CN118269929B

CN118269929B - Longitudinal and transverse control method and device for automatic driving automobile

Info

Publication number: CN118269929B
Application number: CN202410704977.5A
Authority: CN
Inventors: 张宁; 陈新度; 冯桑; 黄晓涛
Original assignee: Guangdong University of Technology
Current assignee: Guangdong University of Technology
Priority date: 2024-06-03
Filing date: 2024-06-03
Publication date: 2024-07-23
Anticipated expiration: 2044-06-03
Also published as: CN118269929A

Abstract

The invention discloses a longitudinal and transverse control method and a device for an automatic driving automobile, comprising the steps of acquiring transverse data and longitudinal speed of the automobile; inputting the automobile transverse data to a preset interval two-type model transverse controller for transverse control, and outputting a target steering angle; longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval two-type model longitudinal controller to generate a target torque; determining a target longitudinal force according to the target torque and the longitudinal speed of the automobile; controlling the automatic driving automobile to run through the target steering angle and the target longitudinal force; the technical problem that the stability of the overall performance of the vehicle is poor in the existing longitudinal and transverse control method of the automobile is solved.

Description

Longitudinal and transverse control method and device for automatic driving automobile

Technical Field

The invention relates to the technical field of automatic driving automobile motion control, in particular to a longitudinal and transverse control method and a longitudinal and transverse control device for an automatic driving automobile.

Background

With the continuous upgrading of automobile intellectualization and networking and the rapid development of artificial intelligence technology, intelligent automobiles have become a trend of the transformation of the traditional automobile industry and a research hotspot of worldwide vehicle engineering in the background. The intelligent automobile is hopeful to relieve people from tedious long distance driving, and has great potential for reducing traffic jams and traffic accidents.

The automatic driving control of the automobile can be divided into longitudinal control and transverse control, wherein the longitudinal control refers to controlling the longitudinal speed of the automobile, namely controlling the accelerator and the brake; lateral control refers to controlling the steering angle of the front wheels of the vehicle so that the actual running track of the vehicle approaches the reference running track, i.e., track following. The control algorithm based on the vehicle dynamics model can realize better road utilization rate and higher safety, for example, operations such as track tracking, obstacle avoidance and the like can be performed under different road adhesion coefficients.

Most of the existing automobile longitudinal and transverse control methods are based on analysis mechanics technology to establish a vehicle kinematics or dynamics model, but the process has the problems that data distribution is uneven, model parameters cannot correspond to real world physical parameters, so that the model has insufficient interpretability, and the stability of the overall performance of the vehicle is poor.

Disclosure of Invention

The invention provides a longitudinal and transverse control method and device for an automatic driving automobile, which are used for solving the technical problem that the stability of the overall performance of the automobile is poor in the conventional longitudinal and transverse control method for the automobile.

The invention provides a longitudinal and transverse control method of an automatic driving automobile, which comprises the following steps:

acquiring transverse data and longitudinal speed of the automobile;

inputting the automobile transverse data to a preset interval two-type model transverse controller for transverse control, and outputting a target steering angle;

longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval two-type model longitudinal controller to generate a target torque;

Determining a target longitudinal force based on the target torque and the vehicle longitudinal speed;

And controlling the automatic driving automobile to run through the target steering angle and the target longitudinal force.

Optionally, the preset interval two-type model transverse controller comprises a transverse input layer, a transverse normalization layer, a transverse blurring layer, a transverse corresponding layer, a transverse back piece layer and a transverse output layer; the step of inputting the automobile transverse data to a preset interval two-type model transverse controller for transverse control and outputting a target steering angle comprises the following steps:

Transmitting the automobile transverse data to a transverse normalization layer through a transverse input layer to perform linear combination and normalization to generate a plurality of transverse normalization linear combination values;

fuzzifying each transverse normalized linear combination value by adopting a transverse fuzzifying layer to generate a plurality of transverse boundary membership function values and a plurality of transverse lower boundary membership function values;

inputting each transverse boundary membership function value and each transverse lower boundary membership function value into a transverse corresponding layer to carry out fuzzy product operator, and outputting a plurality of transverse closed interval sets;

determining a transverse fuzzy back part definition value corresponding to each transverse normalized linear combination value according to each transverse normalized linear combination value by adopting a preset interval two-type fuzzy transverse rule;

Determining a plurality of automobile transverse neuron node output values by adopting a transverse back-part layer according to the clear values of the transverse fuzzy back-parts;

And performing drop aggregation on the output value of each automobile transverse neuron node and each transverse closed interval set through a transverse output layer to generate a target steering angle.

Optionally, the preset interval two-type model longitudinal controller comprises a longitudinal input layer, a longitudinal normalization layer, a longitudinal blurring layer, a longitudinal corresponding layer, a longitudinal back piece layer and a longitudinal output layer; the step of adopting a preset interval two-type model longitudinal controller to longitudinally control the longitudinal speed of the automobile and generating target torque comprises the following steps:

Transmitting the longitudinal speed of the automobile to a longitudinal normalization layer through a longitudinal input layer to perform linear combination and normalization, and generating a plurality of longitudinal normalization linear combination values;

fuzzifying each longitudinal normalized linear combination value by adopting a longitudinal fuzzifying layer to generate a plurality of longitudinal upper bound membership function values and a plurality of longitudinal lower bound membership function values;

inputting each longitudinal upper bound membership function value and each longitudinal lower bound membership function value into a longitudinal corresponding layer to carry out fuzzy product operator, and outputting a plurality of longitudinal closed interval sets;

Determining a longitudinal fuzzy back piece definition value corresponding to each longitudinal normalized linear combination value according to each longitudinal normalized linear combination value by adopting a preset interval two-type model longitudinal rule;

determining a plurality of longitudinal neuron node output values of the automobile by adopting a longitudinal back part layer according to the clear value of each longitudinal fuzzy back part;

And performing drop aggregation on the output value of each longitudinal neuron node of the automobile and each longitudinal closed interval set through a longitudinal output layer to generate target torque.

Optionally, the step of determining a target longitudinal force according to the target torque and the longitudinal speed of the vehicle comprises:

inputting the target torque to a preset proportional-integral-derivative controller for adjustment, and outputting a longitudinal force demand;

determining the longitudinal force of the tire according to the longitudinal speed of the automobile by adopting a preset longitudinal force function of the tire;

A target longitudinal force is calculated using the longitudinal force demand and the tire longitudinal force.

Optionally, before the step of acquiring the vehicle lateral data and the vehicle longitudinal speed, the method includes:

acquiring transverse data and longitudinal speed of the automobile to be trained, inputting the transverse data of the automobile to be trained into an initial interval two-type model transverse controller for transverse control, and determining a steering angle to be trained;

Based on motion dynamics, constructing a vehicle model corresponding to the automatic driving automobile, and determining and adjusting automobile transverse data by adopting the vehicle model according to the steering angle to be trained;

Optimizing the initial interval two-type model transverse controller by adopting the adjusting automobile transverse data, and determining the preset interval two-type model transverse controller;

And optimizing the initial interval two-type model longitudinal controller by adopting the longitudinal speed of the automobile to be trained, and determining the preset interval two-type model longitudinal controller.

Optionally, the step of optimizing the initial interval two-type model transversal controller by using the adjusted car transversal data to determine the preset interval two-type model transversal controller includes:

Calculating a transverse loss value according to the adjusted automobile transverse data, the preset reference transverse displacement and the preset reference transverse swing angle;

updating the super parameters of the initial interval two-type model transverse controller by adopting the transverse loss value, determining the intermediate interval two-type model transverse controller, and counting the iteration times in real time;

judging whether the iteration times reach preset training times or not;

And if the iteration times reach the preset training times, using the middle interval two-type model transverse controller as the preset interval two-type model transverse controller.

The second aspect of the present invention provides an automatic driving automobile longitudinal and lateral control device, comprising:

The acquisition module is used for acquiring transverse data and longitudinal speed of the automobile;

The transverse control module is used for inputting the transverse data of the automobile to a preset interval two-type model transverse controller to carry out transverse control and outputting a target steering angle;

the longitudinal control module is used for longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval second-type model longitudinal controller to generate target torque;

Determining a target longitudinal force from the target torque and the vehicle longitudinal speed according to a module;

And the driving module is used for controlling the automatic driving automobile to drive through the target steering angle and the target longitudinal force.

An electronic device according to a third aspect of the present invention includes a memory and a processor, where the memory stores a computer program, and the computer program when executed by the processor causes the processor to execute the steps of the method for controlling a longitudinal and transverse direction of an autopilot according to any one of the above.

A fourth aspect of the present invention provides a computer readable storage medium having stored thereon a computer program/instruction which, when executed by a processor, implements the steps of the method for controlling a longitudinal and lateral direction of an autonomous car as described in any of the above.

A fifth aspect of the invention provides a computer program product comprising computer programs/instructions which when executed by a processor implement the steps of the method for controlling the longitudinal and lateral directions of an autonomous car according to any of the preceding claims.

From the above technical scheme, the invention has the following advantages:

The technical scheme of the invention provides a longitudinal and transverse control method for an automatic driving automobile, which is used for acquiring transverse data and longitudinal speed of the automobile; inputting the automobile transverse data to a preset interval two-type model transverse controller for transverse control, and outputting a target steering angle; longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval two-type model longitudinal controller to generate a target torque; determining a target longitudinal force according to the target torque and the longitudinal speed of the automobile; controlling the automatic driving automobile to run through the target steering angle and the target longitudinal force; based on the scheme, the transverse data of the automobile is input to the preset interval secondary type model transverse controller to carry out transverse control, and the preset interval secondary type model longitudinal controller is adopted to carry out longitudinal control on the longitudinal speed of the automobile, so that the target steering angle and the target longitudinal force are obtained, the automatic driving automobile is controlled to run, the longitudinal movement and the transverse movement of the automatic driving automobile can be controlled simultaneously in the process, the important characteristic of speed tracking is considered, and the stability of the overall performance of the automobile is further improved.

Drawings

In order to more clearly illustrate the embodiments of the invention or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the invention, and that other drawings can be obtained from these drawings without inventive faculty for a person skilled in the art.

FIG. 1 is a flow chart of steps of a method for controlling longitudinal and transverse directions of an automatic driving automobile according to an embodiment of the present invention;

FIG. 2 is a flow chart of steps of another aspect control method for an automatic driving vehicle according to an embodiment of the present invention;

FIG. 3 is a schematic diagram of a process for constructing and training an initial interval two-type model transverse controller and an initial interval two-type model longitudinal controller according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of an autopilot vehicle dynamics model provided by an embodiment of the present invention;

FIG. 5 is a block diagram of an automated guided vehicle for simultaneous longitudinal and lateral motion control in accordance with an embodiment of the present invention;

Fig. 6 is a block diagram of an automatic driving car longitudinal and transverse control device according to an embodiment of the present invention.

Detailed Description

The embodiment of the invention provides a longitudinal and transverse control method and device for an automatic driving automobile, which are used for solving the technical problem that the stability of the overall performance of the automobile is poor in the conventional longitudinal and transverse control method for the automobile.

In order to make the objects, features and advantages of the present invention more comprehensible, the technical solutions in the embodiments of the present invention are described in detail below with reference to the accompanying drawings, and it is apparent that the embodiments described below are only some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.

Referring to fig. 1, fig. 1 is a flowchart illustrating steps of a method for controlling a longitudinal direction and a transverse direction of an automatic driving automobile according to an embodiment of the present invention.

and 101, acquiring transverse data and longitudinal speed of the automobile.

The automobile transverse data comprise transverse displacement, a transverse swing angle, a transverse displacement error and a transverse swing angle error, wherein the transverse displacement is obtained by controlling a steering wheel of an automatic driving automobile so as to control the vertical distance between a front wheel turning angle and a central line; the yaw angle is obtained through an angle signal of a steering angle sensor of a steering wheel of the automatic driving automobile; the transverse displacement error is obtained by performing difference between preset reference transverse displacement and the acquired transverse displacement; the yaw angle error is obtained by subtracting the obtained yaw angle from a preset reference yaw angle value.

In the present embodiment, the vehicle lateral data and the vehicle longitudinal speed are acquired.

And 102, inputting the automobile transverse data into a preset interval two-type model transverse controller to carry out transverse control, and outputting a target steering angle.

The preset interval secondary fuzzy horizontal controller and the preset interval secondary fuzzy vertical controller are both composed of an IT2FSNN ((INTERVAL TYPE-2 Fuzzy Set Neutral Network) interval secondary fuzzy neural network, wherein the preset interval secondary fuzzy horizontal controller comprises a horizontal input layer, a horizontal normalization layer, a horizontal fuzzification layer, a horizontal corresponding layer, a horizontal back piece layer and a horizontal output layer, and the preset interval secondary fuzzy vertical controller comprises a vertical input layer, a vertical normalization layer, a vertical fuzzification layer, a vertical corresponding layer, a vertical back piece layer and a vertical output layer.

Further, the longitudinal speed of the automobile is longitudinally controlled by adopting a preset interval two-type model longitudinal controller, which comprises the following steps: the method comprises the steps of transmitting automobile transverse data to a transverse normalization layer through a transverse input layer to perform linear combination and normalization to generate a plurality of transverse normalization linear combination values, wherein the input of a section two-type fuzzy neural network is the same as that of a traditional neural network, nodes of the section two-type fuzzy neural network are connected with input signals (automobile transverse data), the input signals come from information transmitted by a controller and are distributed to the next layer, the input signals are applied to a hidden layer through linked weights, and a normalization program is executed before the input linear combination is applied to neurons of the hidden layer; the processing procedure of the transverse normalized linear combination value comprises the following specific steps:

Wherein, For input to hidden layerA lateral linear combination of individual neurons; The number of nodes in the transverse input layer; Is the first Hidden layer of input connectionLateral weights of individual neurons, i.e. firstThe preset interval is two-level type of transversely regular transverse weights; To be input to the first The method comprises the steps of inputting automobile transverse data of layer nodes transversely, wherein the automobile transverse data comprise transverse displacement, a yaw angle, a transverse displacement error and a yaw angle error; For input to hidden layer Linear deviations of individual neurons; Linear combinations of inputs for the 1 st neuron in the hidden layer; To hide the first layer Linear combinations of individual neuron inputs; For input to hidden layer Laterally normalized linear combination values of the individual neurons; As a function of the maximum value.

Further, a transverse blurring layer is adopted to blur each transverse normalized linear combination value, and a plurality of transverse boundary membership function values and a plurality of transverse lower boundary membership function values are generated; the horizontal fuzzification layer fuzzifies the obtained horizontal normalized linear combination value subjected to linear combination and normalization treatment, and outputs a horizontal boundary membership function value and a horizontal lower boundary membership function value as each input value; the processing process of the transverse boundary membership function value and the transverse lower boundary membership function value specifically comprises the following steps:

Wherein, Membership function value for transverse boundary; membership function value for transverse lower bound; For input to hidden layer Laterally normalized linear combination values of the individual neurons; the center of a membership function of a transverse rule of a second type model in the next preset interval; is the center of the membership function; the center of a membership function of a transverse rule of a second type model in the previous preset interval; To be in hidden layer The vertical distance of the cross-domain membership function value of the i-th preset interval two-type model cross rule used in each neuron.

Further, each transverse boundary membership function value and each transverse lower boundary membership function value are input to a transverse corresponding layer to carry out fuzzy product operator, and a plurality of transverse closed interval sets are output; wherein the determined level of each rule of the IT2FSNN network is calculated in the neurons of the layer, each neuron node is formed byAndCarrying out fuzzy cross product operator, and outputting value as closed interval set; the processing process of the transverse closed interval set specifically comprises the following steps:

Wherein, For input to hidden layerA set of lateral amenities of individual neurons; is the lower boundary of the activation intensity; Is an upper boundary of activation intensity; membership function value for transverse lower bound; membership function value for transverse boundary; the number of the horizontal boundary membership function values and the horizontal lower boundary membership function values which are input to the horizontal corresponding layers.

Further, a preset interval second-type model transverse rule is adopted to determine a transverse fuzzy post-part definition value corresponding to each transverse normalized linear combination value according to each transverse normalized linear combination value; the preset interval two-type model transverse rules provided by the invention have three, and the three are specifically as follows:

Wherein, Presetting interval two-type model transverse rules for the ith; For input to hidden layer Laterally normalized linear combination values of the individual neurons; To hide layer no First used in neuronsA front piece with two preset interval types and transversely regular patterns; outputting values of the transverse rules of the second type model in the ith preset interval; To hide layer no First used in neuronsAnd a preset interval two-type model transverse fuzzy back piece definition value with transverse rules.

Further, a transverse back-part layer is adopted to determine a plurality of output values of the automobile transverse neuron nodes according to the clear values of the transverse fuzzy back-parts; wherein each transverse normalized linear combination value has a corresponding transverse fuzzy back part definition value to be obtainedAnd (3) withMultiplying to obtain the output of the neuron node, namely the output value of the automobile transverse neuron node; the processing process of the output value of the automobile transverse neuron node specifically comprises the following steps:

Wherein, To hide layer noThe output value of the automobile transverse neuron node of each neuron; Is the first Hidden layer of input connectionLateral weights of individual neurons; To hide layer no First used in neuronsA second preset interval type model transversely regular transverse fuzzy post-part definition value; To hide layer no Linear deviation of individual neurons to the output layer.

Further, the output values of the cross neuron nodes of each automobile are subjected to drop aggregation through a cross output layer, and a target steering angle is generated; wherein, the output set is needed in the fuzzy inference calculationLeft and right end point boundary values of (2)、The output of the transverse control is obtained by the following formula in the process of the degradation of the IT2FSNN network in order to reduce the CPU calculation load through the neuron calculation of the transverse output layerAnd putting the sameAs target steering angleThe formula is specifically as follows:

Wherein, Is the output of the transverse control; is an adaptive proportional variable; is the left endpoint boundary value; Is the right endpoint boundary value; To hide layer no The output value of the automobile transverse neuron node of each neuron; is the lower boundary of the activation intensity; For input to hidden layer A set of lateral amenities of individual neurons; Is an upper boundary of activation intensity; Is the number of nodes in the lateral input layer.

In this embodiment, the vehicle lateral data is input to the preset section two-type model lateral controller to perform lateral control, and the target steering angle is output.

And 103, longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval two-type model longitudinal controller to generate target torque.

The preset interval two-type fuzzy horizontal controller and the preset interval two-type fuzzy vertical controller are both composed of IT2FSNN ((INTERVAL TYPE-2 Fuzzy Set Neutral Network) interval two-type fuzzy neural network, so that the principle of a data processing process of longitudinal control of the preset interval two-type fuzzy vertical controller is the same as that of the preset interval two-type fuzzy horizontal controller, the data processing process is processed according to different inputs to obtain corresponding output, the longitudinal speed control process of the automobile is carried out by adopting the preset interval two-type fuzzy vertical controller, the longitudinal speed of the automobile is transmitted to a longitudinal normalization layer through a longitudinal input layer to carry out linear combination and normalization to generate a plurality of longitudinal normalization linear combination values, the longitudinal normalization linear combination values are subjected to fuzzification by adopting a longitudinal fuzzification layer to generate a plurality of longitudinal boundary function values and a plurality of longitudinal lower boundary function values, the longitudinal upper boundary membership value and the longitudinal lower boundary membership function value are input to a longitudinal corresponding layer to carry out fuzzy multiplication operator to output a plurality of longitudinal closed interval sets, the longitudinal direction two-type fuzzy longitudinal combination values are adopted to carry out longitudinal normalization combination according to preset interval two-type longitudinal rules, the longitudinal speed of the automobile is subjected to longitudinal normalization combination values, the longitudinal combination values of the automobile is subjected to linear combination values are subjected to linear combination of the longitudinal normalization, the longitudinal combination values are obtained through a plurality of the longitudinal normalization neural nodes, and the output node sets are adopted after the longitudinal node clear output values are subjected to output values are determined through the longitudinal node clear-down of the automobile output nodes of the automobile comparison valuesAs the target torque.

In the embodiment, a preset interval two-type model longitudinal controller is adopted to longitudinally control the longitudinal speed of the automobile, so that the target torque is generated.

Step 104, determining a target longitudinal force according to the target torque and the longitudinal speed of the automobile.

Specifically, inputting a target torque to a preset proportional-integral-derivative controller for adjustment, and outputting a longitudinal force demand; determining the longitudinal force of the tire according to the longitudinal speed of the automobile by adopting a preset longitudinal force function of the tire; calculating a target longitudinal force by adopting the longitudinal force demand and the longitudinal force of the tire; the preset proportional-integral-differential controller is PID (Proportion Integration Differentiation), and the sum of the torques of all the wheels of the automatic driving automobile is required to meet the longitudinal force requirement. The invention obtains the longitudinal force of the vehicle according to the difference between the actual speed and the expected speed, and obtains the longitudinal force demand of the vehicle according to the target torque by adopting a PID controller, wherein the processing procedure of the longitudinal force demand comprises the following concrete steps:

Wherein, Is a longitudinal force requirement; Is a proportionality coefficient; is the target torque; is an integral coefficient; sampling time; is a differential coefficient.

Further, presetting a tire longitudinal force function, specifically:

Wherein, Is the tire longitudinal force; Is the longitudinal stiffness of the tire; Is the tire slip ratio; to describe a nonlinear characteristic function caused by tire slip; a braking torque parameter for driving; is the radius of the tire; is the inertia of the tire; Is the longitudinal speed of the car.

Further, torque information transmitted from the preset interval two-type model longitudinal controllerObtaining output of PID controller by PID controlSolving with a preset tire longitudinal force functionThe target longitudinal force is output through a linear relationship and transmitted to the vehicle. The calculation process of the target longitudinal force specifically comprises the following steps:

Wherein, Is the target longitudinal force; is the tire longitudinal force; Is a longitudinal force requirement.

In the present embodiment, the target longitudinal force is determined based on the target torque and the vehicle longitudinal speed.

Step 105, controlling the automatic driving automobile to run through the target steering angle and the target longitudinal force.

It should be noted that the steering angle and the longitudinal force control the running of the automated guided vehicle and control the longitudinal and transverse dynamics of the automated guided vehicle. In addition to lateral control, taking into account the important characteristics of speed tracking, longitudinal control of the vehicle is integrated according to adjusting the accelerating braking portion, improving the stability of the overall performance of the vehicle.

In the present embodiment, the autonomous vehicle is controlled to travel by the target steering angle and the target longitudinal force.

In the embodiment of the invention, the invention provides a longitudinal and transverse control method for an automatic driving automobile, which is used for acquiring transverse data and longitudinal speed of the automobile; inputting the automobile transverse data to a preset interval two-type model transverse controller for transverse control, and outputting a target steering angle; longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval two-type model longitudinal controller to generate a target torque; determining a target longitudinal force according to the target torque and the longitudinal speed of the automobile; controlling the automatic driving automobile to run through the target steering angle and the target longitudinal force; based on the scheme, the transverse data of the automobile is input to the preset interval secondary type model transverse controller to carry out transverse control, and the preset interval secondary type model longitudinal controller is adopted to carry out longitudinal control on the longitudinal speed of the automobile, so that the target steering angle and the target longitudinal force are obtained, the automatic driving automobile is controlled to run, the longitudinal movement and the transverse movement of the automatic driving automobile can be controlled simultaneously in the process, the important characteristic of speed tracking is considered, and the stability of the overall performance of the automobile is further improved.

Referring to fig. 2, fig. 2 is a flowchart illustrating steps of another aspect control method of an automatic driving automobile according to an embodiment of the present invention.

And 201, acquiring transverse data of the automobile to be trained and longitudinal speed of the automobile to be trained, inputting the transverse data of the automobile to be trained into an initial interval two-type model transverse controller for transverse control, and determining a steering angle to be trained.

The automobile transverse data to be trained are automobile transverse data used for training the initial interval two-type model transverse controller; the longitudinal speed of the automobile to be trained is the longitudinal speed of the automobile for the initial interval two-type model longitudinal controller; the automobile transverse data comprise transverse displacement, a transverse swing angle, a transverse displacement error and a transverse swing angle error, wherein the transverse displacement is obtained by controlling a steering wheel of an automatic driving automobile so as to control the vertical distance between a front wheel turning angle and a central line; the yaw angle is obtained through an angle signal of a steering angle sensor of a steering wheel of the automatic driving automobile; the transverse displacement error is obtained by performing difference between preset reference transverse displacement and the acquired transverse displacement; the yaw angle error is obtained by subtracting the obtained yaw angle from a preset reference yaw angle value.

In the embodiment, the transverse data of the automobile to be trained and the longitudinal speed of the automobile to be trained are obtained, the transverse data of the automobile to be trained are input to the initial interval two-type model transverse controller for transverse control, and the steering angle to be trained is determined.

Step 202, constructing a vehicle model corresponding to an automatic driving automobile based on motion dynamics, and determining and adjusting automobile transverse data by adopting the vehicle model according to a steering angle to be trained.

The steering angle to be trained is a vehicle steering angle used for training the initial interval two-type model transverse controller, the vehicle steering angle is an included angle between a limit position which can be reached through the rotation of the front wheels of the vehicle towards the left and right directions and a central line when the wheels are not deflected, and a steering angle sensor is arranged below the steering column of the vehicle and can detect the rotation angle and the steering direction of the steering wheel.

It should be noted that, referring to fig. 3, the process of constructing and training the initial section two-type model transverse controller and the initial section two-type model longitudinal controller is divided into six steps: vehicle dynamics modeling, fuzzy logic design, section two-type fuzzy neural network model design, section two-type fuzzy transverse controller design based on an IT2FSNN network, section two-type fuzzy longitudinal controller design based on an IT2FSNN network and automatic driving automobile implementation parameter setting; for vehicle dynamics modeling, namely, analysis of vehicle path tracking control, to analyze and design the dynamics of the vehicle under the real condition, the dynamics modeling includes seven aspects in total: yaw, lateral motion, longitudinal motion, and rotation of the four wheels. And respectively modeling a vehicle model and a wheel tire model corresponding to the automatic driving automobile based on motion dynamics, and determining the adjusted transverse displacement and yaw angle, namely adjusting the transverse displacement and the adjusted yaw angle, according to the steering angle to be trained through the vehicle model after the vehicle model is obtained. In a vehicle ground coordinate system, the x direction is a longitudinal direction, the y direction is a transverse direction, and the vehicle motion relationship is as follows:

Wherein, Is a lateral position in a ground coordinate system; is the longitudinal position in the ground coordinate system; is the longitudinal speed of the vehicle; Is the lateral speed of the vehicle; Is the yaw angle.

Further, in the design of a vehicle transverse dynamics model, the relation between longitudinal and transverse movements is first determined, which results from the interaction between the road and the vehicle tyre. In the transverse dynamics model, a yaw and transverse dynamics equation is established:

Wherein, Is the moment of inertia; Is yaw rate; longitudinal force of the front right tire; Longitudinal force of the front left tire; Is the front wheel corner; is the wheelbase of the front wheel; a is the distance from the centroid to the front axis; b is the distance from the centroid to the rear axis; is the rear wheel axle distance; Is the lateral force of the front right tire; Is the lateral force of the front left tire; is the total mass of the vehicle; Is the transverse acceleration; is the lateral force of the rear left tire; Is the lateral force of the rear right tire; is the longitudinal force of the rear left tire; is the longitudinal force of the rear right tire.

Further, in vehicle longitudinal dynamics model design, it is necessary to analyze vehicle longitudinal motion, and its stress condition is mainly composed of resistance and longitudinal force. According to newton's second law, vehicle longitudinal dynamics are expressed as follows:

Wherein, Is the total mass of the vehicle; is the longitudinal acceleration; is the tire longitudinal force; Is windage; is rolling resistance; Is the rolling resistance coefficient; g is gravity acceleration; f is the resistance value.

Further, the transverse force and the longitudinal force of the tire need to be set in the design of the tire model, and a preset tire transverse force function and a preset tire longitudinal force function are constructed, specifically:

Wherein, Is the tire lateral force; is the tire longitudinal force; Is the longitudinal stiffness of the tire; Is the tire slip ratio; to describe a nonlinear characteristic function caused by tire slip; a braking torque parameter for driving; is the radius of the tire; is the inertia of the tire; is the longitudinal speed of the automobile; is the lateral stiffness of the tire; Is the steering angle.

Further, for fuzzy logic design, in the design of the invention, a neural network which adopts a section two-type fuzzy logic system (INTERVAL TYPE-2 fuzzy set, IT2 FS) as an original activation function is adopted to adjust the complex dynamics of the automatic driving automobile; specifically, the basic principle of IT2FS is as follows: two-stage model logic system (T2 FS) architecture. Generally, the T2FS is formed by five parts, including a fuzzifier, a rule base, an reasoner, a degrader and a defuzzifier, namely a mapping function, and finally an accurate output is obtained; for T2FS, there are s inputsAn output. The T2FS rule base is composed of IF-THEN rules:

Wherein, The i-th fuzzy rule; inputting a first element of a variable for M dimensions of the fuzzy model; The excitation interval of the ith rule which is the second type ambiguity of the first front part interval is a first type ambiguity set; inputting the s-th element of the variable for M-dimension of the fuzzy model; the two-type fuzzy set is a front piece; Is the output value of the fuzzy rule; the method is a second fuzzy set of the back part; Is a rule base;

Further, membership functions of type II Represented by the following formula, wherein,Is the interval between [0,1], specifically:

Wherein, A second type of fuzzy set for the interval; is the main variable; Is a minor variable; is a membership function of type II; is the discourse of x.

Further, IT2FS needs to be designed for interval two-type modular logic system (IT 2 FS) construction to reduce the computational complexity of T2 FS. In IT2FS, all membership functions of type 2 are considered to be equal to 1, and conversion of the above formula can result:

Wherein, A second type of fuzzy set for the interval; is the main variable; Is a minor variable; is the discourse of x.

Further, for IT2FS, the uncertainty of the interval type two fuzzy set main membership functions constitutes a bounded region, called uncertainty trajectory (Footprint of uncertainty, FOU); the uncertainty trajectory can be expressed as:

Wherein, Representing a region between an upper membership function and a lower membership function as an uncertainty coverage field; is the main variable; Is a minor variable; a discourse domain of x; Is the lower limit of the interval; Is the upper limit of the interval.

Further, in the basic principle of IT2FS, the section two-type fuzzy neural network model is designed, and after the uncertainty influence is applied to the front part of the T2FS rule and the deduction, reduction and defuzzification processes are completed, the output of the section two-type fuzzy neural network (IT 2 FSNN) can be obtained, which indicates that the section two-type fuzzy neural network (IT 2 FSNN) is constructed, the fuzzy set is used as the front part set of IT2FSNN, the IT2FS is used as the activation function of the neural network, and the values obtained through the processes of fuzzification, reasoning, defuzzification and the like are used as the back part of the neural network. For the section two-type fuzzy transverse controller design based on the IT2FSNN network, the section two-type fuzzy longitudinal controller design based on the IT2FSNN network and unique uncertainty modeling characteristics existing in the IT2FS structure, the built section two-type fuzzy neural network (IT 2 FSNN) is applied to control the nonlinearity and the complex dynamics of an automatic driving automobile, so that the design of the initial section two-type fuzzy transverse controller and the initial section two-type fuzzy longitudinal controller is completed, and finally, the model training is carried out on the initial section two-type fuzzy transverse controller and the initial section two-type fuzzy longitudinal controller, so that the trained preset section two-type fuzzy transverse controller and the preset section two-type fuzzy longitudinal controller can be obtained.

Further, parameter setting is implemented for the automatic driving automobile, namely, the setting of the parameter value of the automatic driving automobile according to the invention is implemented by the following specific parameter formats: total mass of vehicle: 1294; yaw inertia: 1630; front wheel and vehicle center of gravity spacing: 1, a step of; distance of rear wheel from center of gravity of vehicle: 1.459; lateral stiffness of the front wheels: 90000; lateral stiffness of the rear wheels: 90000; radius of tire: 0.35; coefficient of rolling resistance: 0.01.

Further, referring to fig. 4, in the design of the vehicle dynamics model, various forces and moments of the vehicle in longitudinal and lateral motions, including the physical quantities related to factors such as the vehicle mass, the tire characteristics, the driving torque, etc., are described based on the above equations, so as to obtain an autopilot vehicle dynamics model, which is very important for designing the path tracking controller (the initial section two-type fuzzy lateral controller and the initial section two-type fuzzy longitudinal controller) and other control systems of the vehicle.

In the embodiment, a vehicle model corresponding to an automatic driving automobile is constructed based on motion dynamics, and the vehicle model is adopted to determine and adjust automobile transverse data according to a steering angle to be trained.

And 203, optimizing the initial interval two-type model transverse controller by adopting the adjustment automobile transverse data, and determining the preset interval two-type model transverse controller.

Further, step 203 may comprise the sub-steps of:

s31, calculating a transverse loss value according to the transverse data of the adjusted automobile, the preset reference transverse displacement and the preset reference transverse swing angle;

S32, updating the super parameters of the second-type model transverse controller in the initial interval by adopting a transverse loss value, determining the second-type model transverse controller in the middle interval, and counting the iteration times in real time;

S33, judging whether the iteration times reach preset training times or not;

And S34, if the iteration times reach the preset training times, taking the middle section second type model transverse controller as the preset section second type model transverse controller.

Adjusting the vehicle lateral data includes adjusting lateral displacement and adjusting yaw angle.

It should be noted that, assuming that the longitudinal speed is constant, the transverse dynamics of the vehicle is set to be irrelevant to the longitudinal dynamics, and the section two-type model transverse controller designed by the invention comprises 4 inputs: lateral displacement, yaw angle, lateral displacement error, and yaw angle error; according to the transverse data of the automobile, the preset reference transverse displacement and the preset reference yaw angle, the transverse loss value is calculated, and the calculation process of the transverse loss value is specifically as follows:

Wherein ^LCL_f1 is the lateral loss value; a preset reference lateral displacement; to adjust the lateral displacement; A reference yaw angle is preset; To adjust the yaw angle; Is transposed.

Further, the super-parameters of the initial interval two-type model transversal controller comprise an initial transversal connection hidden layer (LLL)Individual neurons and output layerThe weight of the link of the output and the first third preset interval second model transverse rule and hidden layerOffset associated with linear combination of inputs to individual neurons, initial first preset interval second model cross-rule back part portion and hidden layer firstDeviations associated with linear combinations of inputs to individual neurons, vertical distance of lower membership functions of initial second preset interval second model transverse rule, initial transverse connection hidden layerNeurons of the first groupThe current layer of the initial transverse link is connected with the hidden layerDeviations associated with linear combinations of individual neuron inputs; the process for updating the super parameters of the initial interval two-type model transverse controller by adopting the transverse loss value comprises the following steps:

Wherein, Conceal layer for intermediate lateral connectionIndividual neurons and output layerThe weights of the links of the outputs; conceal layer first for initial lateral connection Individual neurons and output layerThe weights of the links of the outputs; Transverse rules and hidden layer(s) of the second type of model for the initial third preset interval Deviations associated with linear combinations of inputs to individual neurons; second type of model transverse rule and hidden layer for the third preset interval in the middle Deviations associated with linear combinations of inputs to individual neurons; Two-level model of the middle first preset interval is transversely regular back part and hidden layer Deviations associated with linear combinations of inputs to individual neurons; The second section of the model is a horizontally regular back part and hidden layer for the first preset section Deviations associated with linear combinations of inputs to individual neurons; Hidden layer for middle lateral current layer connection Deviations associated with linear combinations of individual neuron inputs; Hidden layer for initial lateral current layer connection Deviations associated with linear combinations of individual neuron inputs; the vertical distance of the lower membership function of the second model transverse rule in the second preset interval in the middle; The vertical distance of the lower membership function of the second model transverse rule of the initial second preset interval; conceal layer for intermediate lateral connection Neurons of the first groupWeights of the links of the inputs; conceal layer first for initial lateral connection Neurons of the first groupWeights of the links of the inputs; is the transverse loss value; Is a parameter of preset convergence rate.

Further, the partial derivative required for the process of updating the super parameter of the model transversal controller in the initial interval two-stage model can be obtained by the following formula:

Wherein, For the Jacobian matrix of the vehicle model,；Is the steering angle; hidden layer number in a cross controller for an initial interval two-type model The output of the individual neurons; hidden layer number in a cross controller for an initial interval two-type model Normalized linear combination values of the individual neuron inputs.

Further, if the iteration times reach the preset training times, taking the middle section second type model transverse controller as the preset section second type model transverse controller; if the iteration number does not reach the preset training number, taking the middle section two-type model transverse controller as a new initial section two-type model transverse controller, jumping to execute the step 201 until the iteration number reaches the preset training number, and taking the middle section two-type model transverse controller as the preset section two-type model transverse controller when the iteration number reaches the preset training number to determine.

In this embodiment, the initial section two-type model transverse controller is optimized by adjusting the vehicle transverse data, and the preset section two-type model transverse controller is determined.

And 204, optimizing the initial interval two-type model longitudinal controller by adopting the longitudinal speed of the automobile to be trained, and determining the preset interval two-type model longitudinal controller.

It should be noted that, the training principle of the initial interval second-type model longitudinal controller is consistent with the training principle of the initial interval second-type model transverse controller, specifically, the initial interval second-type model longitudinal controller is used for longitudinally controlling the longitudinal speed of the automobile to be trained, outputting the torque to be trained, and calculating the longitudinal loss value by adopting the torque to be trained and the preset reference torque, wherein the calculation process of the longitudinal loss value specifically includes:

Wherein ^LCL_f2 is the longitudinal loss value; The torque is preset reference torque; Is the torque to be trained.

Further, in designing a proper path-following control method, in addition to steering control, speed-following should be considered, which involves adjustment of an acceleration/braking section, adjusting the vehicle speed according to the path traversed, the present invention also designs an initial interval two-section fuzzy longitudinal controller, and evaluates the performance of the controller in the speed-following problem; based on the torque adjustment and longitudinal speed control relationship in the above formula, the longitudinal controller is based on the appropriate torque adjustment.

Further, updating the super parameters of the initial interval two-type model longitudinal controller through the longitudinal loss value to obtain an intermediate interval two-type model longitudinal controller, counting the current iteration times, and taking the intermediate interval two-type model longitudinal controller as a trained preset interval two-type model longitudinal controller if the current iteration times reach a preset training time threshold; if the current iteration number does not reach the preset training number threshold, taking the middle section second type fuzzy longitudinal controller as a new initial section second type fuzzy longitudinal controller, jumping to execute the steps of longitudinally controlling the longitudinal speed of the automobile to be trained through the initial section second type fuzzy longitudinal controller and outputting the torque to be trained until the current iteration number reaches the preset training number threshold, and taking the middle section second type fuzzy longitudinal controller as a trained preset section second type fuzzy longitudinal controller when the current iteration number reaches the preset training number threshold; wherein, because the initial section two-type fuzzy longitudinal controller and the initial section two-type fuzzy longitudinal controller both apply the section two-type fuzzy neural network (IT 2 FSNN), the principle of updating the super-parameters of the initial section two-type fuzzy longitudinal controller is consistent with the principle of updating the super-parameters of the initial section two-type fuzzy transverse controller through the transverse loss value, and the Jacobian matrix required by the updating process of the super-parameters of the initial section two-type fuzzy longitudinal controller is obtained through the IT2FSNN identifier; the super parameters of the initial interval two-type model longitudinal controller comprise an initial longitudinal connection hiding layerIndividual neurons and output layerThe weight of the output link, the initial third preset interval second model longitudinal rule and the hidden layerOffset associated with linear combination of inputs to individual neurons, initial first preset interval second model longitudinal rule back part and hidden layer firstDeviation associated with linear combination of inputs of individual neurons, vertical distance of lower membership function of initial second preset interval second model longitudinal rule, initial longitudinal connection hidden layerNeurons of the first groupThe weights of the links of the inputs.

In the embodiment, the longitudinal speed of the automobile to be trained is adopted to optimize the longitudinal controller of the secondary model in the initial interval, and the longitudinal controller of the secondary model in the preset interval is determined.

Step 205, acquiring automobile transverse data and automobile longitudinal speed.

And 206, inputting the automobile transverse data into a preset interval two-type model transverse controller to carry out transverse control, and outputting a target steering angle.

The preset interval two-type model transverse controller comprises a transverse input layer, a transverse normalization layer, a transverse blurring layer, a transverse corresponding layer, a transverse back piece layer and a transverse output layer.

Further, step 206 may include the sub-steps of:

s61, transmitting the automobile transverse data to a transverse normalization layer through a transverse input layer to perform linear combination and normalization to generate a plurality of transverse normalized linear combination values;

S62, blurring each transverse normalized linear combination value by adopting a transverse blurring layer to generate a plurality of transverse boundary membership function values and a plurality of transverse lower boundary membership function values;

S63, inputting each transverse boundary membership function value and each transverse lower boundary membership function value to a transverse corresponding layer to carry out fuzzy product operator, and outputting a plurality of transverse closed interval sets;

s64, determining a clear value of the transverse fuzzy post-part corresponding to each transverse normalized linear combination value according to each transverse normalized linear combination value by adopting a preset interval two-level model transverse rule;

S65, determining a plurality of automobile transverse neuron node output values by adopting a transverse back-part layer according to the clear values of all transverse fuzzy back-parts;

S66, performing drop aggregation on the output values of the cross neuron nodes of each automobile and the cross closed zone sets through the cross output layer to generate a target steering angle.

And 207, longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval two-type model longitudinal controller to generate target torque.

The preset interval two-type model longitudinal controller comprises a longitudinal input layer, a longitudinal normalization layer, a longitudinal blurring layer, a longitudinal corresponding layer, a longitudinal back piece layer and a longitudinal output layer.

Further, step 207 may comprise the sub-steps of:

S71, transmitting the longitudinal speed of the automobile to a longitudinal normalization layer through a longitudinal input layer to perform linear combination and normalization, and generating a plurality of longitudinal normalized linear combination values;

S72, blurring each longitudinal normalized linear combination value by adopting a longitudinal blurring layer to generate a plurality of longitudinal upper bound membership function values and a plurality of longitudinal lower bound membership function values;

s73, inputting each longitudinal upper-bound membership function value and each longitudinal lower-bound membership function value into a longitudinal corresponding layer to carry out fuzzy product operator, and outputting a plurality of longitudinal closed interval sets;

S74, determining a longitudinal fuzzy back piece clear value corresponding to each longitudinal normalized linear combination value according to each longitudinal normalized linear combination value by adopting a preset interval second type model longitudinal rule;

S75, determining a plurality of longitudinal neuron node output values of the automobile by adopting a longitudinal back part layer according to the clear values of the longitudinal fuzzy back parts;

s76, performing drop aggregation on the output values of the longitudinal neuron nodes of each automobile and the longitudinal closed interval sets through the longitudinal output layer to generate target torque.

Step 208, determining a target longitudinal force according to the target torque and the longitudinal speed of the automobile.

Further, step 208 may include the sub-steps of:

s81, inputting a target torque to a preset proportional-integral-derivative controller for adjustment, and outputting a longitudinal force demand;

s82, determining the longitudinal force of the tire according to the longitudinal speed of the automobile by adopting a preset longitudinal force function of the tire;

s83, calculating target longitudinal force by adopting the longitudinal force demand and the longitudinal force of the tire.

Step 209, controlling the automatic driving automobile to run through the target steering angle and the target longitudinal force.

It should be noted that referring to fig. 5, the overall block diagram for controlling the longitudinal and lateral movements of the automatic driving vehicle simultaneously includes four basic parts: a reference information part, a control module (a preset interval two-type model transverse controller and a preset interval two-type model longitudinal controller), an information collection module and a vehicle model; in particular, the reference information part comprises a preset reference lateral displacement for the training phasePreset reference yaw angle for maintaining vehicle stabilityAnd preset reference torque。

Further, in the control module, the preset interval two-type model transverse controller and the preset interval two-type model longitudinal controller simultaneously transmit steering angles and torques to a steering part and an acceleration braking part in the vehicle model by using the IT2FSNN model. Enabling the vehicle to perform path and speed tracking tasks simultaneously with high accuracy.

Further, the information collection module receives state vectors of the longitudinal and lateral motion models of the vehicle via the sensorsAnd control signals for steering angle and torque.

As a comparison of technical effects, which can be referred to in combination with the prior art, vehicle driving safety is a core requirement of modern road transportation, and proper control systems need to be designed and developed to minimize the occurrence of decision errors. With the increasing traffic on roads and the acceleration of the urban process, the need for vehicle safety has become more and more stringent. Therefore, designing and developing an appropriate control system to minimize accident risk and improve driver's safety is one of the primary tasks of the industry. The modern automotive industry is faced with a great challenge because the vehicle control system must perform well in a variety of complex road conditions and extreme weather conditions. This not only means that the control system must be able to make decisions quickly and accurately, but also needs to take into account factors such as the handling and stability of the vehicle. These challenges are pushing the continual progress and innovation of driving assistance systems and autopilot technology.

Path tracking is a vital task in autonomous vehicles that requires the control system to accurately identify and follow the road, making adjustments based on real-time traffic conditions. However, in implementing path tracking, in addition to considering lateral movement of the vehicle, longitudinal movement must also be considered at the same time, adding to the complexity of the control system design.

In the design of control systems, the application of neural network technology provides new ideas for handling vehicle dynamics and environmental changes. The neural network is powerful in that it is capable of learning complex nonlinear relationships from input data, thereby enabling more accurate control and decision making. At the same time, environmental uncertainties, such as abrupt weather changes or other environmental disturbances, and uncertainties in physical parameters, are considered, which can significantly affect the performance of the controller. The intelligent control technology based on fuzzy set also becomes an alternative scheme, which can better cope with the influence of uncertainty factors on the system performance, but the prior art has the following defects: model uncertainty: autopilot vehicles face complex and varied environments on real roads, and the dynamics models of the autopilot vehicles are often affected by uncertainty. Uncertainty in terms of nonlinear dynamics of the vehicle, road condition changes, vehicle quality, etc. may lead to degraded performance of the path tracking controller; sensor error: autopilot vehicles rely on various sensors to sense the surrounding environment, such as lidar, cameras, radar, etc. These sensors may be affected by weather, light, etc., resulting in sensing errors, thereby affecting the accuracy of path tracking; complex environment: automatic driving automobiles travel in various complex environments such as cities, highways and the like, including intersections, crosswalks, buildings and the like. These scenarios may place higher demands on the path tracking controller, requiring more challenges to be overcome; robustness: the path tracking controller needs to exhibit robustness under a variety of different road conditions and traffic conditions. For example, the control system should still be able to track the path efficiently in the face of an emergency or bad weather.

In view of the above problems, the present invention provides a method for controlling the longitudinal and transverse directions of an automatic driving vehicle, which combines a two-section type fuzzy rule system with a two-section type fuzzy neural network (IT 2 FSNN). The fuzzy set is used as a front piece set of IT2FSNN, IT2FS is used as an activation function of a neural network, values obtained through the processes of fuzzification, reasoning, de-modeling and the like are used as rear pieces of the neural network, an output result of the fuzzy neural network is used as a target steering angle output by a preset interval two-type fuzzy transverse controller and a target torque output by a preset interval two-type fuzzy longitudinal controller, meanwhile, a PID controller is used for adjusting longitudinal force, longitudinal force obtained by transmitting output parameters of the longitudinal controller module to an acceleration braking module and longitudinal force obtained by transmitting the longitudinal force to PID optimization are used for optimally adjusting vehicle longitudinal force through linear combination, and therefore final target longitudinal force is output, and the neural network suitable for generating linear and complex mapping is selected: through the neural network, the system can learn and adapt to complex vehicle dynamics characteristics, and more flexible control is realized; IT2FS is combined with a neural network to generate a section two-type model neural network, so that the adaptability of the network to complex mapping is improved, and the uncertainty of parameters is effectively treated; the longitudinal controller is combined with the transverse controller and is used for simultaneously controlling longitudinal and transverse dynamic movements of the automatic driving automobile, the simultaneous control of the transverse and longitudinal movements of the automobile is realized by using the input of the angle and the torque of the steering wheel, the path tracking and the speed tracking targets can be simultaneously realized, and the overall movement control performance of the automatic driving automobile is improved; the PID controller is introduced to adjust the longitudinal force and directly transmit the longitudinal force linear adjustment to the output of the acceleration braking control module, so that better path tracking and speed tracking capabilities are shown.

Referring to fig. 6, fig. 6 is a block diagram illustrating a longitudinal and transverse control device for an automatic driving vehicle according to an embodiment of the present invention.

The invention provides an automatic driving automobile longitudinal and transverse control device, which comprises:

an acquisition module 601, configured to acquire vehicle transverse data and vehicle longitudinal speed;

the transverse control module 602 is used for inputting the transverse data of the automobile to a preset interval two-type model transverse controller to carry out transverse control and outputting a target steering angle;

the longitudinal control module 603 is used for longitudinally controlling the longitudinal speed of the automobile by adopting a preset interval two-type model longitudinal controller to generate target torque;

According to block 604, a target longitudinal force is determined based on the target torque and the longitudinal speed of the vehicle;

the driving module 605 is used for controlling the automatic driving automobile to drive through the target steering angle and the target longitudinal force.

Further, the preset interval two-type model transverse controller comprises a transverse input layer, a transverse normalization layer, a transverse blurring layer, a transverse corresponding layer, a transverse back piece layer and a transverse output layer; a lateral control module 602, comprising:

The first transverse sub-module is used for transmitting the transverse data of the automobile to the transverse normalization layer through the transverse input layer to perform linear combination and normalization to generate a plurality of transverse normalized linear combination values;

the second transverse submodule is used for fuzzifying all transverse normalized linear combination values by adopting a transverse fuzzifying layer to generate a plurality of transverse boundary membership function values and a plurality of transverse lower boundary membership function values;

the third transverse submodule is used for inputting each transverse boundary membership function value and each transverse lower boundary membership function value to a transverse corresponding layer to carry out fuzzy product operator and outputting a plurality of transverse closed interval sets;

the fourth transverse submodule is used for determining a clear value of the transverse fuzzy part corresponding to each transverse normalized linear combination value according to each transverse normalized linear combination value by adopting a preset interval two-type fuzzy transverse rule;

a fifth transverse sub-module, configured to determine a plurality of output values of the vehicle transverse neuron nodes according to the clear values of the transverse fuzzy afterparts by using the transverse afterparts layer;

and the sixth transverse submodule is used for carrying out drop aggregation on the output value of each automobile transverse neuron node and each transverse closed interval set through the transverse output layer to generate a target steering angle.

Further, the preset interval two-type model longitudinal controller comprises a longitudinal input layer, a longitudinal normalization layer, a longitudinal blurring layer, a longitudinal corresponding layer, a longitudinal back piece layer and a longitudinal output layer; a longitudinal control module 603 comprising:

the first longitudinal submodule is used for transmitting the longitudinal speed of the automobile to the longitudinal normalization layer through the longitudinal input layer to perform linear combination and normalization to generate a plurality of longitudinal normalized linear combination values;

The second longitudinal submodule is used for carrying out fuzzification on each longitudinal normalized linear combination value by adopting a longitudinal fuzzification layer to generate a plurality of longitudinal upper bound membership function values and a plurality of longitudinal lower bound membership function values;

the third longitudinal sub-module is used for inputting each longitudinal upper bound membership function value and each longitudinal lower bound membership function value into a longitudinal corresponding layer to carry out fuzzy product operator and outputting a plurality of longitudinal closed interval sets;

The fourth longitudinal submodule is used for determining a longitudinal fuzzy back part clear value corresponding to each longitudinal normalized linear combination value according to each longitudinal normalized linear combination value by adopting a preset interval second type fuzzy longitudinal rule;

A fifth longitudinal sub-module, configured to determine a plurality of longitudinal neuron node output values of the vehicle by using a longitudinal back-piece layer according to the respective longitudinal blurred back-piece definition values;

And the sixth longitudinal submodule is used for carrying out drop aggregation on the output value of each longitudinal neuron node of the automobile and each longitudinal closed interval set through the longitudinal output layer to generate target torque.

Further, according to block 604, comprising:

The adjusting submodule is used for inputting the target torque to a preset proportional-integral-derivative controller for adjustment and outputting the longitudinal force requirement;

The longitudinal force sub-module is used for determining the longitudinal force of the tire according to the longitudinal speed of the automobile by adopting a preset longitudinal force function of the tire;

And the target sub-module is used for calculating the target longitudinal force by adopting the longitudinal force requirement and the tire longitudinal force.

In an alternative embodiment, the apparatus further comprises:

The first module is used for acquiring transverse data of the automobile to be trained and the longitudinal speed of the automobile to be trained, inputting the transverse data of the automobile to be trained into the initial interval two-type model transverse controller for transverse control, and determining the steering angle to be trained;

The second module is used for constructing a vehicle model corresponding to the automatic driving automobile based on motion dynamics, and determining and adjusting automobile transverse data by adopting the vehicle model according to the steering angle to be trained;

The third module is used for optimizing the initial interval second-type model transverse controller by adopting the adjustment automobile transverse data and determining the preset interval second-type model transverse controller;

and the fourth module is used for optimizing the initial interval second-type model longitudinal controller by adopting the longitudinal speed of the automobile to be trained and determining the preset interval second-type model longitudinal controller.

Further, a third module, comprising:

The first training submodule is used for calculating a transverse loss value according to the transverse data of the adjusted automobile, the preset reference transverse displacement and the preset reference transverse swing angle;

The second training submodule is used for updating the super parameters of the second-type model transverse controller in the initial interval by adopting the transverse loss value, determining the second-type model transverse controller in the middle interval and counting the iteration times in real time;

The third training sub-module is used for judging whether the iteration times reach preset training times or not;

And the fourth training sub-module is used for taking the middle section second type model transverse controller as the preset section second type model transverse controller if the iteration times reach the preset training times.

It will be clearly understood by those skilled in the art that, for convenience and brevity of description, specific working procedures of the apparatus, modules and sub-modules described above may refer to corresponding procedures in the foregoing method embodiments, which are not described herein again.

In the several embodiments provided by the present application, it should be understood that the disclosed apparatus and method may be implemented in other manners. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

The above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for controlling the longitudinal and transverse directions of an automatic driving automobile, comprising the following steps:

acquiring transverse data and longitudinal speed of the automobile;

Controlling the automatic driving automobile to run through the target steering angle and the target longitudinal force;

the preset interval two-type model transverse controller comprises a transverse input layer, a transverse normalization layer, a transverse blurring layer, a transverse corresponding layer, a transverse back piece layer and a transverse output layer; the step of inputting the automobile transverse data to a preset interval two-type model transverse controller for transverse control and outputting a target steering angle comprises the following steps:

Performing drop aggregation on the output value of each automobile transverse neuron node and each transverse closed interval set through a transverse output layer to generate a target steering angle;

The preset interval two-type model longitudinal controller comprises a longitudinal input layer, a longitudinal normalization layer, a longitudinal blurring layer, a longitudinal corresponding layer, a longitudinal back piece layer and a longitudinal output layer; the step of adopting a preset interval two-type model longitudinal controller to longitudinally control the longitudinal speed of the automobile and generating target torque comprises the following steps:

performing drop aggregation on the output value of each longitudinal neuron node of the automobile and each longitudinal closed interval set through a longitudinal output layer to generate target torque;

the step of determining a target longitudinal force based on the target torque and the vehicle longitudinal speed comprises:

2. The method for controlling the longitudinal and transverse directions of an automatically driven automobile according to claim 1, comprising, before the step of acquiring the lateral data and the longitudinal speed of the automobile:

3. The method of claim 2, wherein the step of optimizing the initial zone two-level fuzzy lateral controller using the adjusted vehicle lateral data to determine the preset zone two-level fuzzy lateral controller comprises:

judging whether the iteration times reach preset training times or not;

4. An automatic steering car longitudinal and lateral control device, characterized by comprising:

The driving module is used for controlling the automatic driving automobile to drive through the target steering angle and the target longitudinal force;

the preset interval two-type model transverse controller comprises a transverse input layer, a transverse normalization layer, a transverse blurring layer, a transverse corresponding layer, a transverse back piece layer and a transverse output layer; the lateral control module comprises:

The second transverse submodule is used for fuzzifying the transverse normalized linear combination values by adopting a transverse fuzzifying layer to generate a plurality of transverse boundary membership function values and a plurality of transverse lower boundary membership functions;

the fourth transverse submodule is used for determining a transverse fuzzy back piece clear value corresponding to each transverse normalized linear combination value according to each transverse normalized linear combination value by adopting a preset interval two-type fuzzy transverse rule;

a fifth transverse sub-module, configured to determine a plurality of output values of the vehicle transverse neuron nodes according to the clear values of the transverse fuzzy afterparts by using a transverse afterparts layer;

the sixth transverse submodule is used for carrying out drop aggregation on the output value of each automobile transverse neuron node and each transverse closed interval set through a transverse output layer to generate a target steering angle;

The preset interval two-type model longitudinal controller comprises a longitudinal input layer, a longitudinal normalization layer, a longitudinal blurring layer, a longitudinal corresponding layer, a longitudinal back piece layer and a longitudinal output layer; a longitudinal control module comprising:

the fourth longitudinal submodule is used for determining a longitudinal fuzzy back part definition value corresponding to each longitudinal normalized linear combination value according to each longitudinal normalized linear combination value by adopting a preset interval two-type model longitudinal rule;

the sixth longitudinal submodule is used for carrying out drop aggregation on the output value of each longitudinal neuron node of the automobile and each longitudinal closed interval set through a longitudinal output layer to generate target torque;

the according module comprises:

the longitudinal force sub-module is used for determining the longitudinal force of the tire according to the longitudinal speed of the automobile by adopting a preset tire longitudinal force function;

A target sub-module for calculating a target longitudinal force using the longitudinal force demand and the tire longitudinal force.

5. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method for controlling the longitudinal and lateral directions of an autonomous car according to any one of claims 1 to 3.

6. A computer readable storage medium having stored thereon a computer program/instruction, which when executed by a processor, implements the steps of the method for controlling the longitudinal and lateral directions of an autonomous car as claimed in any one of claims 1 to 3.

7. A computer program product comprising computer programs/instructions which, when executed by a processor, implement the steps of the method for controlling the longitudinal and transverse directions of an autonomous car as claimed in any of claims 1 to 3.