weighted average method is adopted to calculate the road attachment coefficient mu _k ：

Wherein mu ₁ 、μ ₂ 、μ ₃ 、μ ₄ 、μ ₅ 、μ ₆ The peak adhesion coefficients of the six road surfaces are respectively.

Further, in the step 4), the input of the fuzzy control algorithm is the lane line curvature κ _k Tracking deviation e of road track _k And road adhesion coefficient mu _k According to IF A and B and C then D fuzzy rule, road planning speed u at this time is output _{plan k} The fuzzy rule is:

step 4.1) when the lane line curvature κ _k The speed of the vehicle should be reduced when the vehicle is too large, and the speed of the vehicle should be lower than the speed limit V of the lane _lim Allowing the vehicle to safely pass through a curve with a smaller radius;

step 4.2) when road adhesion coefficient μ _k When the speed is larger, the vehicle speed should be properly increased and the vehicle speed should be lower than the speed limit V of the lane _lim ；

Step 4.3) tracking the deviation e when the track is traced _k When larger, the vehicle speed should be reduced appropriately so that the vehicle re-tracks the lane track over a shorter distance.

Further, in the step 6), the brain emotion learning circuit controller calculates the expected acceleration α _{des k} The method comprises the following steps:

step 6.1), inputting a stimulation signal into the thalamus for deep processing:

wherein the method comprises the steps ofFor inputting the stimulus signal, n represents the number of input stimulus signals,/->A value of a vector of a central point of the jth hidden layer neuron, gamma _j >0 is the width of the gaussian function (j=1, …, m), and m e N is the total number of input signals;

step 6.2), processed signal S _j After entering the sensory cortex, the almond body and the prefrontal cortex enter respectively, and the almond kernel and the prefrontal cortex output are respectively:

wherein V is _j And W is _j (j=1, …, m) are the weight factors of the amygdala and the prefrontal cortex, respectively, weight vector for almond body, ++>Is the weight vector of the orbital cortex; the output of the controller is:

wherein the method comprises the steps ofu is the output alpha of the controller _des ；

Step 6.3), the weight factor updating process is as follows:

wherein S is the signal S after thalamus processing _j F represents an unknown function f to be approximatedx) G represents an unknown function g to be approximatedx)，γ ₁ ～γ ₄ In order to adjust the parameters of the device,for error matrix, e is e _v U is the output alpha of the controller _des P is a positive definite matrix and satisfies Λ ^T P + pΛ = -Q, Q being any n x n order positive definite matrix,b＝[00...01] ^T _n ，V _fk 、V _gk respectively corresponding to the time f of the almond body kx)、g(x) Weight factor, W of _fk 、W _gk Respectively corresponding to the time f of the forehead cortex kx)、g(x) Weight factor of->Is the first derivative;

step 6.4), the control quantity u is the vehicle expected acceleration alpha _des The calculation process of (2) is as follows:

g(x)＝(V _g -W _g )S

u＝g ^-1 (x)[-f(x)+x _d ⁽ⁿ⁾ +K ^T E+u _r ]

wherein S is the signal S after thalamus processing _j ，F is a state variable of the systemx) Is an unknown smooth function and f%x) The bounded is marked as f%x)||≤f<∞，g(x) Is unknown smooth function and g%x) The limit is marked as g%x)||≤g<∞；/>Respectively f%x)，g(x) Is a function of the estimated value of (2); t represents time, ε is a number greater than zero, x _d ⁽ⁿ⁾ For the n-th derivative of the desired speed, k= [ K ] _n ,k _n-1 ,k _n-2 ,…,k ₁ ] ^T Is a coefficient matrix, d%xT) is the interference quantity related to time and meets the requirement of d%x,t)||≤ε _d <Infinity, e is e _v Error matrix is +.>u _r For the system robust term, ρ (t) is w+d #xUpper bound of t)>

Further, the brain emotion learning circuit controller in the step 7) calculates an expected steering angle delta _k The specific steps of (a) are as follows:

step 7.1), inputting a stimulation signal into the thalamus for deep processing:

wherein the method comprises the steps ofFor inputting the stimulus signal, n represents the number of input stimulus signals,/->The central point vector value, sigma, for the jth hidden layer neuron _j >0 is the width of the gaussian function (j=1, …, m), and m e N is the total number of input signals;

step 7.2), processed signal S _j After entering the sensory cortex, the almond body and the prefrontal cortex enter respectively, and the almond kernel and the prefrontal cortex output are respectively:

wherein V is _j And W is _j (j=1, …, m) are the weight factors of the amygdala and the prefrontal cortex, respectively, weight vector for almond body, ++>Is the weight vector of the orbital cortex; the output of the controller is:

wherein the method comprises the steps ofu is the output delta of the controller _k ；

Step 7.3), the weight factor updating process is as follows:

wherein S is the signal S after thalamus processing _j F represents an unknown function f to be approximatedx) G represents an unknown function g to be approximatedx)，γ ₁ ～γ ₄ In order to adjust the parameters of the device,for error matrix, e is e _k U is the output delta of the controller _k P is a positive definite matrix and satisfies Λ ^T P + pΛ = -Q, Q being any n x n order positive definite matrix,b＝[0 0 ... 0 1] ^T _n ，V _fk 、V _gk respectively corresponding to the time f of the almond body kx)、g(x) Weight factor, W of _fk 、W _gk Respectively corresponding to the time f of the forehead cortex kx)、g(x) Weight factor of->Is the first derivative;

step 7.4), the control quantity u is the vehicle expected acceleration delta _k The calculation process of (2) is as follows:

g(x)＝(V _g -W _g )S

u＝g ^-1 (x)[-f(x)+x _d ⁽ⁿ⁾ +K ^T E+u _r ]

wherein S is the signal S after thalamus processing _j ，F is a state variable of the systemx) Is an unknown smooth function and f%x) The bounded is marked as f%x)||≤f<∞，g(x) Is unknown smooth function and g%x) The limit is marked as g%x)||≤g<∞；/>Respectively f%x)，g(x) Is a function of the estimated value of (2); t represents time, ε is a number greater than zero, x _d ⁽ⁿ⁾ For the n-th derivative of the desired trajectory, k= [ K ] _n ,k _n-1 ,k _n-2 ,…,k ₁ ] ^T Is a coefficient matrix, d%xT) is the interference quantity related to time and meets the requirement of d%x,t)||≤ε _d <Infinity, e is e _k Error matrix is +.>u _r For the system robust term, ρ (t) is w+d #xUpper bound of t)>

The beneficial effects are that:

1) The intelligent fuzzy algorithm is adopted to plan the speeds of different points of the curve in real time, the expected speeds of the vehicle at different road positions are planned in real time by combining the vehicle dynamics constraint, and corresponding smoothing processing is carried out to obtain a curve expected speed curve, so that the running stability of the vehicle at curves with different curvatures is ensured.

2) Most vehicles have a vehicle body stabilizing system, but speed planning is performed on curve driving in advance, so that the unstable state of the vehicles can be effectively reduced, and double insurance is applied to curve driving safety.

3) The brain emotion learning loop model in the brain-like calculation field is utilized to design a track tracking controller, so that the speed tracking and track tracking precision is increased, the tracking error is reduced, the lag response time of a driver is reduced, and the stability of the tracking process is improved.

Drawings

FIG. 1 is a logic block diagram of a driving curve trajectory tracking control system;

FIG. 2 is a graph of a radial basal brain emotional nerve learning model;

fig. 3 is a block diagram of the trajectory tracking control system input u.

Detailed Description

The invention is further explained below with reference to the drawings.

As shown in fig. 1, the logic block diagram of the driving-assisted curve track following control system of the present invention can be divided into seven functional parts: the system comprises an environment sensing unit, a vehicle sensor, a CAN bus, a VCU, a wire control and movement unit, a speed control unit and a wire control and movement unit. The following is a specific explanation of the flow of the block diagram:

the environment sensing unit comprises a camera, a laser radar and a millimeter wave radar; the self-vehicle sensor unit comprises a vehicle speed sensor, a wheel speed sensor, an acceleration sensor and a front wheel steering angle sensor; the camera is arranged right above the front windshield of the vehicle and used for identifying lane line information, barrier information and lane speed limit and transmitting image information into the VCU; the laser radars are at least 2 in number (the safety of front end detection is ensured, the danger caused by single radar failure is prevented), are respectively arranged on a front end cabin cover and a roof and are used for detecting the relative distance between a front obstacle and a vehicle and the speed and the acceleration of the front vehicle, and information is stored on a CAN bus for being called and processed by a VCU in real time; the millimeter wave radar is at least 1 in number, is arranged on an air inlet barrier at the front part of the vehicle, and is used for detecting the relative distance between a remote vehicle and a host vehicle, and storing information on a CAN bus for being called and processed by a VCU in real time; the vehicle speed sensor, the wheel speed sensor, the acceleration sensor and the front wheel steering angle sensor are respectively used for collecting the speed, the wheel rotating speed, the longitudinal acceleration and the front wheel steering angle of the vehicle, and storing information on a CAN bus for being called and processed by a VCU in real time; the steering-by-wire unit comprises a steering power-assisted motor and a steering controller and is used for receiving a steering signal of the VCU and steering; the brake-by-wire unit comprises a brake cylinder and is used for receiving a brake signal of the VCU and braking; the wheel speed control unit comprises a wheel motor and is used for receiving a wheel speed signal of the VCU and controlling the vehicle speed; the electronic control unit VCU realizes the functions of calculating, judging and sending control signals and is used for calculating the expected vehicle acceleration and the expected front wheel rotation angle respectively according to the received values of the speed, the longitudinal acceleration, the front wheel rotation angle, the lane line curvature, the track tracking deviation, the lane speed limit, the road attachment coefficient and the like of the vehicle; the VCU comprises a speed planning module and a track tracking module, wherein the speed planning module calculates a speed curve of a curve according to the curvature of a lane line, track tracking deviation and road adhesion coefficient, and the speed curve of the curve is processed by Kalman filtering and vehicle dynamics constraint so as to be smoother and more reasonable; the track tracking control module builds a tracking controller by using a brain emotion learning loop model, and calculates a spring expected front wheel corner by combining dynamics constraint; the desired front wheel rotation angle and the desired acceleration bring the vehicle into the next running state.

The invention also discloses a driving curve track assisting control method, which comprises the following steps:

step 1), a camera collects lane line information, obstacle information and lane speed limit information of a lane at the current moment k, and the curvature kappa of a lane line of a lane where a recognized own vehicle is located _k =1/R, road track tracking deviation e _k ＝d _{real k} -d _{desire k} Speed limit V of lane _{lim k} Transmitting to CAN bus for VCU to fetch for calculation and judgment; l and R are the length and the radius of the lane line acquired by the camera at the moment k; d, d _{real k} For the true deviation of the longitudinal axis of the vehicle from the lane line at time k, d _{desire k} The expected deviation of the longitudinal axis of the vehicle from the lane line at time k;

step 2), the speed u of the automobile at the moment k is collected by a speed sensor, an acceleration sensor and a front wheel steering angle sensor respectively _{v k} Wheel speed u _{w k} Acceleration alpha _k Front wheel angle delta _k The collected data is input to a CAN bus for VCU calculation;

step 3), calculating the longitudinal force F of the tire by adopting a magic formula tire model _x Calculating road surface braking coefficient mu according to a vehicle dynamics formula _b Estimating the road adhesion coefficient mu at the moment k according to a fuzzy estimation algorithm _k ；

Step 3.1) calculating the tire longitudinal force F at the current moment k according to the magic formula tire model _{x k} ：

F _x ＝D _x sin{C _x arctan[B _x λ-E _x (B _x λ-arctan B _x λ)]}

C _x ＝1.62

D _x ＝a ₁ F _z ² +a ₂ F _z

B _x C _x D _x ＝a ₃ sin[a ₄ arctan(a ₅ F _z )](1-a ₁₂ |λ|)

B _x C _x D _x ＝B _x C _x D _x /C _x D _x

E _x ＝a ₆ F _z ² +a ₇ F _z +a ₈

Wherein a is ₁ 、a ₂ 、a ₃ 、a ₄ 、a ₅ 、a ₆ 、a ₇ 、a ₈ 、a ₁₂ Fitting parameters of the magic formula tire, F _z For vertical load of the tyre, x represents longitudinal direction of the vehicle, lambda is slip ratio, D _x Is the peak factor, B _x As a rigidity factor, C _x Is a curve shape factor, E _x Is a curve curvature factor;

step 3.2) fitting six road surfaces through a large amount of experimental data: ice, snow, wet cobble, wet asphalt, dry cement, dry asphalt semi-empirical tire-road mathematical model:

wherein C is ₁ 、C ₂ 、C ₃ Three parameters for the pavement model; then drawing lambda-mu curves of six road surfaces according to the corresponding parameters; λ represents the slip ratio, μ is the road surface peak adhesion coefficient;

step 3.3) calculating the brake force coefficient μ _b And slip ratio λ:

λ＝(u _w -u _v )/u _w

step 3.4) setting a fuzzy controller, wherein the input is mu _b And lambda, output as the similarity s of the current road surface and six road surfaces ₁ 、s ₂ 、s ₃ 、s ₄ 、s ₅ 、s ₆ The method comprises the steps of carrying out a first treatment on the surface of the Then the weighted average method is adopted to calculate the road attachment coefficient mu _k ：

Wherein mu ₁ 、μ ₂ 、μ ₃ 、μ ₄ 、μ ₅ 、μ ₆ The peak adhesion coefficients of the six road surfaces are respectively.

Step 4), VCU according to the received lane line curvature kappa _k Tracking deviation e of road track _k Road adhesion coefficient mu _k Calculating planning speed u at curve k point position at current k moment by using fuzzy control algorithm _{plan k} The method comprises the steps of carrying out a first treatment on the surface of the Combining the planned speeds at each time before the k time, the speed planning curve described in step 5) can be obtained.

The input of the fuzzy control algorithm is lane line curvature kappa _k Tracking deviation e of road track _k And road adhesion coefficient mu _k According to IF A and B and C then D fuzzy rule, road planning speed u at this time is output _{plan k} The fuzzy rule is:

step 4.1) when the lane line curvature κ _k The speed of the vehicle should be reduced when the vehicle is too large, and the speed of the vehicle should be lower than the speed limit V of the lane _lim Allowing the vehicle to safely pass through a curve with a smaller radius;

step 4.2) when road adhesion coefficient μ _k When the speed is larger, the vehicle speed should be properly increased and the vehicle speed should be lower than the speed limit V of the lane _lim ；

Step 4.3) tracking the deviation e when the track is traced _k When larger, the vehicle speed should be reduced appropriately so that the vehicle re-tracks the lane track over a shorter distance.

Step 5), the camera detects road traffic signs in real time, and when the road speed limit is detected to be V _{lim k} When VCU uses Kalman filtering to smooth speed planning curve and is smaller than road speed limit V _{lim k} If no road speed limit exists at the moment k, no speed limit processing is performed, and a smooth and reasonable speed curve is obtained;

step 6), VCU calculates the planning speed u at k moment _{pla n} And the real speed u _{rea l} Deviation e of (2) _vk ＝u _plank -u _rea Calculating expected acceleration alpha at k moment through brain emotion learning loop controller _{des k} Judging that the vehicle needs accelerating, decelerating or idling according to a braking driving switching strategy, and finally calculating the automatic pedal pressure and the throttle opening according to an inverse engine model and an inverse brake model respectively so as to control the acceleration and deceleration of the vehicle;

the brain emotion learning loop controller calculates the expected acceleration alpha _des The method comprises the following steps:

step 6.1), inputting a stimulation signal into the thalamus for deep processing:

wherein the method comprises the steps ofFor inputting the stimulus signal, n represents the number of input stimulus signals,/->The central point vector value, sigma, for the jth hidden layer neuron _j >0 is the width of the gaussian function (j=1, …, m), and m e N is the total number of input signals;

step 6.2), processed signal S _j After entering the sensory cortex, the almond body and the prefrontal cortex enter respectively, and the almond kernel and the prefrontal cortex output are respectively:

wherein V is _j And W is _j (j=1, …, m) are the weight factors of the amygdala and the prefrontal cortex, respectively, weight vector for almond body, ++>Is the weight vector of the orbital cortex; the output of the controller is:

wherein the method comprises the steps ofu is the output alpha of the controller _des ；

Step 6.3), the weight factor updating process is as follows:

wherein S is the signal S after thalamus processing _j F represents an unknown function f to be approximatedx) G represents an unknown function g to be approximatedx)，γ ₁ ～γ ₄ In order to adjust the parameters of the device,for error matrix, e is e _v U is the control output alpha _des P is a positive definite matrix and satisfies Λ ^T P + pΛ = -Q, Q being any n x n order positive definite matrix,b＝[0 0 ... 0 1] ^T _n ，V _fk 、V _gk respectively corresponding to the time f of the almond body kx)、g(x) Weight factor, W of _fk 、W _gk Respectively corresponding to the time f of the forehead cortex kx)、g(x) Weight factor of->Is the first derivative;

step 6.4), the control quantity u is the vehicle expected acceleration alpha _des The calculation process of (2) is as follows:

g(x)＝(V _g -W _g )S

u＝g ^-1 (x)[-f(x)+x _d ⁽ⁿ⁾ +K ^T E+u _r ]

wherein S is the signal S after thalamus processing _j ，F is a state variable of the systemx) Is an unknown smooth function and f%x) The bounded is marked as f%x)||≤f<∞，g(x) Is unknown smooth function and g%x) The limit is marked as g%x)||≤g<∞；/>Respectively are provided withIs f%x)，g(x) Is a function of the estimated value of (2); t represents time, ε is a number greater than zero, x _d ⁽ⁿ⁾ For the n-th derivative of the desired speed, k= [ K ] _n ,k _n-1 ,k _n-2 ,…,k ₁ ] ^T Is a coefficient matrix, d%xT) is the interference quantity related to time and meets the requirement of d%x,t)||≤ε _d <Infinity, e is e _v Error matrix is +.>u _r For the system robust term, ρ (t) is w+d #xUpper bound of t)>

Step 6.5), according to the idle acceleration value α, two switching thresholds h1 and h2 are set near the idle acceleration, forming a transition region:

table 1 actuator switch strategy

The two switching thresholds h1 and h2 determine the control timing of braking and driving. Therefore, the value of h1 should be reasonable, especially, when the value of h1 is too small, the vehicle frequently and slightly shakes, and the riding comfort is affected. h1 is too large, and braking is delayed. The value of h2 can influence the control of driving, too small h2 can cause frequent acceleration, and too large h2 can cause acceleration delay of the system. Finally, through experimental balance, the thresholds h1 and h2 are set to 0.05 and 0.3 respectively.

Step 6.6), when the control strategy output corresponds to braking, at which time the vehicle needs to perform braking operation, according to the inverse brake model, taking into account the air resistance and the rolling resistance, the desired braking pressure is finally calculated by the inverse brake model:

t in _wb,des Indicating the expected braking pressure, m indicating the mass of the whole main vehicle, g indicating the gravitational acceleration, f indicating the rolling resistance coefficient, r _eff Indicating the radius of the wheel, C _D Represents the air resistance coefficient, A represents the windward area v represents the vehicle speed, kb is the braking gain coefficient and 20, P are taken here _W Indicating a desired brake pressure;

when the control strategy output corresponds to acceleration, the vehicle needs to perform acceleration operation at the moment, and the throttle opening corresponding to the acceleration is obtained according to the three-dimensional lookup table of the inverse engine model, so that the acceleration operation is realized;

step 7), the camera detects the position of the lane line of the road in real time, and the real deviation d of the longitudinal axis of the vehicle at the moment k and the lane line is obtained _{real k} Transmitting the deviation to a CAN bus, and calculating the expected deviation d of the longitudinal axis of the vehicle and the lane line at the moment k by the VCU in real time _{plan k} From true deviation d _{real k} E of the difference e of (2) _k ＝d _{real k} -d _{plan k} And the deviation value e _k The front wheel angle delta required by the vehicle to track the lane line at the current moment is calculated by inputting the front wheel angle delta into a brain emotion learning loop controller _k Finally, delta is transmitted through a CAN bus _k The input vehicle steer-by-wire unit controls the vehicle to track the lane line in real time;

the brain emotion learning loop controller calculates the expected steering angle delta _k The specific steps of (a) are as follows:

step 7.1), inputting a stimulation signal into the thalamus for deep processing:

wherein the method comprises the steps ofFor inputting the stimulus signal, n represents the number of input stimulus signals,/->The central point vector value, sigma, for the jth hidden layer neuron _j >0 is the width of the gaussian function (j=1, …, m), and m e N is the total number of input signals;

step 7.2), processed signal S _j After entering the sensory cortex, the almond body and the prefrontal cortex enter respectively, and the almond kernel and the prefrontal cortex output are respectively:

wherein V is _j And W is _j (j=1, …, m) are the weight factors of the amygdala and the prefrontal cortex, respectively, weight vector for almond body, ++>Is the weight vector of the orbital cortex; the output of the controller is:

wherein the method comprises the steps ofu is the output delta of the controller _k ；

Step 7.3), the weight factor updating process is as follows:

wherein S is the signal S after thalamus processing _j F represents an unknown function f to be approximatedx) G represents an unknown function g to be approximatedx)，γ ₁ ～γ ₄ In order to adjust the parameters of the device,for error matrix, e is e _k U is the control output delta _k P is a positive definite matrix and satisfies Λ ^T P + pΛ = -Q, Q being any n x n order positive definite matrix,b＝[0 0 ... 0 1] ^T _n ，V _fk 、V _gk respectively corresponding to the time f of the almond body kx)、g(x) Weight factor, W of _fk 、W _gk Respectively corresponding to the time f of the forehead cortex kx)、g(x) Weight factor of->Is the first derivative;

step 7.4), the control quantity u is the vehicle expected acceleration delta _k The calculation process of (2) is as follows:

g(x)＝(V _g -W _g )S

u＝g ^-1 (x)[-f(x)+x _d ⁽ⁿ⁾ +K ^T E+u _r ]

wherein S is the signal S after thalamus processing _j ，F is a state variable of the systemx) Is an unknown smooth function and f%x) The bounded is marked as f%x)||≤f<∞，g(x) Is unknown smooth function and g%x) The limit is marked as g%x)||≤g<∞；/>Respectively f%x)，g(x) Is a function of the estimated value of (2); t represents time, ε is a number greater than zero, x _d ⁽ⁿ⁾ For the n-th derivative of the desired trajectory, k= [ K ] _n ,k _n-1 ,k _n-2 ,…,k ₁ ] ^T Is a coefficient matrix, d%xT) is the interference quantity related to time and meets the requirement of d%x,t)||≤ε _d <Infinity, e is e _k Error matrix is +.>u _r For the system robust term, ρ (t) is w+d #xUpper bound of t)>

And 8), repeating the steps 1) -7) in sequence, and calculating the acceleration and the front wheel rotation angle at the moment k+1.

Claims (3)

1. The auxiliary driving curve track tracking control method is characterized by comprising the following steps of:

step 1), a camera collects k lane line information, obstacle information and lane speed limit information at the current moment, and a host vehicle station to be identified

Curvature k of lane line in lane _k =1/R, road track tracking deviation e _k ＝d _{real k} -d _{desire k} Speed limit V of lane _{lim k} Transmitting to CAN bus for VCU to fetch for calculation and judgment; l and R are the length and the radius of the lane line acquired by the camera at the moment k; d, d _{real k} For the true deviation of the longitudinal axis of the vehicle from the lane line at time k, d _{desire k} The expected deviation of the longitudinal axis of the vehicle from the lane line at time k;

step 2), the speed u of the automobile at the moment k is collected by a speed sensor, an acceleration sensor and a front wheel steering angle sensor respectively _{v k} Wheel speed u _{w k} Acceleration alpha _k Front wheel angle delta _k The collected data is input to a CAN bus for VCU calculation;

step 3), calculating the longitudinal force F of the tire by adopting a magic formula tire model _x Calculating road surface braking coefficient mu according to a vehicle dynamics formula _b Estimating the road adhesion coefficient mu at the moment k according to a fuzzy estimation algorithm _k ；

Step 4), VCU according to the received lane line curvature kappa _k Tracking deviation e of road track _k Road adhesion coefficient mu _k Calculating the k-moment planning speed u of the curve at the current k moment by using a fuzzy control algorithm _{plan k} The method comprises the steps of carrying out a first treatment on the surface of the Combining the planning speed at each moment before the moment k to obtain a speed planning curve;

step 5), the camera detects road traffic signs in real time, and when the road speed limit is detected to be V _{lim k} When VCU uses Kalman filtering to smooth speed planning curve and is smaller than road speed limit V _{lim k} If no road speed limit exists at the moment k, no speed limit processing is performed, and a smooth and reasonable speed curve is obtained;

step 6), VCU calculates the planning speed u at k moment _{plan k} And the real speed u _{real k} Deviation e of (2) _vk ＝u _{plan k} -u _{real k} Calculating expected acceleration alpha at k moment through brain emotion learning loop controller _{des k} Judging that the vehicle needs accelerating, decelerating or idling according to a braking driving switching strategy, and finally calculating automatic pedal pressure according to the throttle opening of the reverse engine model and the reverse brake model respectively so as to control the acceleration and deceleration of the vehicle;

step 7), the camera detects the position of the lane line of the road in real time, and the real deviation d of the longitudinal axis of the vehicle at the moment k and the lane line is obtained _{real k} Transmitting the deviation to a CAN bus, and calculating the expected deviation d of the longitudinal axis of the vehicle and the lane line at the moment k by the VCU in real time _{plan k} From true deviation d _{real k} E of the difference e of (2) _k ＝d _{real k} -d _{plan k} And the deviation value e _k The front wheel angle delta required by the vehicle to track the lane line at the current moment is calculated by inputting the front wheel angle delta into a brain emotion learning loop controller _k Finally, delta is transmitted through a CAN bus _k The input vehicle steer-by-wire unit controls the vehicle to track the lane line in real time;

step 8), repeating the steps 1) -7) in sequence, and calculating the acceleration and the front wheel rotation angle at the moment k+1;

in the step 6), the brain emotion learning loop controller calculates the expected acceleration alpha _{des k} The method comprises the following steps:

step 6.1), inputting a stimulation signal into the thalamus for deep processing:

wherein the method comprises the steps ofFor inputting the stimulus signal, n represents the number of input stimulus signals,/-> The central point vector value, sigma, for the jth hidden layer neuron _j >0 is the width of the gaussian function (j=1, …, m), and m e N is the total number of input signals;

step 6.2), processed signal S _j After entering the sensory cortex, the almond body and the prefrontal cortex enter respectively, and the almond kernel and the prefrontal cortex output are respectively:

wherein V is _j And W is _j (j=1, …, m) are the weight factors of the amygdala and the prefrontal cortex, respectively, weight vector for almond body, ++>Is the weight vector of the orbital cortex; the output of the controller is:

wherein the method comprises the steps ofu is the output alpha of the controller _des ；

Step 6.3), the weight factor updating process is as follows:

wherein S is the signal S after thalamus processing _j F represents an unknown function f (x) to be approximated, g represents an unknown function g # -, to be approximatedx)，γ ₁ ～γ ₄ In order to adjust the parameters of the device,for error matrix, e is e _v U is the output of the controller, α _des For the desired acceleration of the vehicle, P is a positive definite matrix and satisfies Λ ^T P + pΛ = -Q, Q being any n x n order positive definite matrix,b＝[0 0…01] ^T _n ，V _fk 、V _gk respectively corresponding to the time f of the almond body kx)、g(x) Weight factor, W of _fk 、W _gk Respectively corresponding to the time f of the forehead cortex kx)、g(x) Weight factor of->Is the first derivative;

step 6.4), the control quantity u is the vehicle expected acceleration alpha _des The calculation process of (2) is as follows:

wherein the method comprises the steps ofF is a state variable of the systemx) Is an unknown smooth function and f%x) The bounded is marked as f%x)||≤f<∞，g(x) Is unknown smooth function and g%x) The limit is marked as g%x)||≤g<∞；Respectively f%x)，g(x) Is a function of the estimated value of (2); t represents time, ε is a number greater than zero, x _d ⁽ⁿ⁾ For the n-th derivative of the desired speed, k= [ K ] _n ，k _n-1 ，k _n-2 ，...，k ₁ ] ^T Is a coefficient matrix, d%xT) is the interference quantity related to time and meets the requirement of d%x，t)||≤ε _d <Infinity, e is e _v Error matrix is +.> u _r For the system robust term, ρ (t) is w+d #xUpper bound of t)>

When the control strategy output corresponds to braking, at this time, the vehicle needs to perform braking operation, and according to the reverse brake model, the desired braking pressure is finally calculated by the reverse brake model in consideration of air resistance and rolling resistance:

t in _wb，des Indicating the expected braking pressure, m indicating the mass of the whole main vehicle, g indicating the gravitational acceleration, f indicating the rolling resistance coefficient, r _eff Indicating the radius of the wheel, C _D Represents the air resistance coefficient, A represents the windward area v represents the vehicle speed, kb is the braking gain coefficient and 20, P are taken here _W Indicating a desired brake pressure;

the brain emotion learning loop controller in the step 7) calculates an expected steering angle delta _k The specific steps of (a) are as follows:

step 7.1), inputting a stimulation signal into the thalamus for deep processing:

wherein the method comprises the steps ofFor inputting the stimulus signal, n represents the number of input stimulus signals,/-> The central point vector value, sigma, for the jth hidden layer neuron _j > 0 is the width of the gaussian function (j=1,., m), m e N is the total of the input signalsA number;

step 7.2), processed signal S _j After entering the sensory cortex, the almond body and the prefrontal cortex enter respectively, and the almond kernel and the prefrontal cortex output are respectively:

wherein V is _j And W is _j (j=1,., m) are weight factors of the amygdala and the prefrontal cortex, respectively, weight vector for almond body, ++>Is the weight vector of the orbital cortex; the output of the controller is:

wherein the method comprises the steps ofu is the output delta of the controller _k ；

Step 7.3), the weight factor updating process is as follows:

wherein f represents an unknown function f to be approximatedx) G represents an unknown function g to be approximatedx)，γ ₁ ～γ ₄ In order to adjust the parameters of the device,for error matrix, e is e _k U is the output delta of the controller _k P is a positive definite matrix and satisfies Λ ^T P+pΛ= -Q, Q being any n x n order positive definite matrix, ++>b＝[0 0…01] ^T _n ，V _fk 、V _gk Respectively corresponding to the time f of the almond body kx)、g(x) Weight factor, W of _fk 、W _gk Respectively corresponding to the time f of the forehead cortex kx)、g(x) Weight factor of->Is the first derivative;

step 7.4), the control quantity u is the vehicle expected acceleration delta _k The calculation process of (2) is as follows:

wherein the method comprises the steps ofF is a state variable of the systemx) Is an unknown smooth function and f%x) The bounded is marked as f%x)||≤f<∞，g(x) Is unknown smooth function and g%x) The limit is marked as g%x)||≤g<∞；/>Respectively f%x)，g(x) Is a function of the estimated value of (2); t represents time, ε is a number greater than zero, x _d ⁽ⁿ⁾ For the n-th derivative of the desired trajectory, k= [ K ] _n ，k _n-1 ，k _n-2 ，...，k ₁ ] ^T Is a coefficient matrix, d%xT) is the interference quantity related to time and meets the requirement of d%x，t)||≤ε _d <Infinity, e is e _k Error matrix is +.> u _r For the system robust term, ρ (t) is w+d #xAn upper bound of t) is defined,

2. the driving curve trajectory tracking control method according to claim 1, wherein the step 3) specifically includes:

step 3.1) calculating the tire longitudinal force F at the current moment k according to the magic formula tire model _x ：

F _x ＝D _x sin{C _x arctan[B _x λ-E _x (B _x λ-arctanB _x λ)]}

C _x ＝1.62

D _x ＝a ₁ F _z ² +a ₂ F _z

B _x C _x D _x ＝a ₃ sin[a ₄ arctan(a ₅ F _z )](1-a ₁₂ |λ|)

B _x C _x D _x ＝B _x C _x D _x /C _x D _x

E _x ＝a ₆ F _z ² +a ₇ F _z +a ₈

Wherein a is ₁ 、a ₂ 、a ₃ 、a ₄ 、a ₅ 、a ₆ 、a ₇ 、a ₈ 、a ₁₂ Fitting parameters of the magic formula tire, F _z For vertical load of the tyre, x represents longitudinal direction of the vehicle, lambda is slip ratio, D _x Is the peak factor, B _x As a rigidity factor, C _x Is a curve shape factor, E _x Is a curve curvature factor;

step 3.2) fitting six road surfaces through a large amount of experimental data: ice, snow, wet cobble, wet asphalt, dry cement, dry asphalt semi-empirical tire-road mathematical model:

wherein C is ₁ 、C ₂ 、C ₃ Three parameters of the pavement mathematical model; then drawing lambda-mu curves of six road surfaces according to the corresponding parameters; λ represents the slip ratio, μ is the road surface peak adhesion coefficient;

step 3.3) calculating the brake force coefficient μ _b And slip ratio λ:

λ＝(u _w -u _v )/u _w

wherein u is _v Indicating the speed of the automobile, u _w Representing the wheel speed of an automobile;

step 3.4) setting a fuzzy controller, wherein the input is mu _b And lambda, output as the similarity s of the current road surface and six road surfaces ₁ 、s ₂ 、s ₃ 、s ₄ 、s ₅ 、s ₆ The method comprises the steps of carrying out a first treatment on the surface of the Then the weighted average method is adopted to calculate the road attachment coefficient mu _k ：

Wherein mu ₁ 、μ ₂ 、μ ₃ 、μ ₄ 、μ ₅ 、μ ₆ The peak adhesion coefficients of the six road surfaces are respectively.

3. The driving-assisted curve track following control method according to claim 1, wherein in the step 4), the input of the fuzzy control algorithm is a lane line curvature κ _k Tracking deviation e of road track _k And road adhesion coefficient mu _k According to IF A and B and C then D fuzzy rule, road planning speed u at this time is output _{plan k} The fuzzy rule is:

step 4.1) when the lane line curvature κ _k The speed of the vehicle should be reduced when the vehicle is too large, and the speed of the vehicle should be lower than the speed limit V of the lane _lim Allowing the vehicle to safely pass through a curve with a smaller radius;

step 4.2) when road adhesion coefficient μ _k When the speed is larger, the vehicle speed should be properly increased and the vehicle speed should be lower than the speed limit V of the lane _lim ；

Step 4.3) tracking the deviation e when the track is traced _k When larger, the vehicle speed should be properly reduced so as to facilitate the vehicleThe lane trajectory is re-tracked within a short distance.