CN115432009B - An automatic driving vehicle trajectory tracking control system - Google Patents

Abstract

The embodiment of the application discloses an automatic driving vehicle track tracking control system, which comprises: the system comprises a signal processing subsystem, a parameter adaptation module, a system model library, an optimization solver, an emergency braking module and a system control module. The system can generate a corresponding prediction time domain according to the current vehicle state information and road information, solves the optimal rotation angle and acceleration by utilizing an optimization function under the condition of meeting the constraint condition of the system, and converts the optimal rotation angle and acceleration into control signals through a system control module, so that an automatic driving vehicle runs according to an expected track and an expected vehicle speed, and the automatic driving vehicle speed change control is realized. When the obstacle suddenly appears in front is detected, the system can automatically generate corresponding acceleration through the current speed and the distance from the vehicle to the obstacle, so that emergency braking obstacle avoidance is realized, and the safety of drivers and passengers is ensured.

Description

An automatic driving vehicle trajectory tracking control system

technical field

The invention relates to the field of automatic driving, and is especially suitable for the field of automatic driving vehicle trajectory tracking control system.

Background technique

The key technologies of autonomous vehicles mainly include: environment perception, behavior decision-making, path planning and vehicle motion control. Motion control is the last link, but also a very important one. Research on motion control is mainly divided into two categories: lateral control and longitudinal control: lateral control is mainly to control the front wheel angle of the car to complete the tracking of the desired trajectory and ensure the stability of the vehicle; longitudinal control is to track the speed of the vehicle. To ensure that the vehicle adapts to different road environments by changing its speed. Generally, vehicles need the cooperation of lateral control and longitudinal control to achieve the purpose of driving in different states. Although many scholars have done research on the lateral control, longitudinal control and integrated control of vehicles, there are still deficiencies:

Many studies use model predictive control method for trajectory tracking control, which can limit various constraints through the objective function, so as to ensure the accuracy and stability of vehicle trajectory tracking. Prediction time domain is a key parameter in model predictive control, and its size directly affects the control effect of trajectory tracking.

However, many studies regard the prediction time domain in the model predictive control method as a fixed value, which makes the method unable to adapt to the working conditions with large speed changes, and when the prediction time domain is not well selected, large trajectory deviations will occur, resulting in relatively large Large tracking lags greatly reduce the accuracy of vehicle trajectory tracking.

In addition, many implementations combine the lateral and longitudinal couplings, which is a very complex design, so most scholars only consider lateral control when designing controllers, that is, control the front wheel angle, and set the vehicle speed to a constant speed. However, the speed of the vehicle needs to change according to the surrounding environment. Only by controlling the front wheel angle to achieve trajectory tracking will cause the vehicle tires to rub improperly on the ground, prone to dangerous conditions such as rear axle slipping or tail flicking, which greatly increases hidden dangers. generation.

To sum up, the research of many model predictive control methods at this stage has a constant prediction time domain, but when the vehicle is driving on the actual road, the road curvature and the road environment are changing all the time, and the speed may be changed at any time. It is possible to maintain a constant speed at all times, which leads to poor control accuracy of the traditional model predictive control method, which is prone to danger when changing speeds, and cannot cope with emergencies.

Contents of the invention

An embodiment of the present invention provides an automatic driving vehicle trajectory tracking control system, including:

a signal processing subsystem for determining the curvature of the road ahead and state information of the vehicle;

A parameter adaptation module, the parameter adaptation module includes a prediction time-domain neural network and an adapter, and the prediction time-domain neural network is used for processing according to the current vehicle speed, the expected vehicle speed and the curvature of the road ahead to obtain the corner prediction time-domain parameters and acceleration Prediction time domain parameters; the adapter is used to select the corresponding rotation angle prediction time domain and acceleration prediction time domain according to the size of the rotation angle prediction time domain parameters and acceleration prediction time domain parameters in the custom corner interval group and acceleration interval group;

A system model library, the system model library is used to obtain the output parameters of the rotation angle prediction model and the output parameters of the acceleration prediction model according to the state information of the vehicle, the rotation angle prediction time domain and the acceleration prediction time domain;

An optimization solver, the optimization solver is used to obtain the rotation angle control amount and the acceleration control amount according to the expected trajectory, the desired vehicle speed, the output parameters of the rotation angle prediction model and the rotation angle prediction time domain, the acceleration prediction model output parameters and the acceleration prediction time domain;

A system control module, the system control module is used to receive the control amount of rotation angle and the acceleration control amount and generate corresponding control commands to control the vehicle to perform corresponding deflection and acceleration and deceleration operations.

The embodiment of the present invention also provides a track tracking control method for an automatic driving vehicle, including:

Detecting the road environment in front of the vehicle according to the object detection module in the signal processing subsystem, and judging whether there is an obstacle in the road ahead;

Generate corresponding corner prediction time domain parameters and acceleration prediction time domain parameters based on the current vehicle speed, expected vehicle speed and the road curvature via the prediction time domain neural network in the parameter adaptation module;

Then the adapter selects the corresponding rotation angle prediction time domain and acceleration prediction time domain within the range based on the size of the rotation angle prediction time domain parameter and the acceleration prediction time domain parameter;

The steering angle prediction model in the system model library obtains the output parameters of the steering angle prediction model according to the vehicle state information obtained by the state estimation module and the steering angle prediction time domain;

The acceleration prediction model in the system model library obtains the output parameters of the acceleration prediction model according to the vehicle state information and the acceleration prediction time domain;

The corner optimization function in the optimization solver is based on the expected trajectory, according to the output parameters of the corner prediction model and the corner control amount obtained in the time domain of the corner prediction;

The acceleration optimization function in the optimization solver is based on the expected vehicle speed, and the acceleration control amount is obtained according to the output parameters of the speed prediction model and the acceleration prediction time domain;

After the logic converter in the system control module receives the angle control amount and the acceleration control amount, the instruction generator in the system control module generates corresponding control instructions to control the vehicle to perform corresponding deflection and acceleration and deceleration operations.

Compared with the prior art, the embodiment of the present invention provides a trajectory tracking control system for an automatic driving vehicle. The system relies on the corner prediction model and the acceleration prediction model to realize the vertical and horizontal comprehensive control of the automatic driving vehicle, so that the vehicle can travel on the actual road. It can automatically adjust the steering angle and driving speed according to changes in road curvature and road environment; moreover, the system can automatically generate the corresponding prediction time domain according to the driving speed and road information, so that the vehicle can meet the accuracy of trajectory tracking when driving at variable speeds. It can also ensure driving stability and improve the dynamic control performance and anti-interference performance of self-driving vehicles; finally, the system can detect the road ahead through the object detection module. When an obstacle suddenly appears in the front, it can automatically generate the corresponding acceleration according to the current vehicle speed and the distance to the obstacle, realize emergency braking and avoid obstacles, and ensure the driving safety of drivers and passengers.

Description of drawings

In order to make the technical solutions and advantages in the embodiments of the present application clearer, the exemplary embodiments of the present application will be further described in detail below in conjunction with the accompanying drawings. Embodiments, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative effort.

Figure 1 is a schematic diagram of the specific composition of the modules of the vehicle trajectory tracking system;

Fig. 2 is a schematic diagram of the workflow of the vehicle trajectory tracking system;

Fig. 3 is a schematic diagram of a vehicle corner control model established in conjunction with a tire model;

Fig. 4 is a schematic flow chart of the logic converter in working mode 1;

Fig. 5 is a schematic flow chart of the logic converter working mode 2;

Fig. 6 is a schematic flowchart of a vehicle trajectory tracking method.

Detailed ways

The solutions provided by the embodiments of this specification are described below in conjunction with the accompanying drawings. It should be understood that the specific embodiments described here are only used to explain related inventions, rather than to limit the invention. The described embodiments are only some of the embodiments in this specification, not all of them. Based on the embodiments in this specification, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present application.

The embodiment of the present invention discloses a trajectory tracking control system for an automatic driving vehicle, which is composed of a signal processing subsystem, a system model library, a parameter adaptation module, an optimization solver and a system control module, wherein the signal processing subsystem includes a state Estimation module and object detection module. When the expected trajectory and expected vehicle speed are known, the system can generate the corresponding prediction time domain according to the current vehicle state information and road information, and use the optimization function to solve the optimal rotation angle and acceleration under the system constraints, and The system control module converts it into a control signal, so that the self-driving vehicle can drive according to the expected trajectory and speed, and realize the variable speed control of the self-driving vehicle. During the driving process, the data of the on-board sensor components is used to continuously update the current vehicle state information and road information through the state estimation module, and repeat the above process, and finally complete the trajectory tracking control of the autonomous driving vehicle. When an obstacle suddenly appears in front of the vehicle, the system can automatically generate the corresponding acceleration based on the current vehicle speed and the distance to the obstacle, so as to realize emergency braking to avoid obstacles and ensure the safety of drivers and passengers.

The specific composition of the modules of the vehicle trajectory tracking system is shown in Figure 1. The specific modules are implemented as follows:

S110: Signal processing subsystem

The signal processing subsystem consists of a state estimation module and an object detection module, including:

S111: state estimation module

The state estimation module can perform estimation calculations based on the measurement data obtained by the on-board sensor components, so as to obtain the state information of the autonomous vehicle.

S112: Object detection module

The object detection module includes a binocular camera outside the vehicle and a laser radar, which can detect the driving environment in front of the vehicle and obtain the road curvature and obstacle information of the road ahead, including the distance to the obstacle.

S120: parameter adaptation module

The parameter adaptation module can generate the corresponding prediction time domain according to the vehicle state information and road curvature, including:

S121: Predicting time-domain neural network

The prediction time-domain neural network takes the current vehicle speed, expected vehicle speed and road curvature as input, and outputs the time-domain parameters of turning angle prediction and acceleration prediction time-domain parameters.

S122: Adapter

The adapter selects the corresponding rotation angle prediction time domain N _P1 and acceleration prediction time domain N _P2 within the range according to the size of the parameters P ₁ and P ₂ output by the prediction time domain neural network.

S130: System model library

The system model library can update the output parameter Y ₁ of the steering angle prediction model and the output parameter Y ₂ of the acceleration prediction model in real time according to the state information of the vehicle and the prediction time domain _NP1 and _NP2 , including:

S131: Acceleration prediction model

The acceleration prediction model uses the acceleration prediction time domain N _P2 to calculate the output parameter Y ₂ of the acceleration prediction model according to the state information of the vehicle.

S132: Corner prediction model

The steering angle prediction model calculates the output parameter Y ₁ of the steering angle prediction model by using the steering angle prediction time domain N _P1 according to the state information of the vehicle.

S140: Optimization solver

The optimization solver can solve the steering angle control variable u 2 and the acceleration control variable _{u 3 according} to the expected trajectory, expected vehicle speed, model output parameters Y ₁ and Y ₂ , and prediction time domain N _P1 and N _P2 , including:

S141: Acceleration optimization function

The acceleration optimization function can obtain the current acceleration control value u ₃ from the expected vehicle speed, the output parameter Y ₂ of the acceleration prediction model and the prediction time domain N _P2 .

S142: Corner optimization function

The corner optimization function can obtain the current corner control quantity u ₂ from the expected trajectory, the output parameter Y ₁ of the corner prediction model and the prediction time domain N _P1 .

S150: Emergency brake module

Emergency braking consists of an acceleration neural network and siren, which includes:

S151: Acceleration Neural Network

The acceleration neural network can generate the corresponding emergency braking acceleration control value u ₁ according to the current vehicle speed and the distance of the obstacle, and control the vehicle to perform emergency braking.

S152: Siren

The siren can send out sound and light alarms to prompt the occupants in the car according to the current vehicle speed and the distance between the current vehicle and the obstacle ahead.

S160: System Control Module

The system control module consists of logic converters and instruction generators, including:

S161: Logic Converter

The logic converter can convert the acceleration control quantity into the accelerator/brake control quantity;

S162: Instruction generator

The instruction generator can generate corresponding control instructions for the steering angle control amount, accelerator control amount, and brake control amount and send them to the current self-driving vehicle.

The specific working process of the automatic driving vehicle trajectory tracking control system disclosed in the embodiments of the present invention is shown in Figure 2. The specific implementation of the system will be described below in conjunction with the accompanying drawings:

The first is to process the signal. When the expected trajectory and expected vehicle speed are known, the signal processing subsystem simultaneously performs vehicle state estimation and obstacle detection to obtain vehicle state information, road curvature and obstacle information, including:

1) Use the state estimation module to estimate the state information of the current driving vehicle

The relevant measurement data is acquired by the on-board sensor components of the autonomous vehicle, and the data is input to the state estimation module. The state estimation module can perform estimation operations based on these measurement data, so as to obtain the current state information of the autonomous vehicle and the road information of the environment, and pass this information to the parameter adaptation module and the system model library.

2) Use the object detection module to detect obstacle information on the road ahead

The object detection module includes a binocular camera outside the vehicle and a laser radar, which can detect the driving environment in front of the vehicle and obtain the road curvature and obstacle information of the road ahead, including the distance to the obstacle.

The binocular camera outside the vehicle is used to collect the first image data and the second image data in the driving environment of the vehicle, wherein the first image data is the road curvature of the road ahead of the vehicle; the second image data is the road environment data in the direction of the vehicle. It is used to judge whether there is an obstacle ahead of the road.

LiDAR collects road data in the vehicle driving environment to determine whether there is an obstacle in front of the road, and to measure the distance a to the obstacle.

The object detection module combines the second image data of the binocular camera and the road data of the lidar to generate obstacle information.

When the system judges that there is an obstacle ahead, it will activate the emergency braking module, issue an alarm and control the vehicle to perform emergency braking to avoid obstacles. The emergency braking module includes the siren and acceleration neural network, which includes:

1) Acceleration generation

The system uses the trained acceleration neural network to obtain the corresponding emergency braking acceleration control value u ₁ , and then sends u ₁ to the system control module.

The input layer of the acceleration neural network has two nodes, which are the current vehicle speed and the distance to the obstacle; the output layer has one node, which is the acceleration control value u ₁ . Taking the current vehicle speed and the distance to the obstacle as input, the corresponding output value is obtained, and after denormalization processing, the acceleration control value u ₁ can be obtained.

After the logic converter in the system control module receives the acceleration control value u ₁ , it starts the working mode 1, and then generates the corresponding control command through the command generator, which can control the vehicle to perform emergency braking according to the braking acceleration u ₁ . Since u ₁ is generated online in real time by neural network, different u ₁ can be produced according to different vehicle speeds and distances to obstacles, which can adapt to different working conditions, and can reduce emergency braking while ensuring the driving safety of drivers and passengers. It can reduce the frustration during driving and improve the ride comfort of the vehicle.

2) siren

The siren includes an alarm circuit composed of a voice announcer and an indicator light, which can be used to alarm according to different situations. The voice announcer broadcasts the distance to obstacles at a certain frequency.

When the distance a to the obstacle is greater than the defined safety distance a ₀ , the voice announcer will broadcast "There are obstacles ahead, please avoid them"; the orange warning light is flashing, and the red warning light is off.

When the distance a to the obstacle is less than or equal to the defined safety distance a ₀ , the voice announcer broadcasts "Dangerous driving, there is a risk of collision ahead"; the orange warning light is off, and the red warning light is flashing.

When the system judges that there is no obstacle ahead, it starts the parameter adaptation module to generate the corresponding prediction time domain according to the current vehicle speed, expected vehicle speed and road curvature. The parameter adaptation module includes predictive temporal neural networks and adapters, including:

1) Prediction time domain parameter generation

The system uses the trained prediction time-domain neural network to obtain the corresponding rotation angle prediction time-domain parameters P ₁ and acceleration prediction time-domain parameters P ₂ , and then send P ₁ and P ₂ to the adapter.

The input layer of the prediction time domain neural network has three nodes, which are the current vehicle speed, expected vehicle speed and road curvature; the output layer has two nodes, which are the time domain parameter P ₁ of the corner prediction and the time domain parameter P ₂ of the acceleration prediction. Taking the current vehicle speed, expected vehicle speed and road curvature as input, the corresponding output value is obtained, and after denormalization processing, the corresponding rotation angle prediction time domain parameter P ₁ and acceleration prediction time domain parameter P in the range of 0-1 can be obtained ₂ .

2) Parameter adaptation

The adapter is generated within 0-1 with a spacing of i ₁ Intervals are combined into an interval group, where i ₁ is the interval parameter of the current interval. Each interval in the interval group has a corresponding forecast time domain. The adapter has two interval groups, which are the corner interval group and the acceleration interval group.

After the neural network outputs the parameters P ₁ and P ₂ , the adapter can find the corresponding rotation angle prediction time domain N _P1 and acceleration prediction time domain N _P2 in the respective interval groups according to the parameters P ₁ and P ₂ , so as to realize the prediction time domain Generated online to improve the accuracy of trajectory tracking control.

Then, the parameter adaptation module sends the obtained N _P1 and N _P2 to the system model library and the optimization solver.

The system model library includes a rotation angle prediction model and an acceleration prediction model, which can update the output parameter Y ₁ of the rotation angle prediction model and the output parameter Y ₂ of the acceleration prediction model according to the state information of the vehicle and the prediction time domain N _P1 and N _P2 , and Send the updated _Y1 and _Y2 to the optimization solver where:

1) Corner prediction model

Based on vehicle dynamics model and tire model, a linear time-varying predictive model based on dynamics is designed by using the principle of model predictive control. The specific principles are as follows:

The first is to model the vehicle dynamics. Due to the complexity of the vehicle system itself, it is difficult to establish an accurate model, and some reasonable assumptions need to be made before modeling. After assumption, Fig. 3 establishes the vehicle corner control model in conjunction with the tire model, and the specific implementation of the corner prediction model is explained below in conjunction with the accompanying drawings:

In this model, the state quantity is The angle control quantity is u ₂ ＝δ _f ; the output quantity is

Among them, m is the mass of the vehicle; a and b are the distances from the center of mass to the front and rear axles; is the center of mass yaw angle; /> is the center-of-mass yaw rate; /> is the yaw angular acceleration of the center of mass; /> and /> are the longitudinal and lateral speeds of the vehicle, respectively; /> and /> are the longitudinal acceleration and the lateral acceleration; I _z is the moment of inertia of the vehicle around the z axis; δ _f is the rotation angle of the front wheel; C _cf and C _cr are _the cornering stiffness of the front and rear _wheels respectively; is the longitudinal stiffness of the front and rear wheels; s _f and s _r are the slip rates of the front and rear wheels respectively; X and Y are the lateral and longitudinal displacements of the vehicle in the inertial coordinate system, respectively. m, a, b, I _z , C _cf , C _cr , C _lf , C _lr , s _f , and s _r are all known values.

Then use the principle of model predictive control to carry out linearization and discretization. Firstly, the Taylor formula is used for first-order expansion, and the above model can be linearized and simplified to obtain:

in,

in,

Then use the forward Euler method to discretize the model, and the discrete state space expression can be obtained:

ξ ₁ (k+1)=A ₁ (k)ξ ₁ (k)+B ₁ (k)u ₂ (k)

Wherein, A ₁ (k)=I+TA ₁ (t); B ₁ (k)=TB ₁ (t); k is the current sampling moment, k+1 is the next sampling moment; T is the sampling period.

Select the angle increment Δu ₂ as the control amount. After solving the control increment Δu ₂ (k) at the current moment, and adding the known control quantity at the previous moment, the control quantity u ₂ (k) at the current moment can be obtained. set up:

This leads to a new state-space expression:

ξ(k+1|t)＝A ₂ ξ(k|t)+B ₂ Δu ₂ (k|t)

Let the model output be:

η(k|t)=C ₁ ξ(k|t)

Set the prediction time domain of this model as N _P1 ; the control time domain is N _C1 , known, and N _C1 <N _P1 . Then the output at time N _P1 in the future can be obtained as:

The current state quantity can be measured by sensors or obtained by state estimation, so ξ(k|t) is known, and the control increment ΔU ₂ (t) in the control time domain can be obtained by calculation, so when predicting The output volume in the domain can be obtained.

Finally, according to the above model, according to the state information of the vehicle, the output parameter Y ₁ of the steering angle prediction model can be calculated by using the steering angle prediction time domain N _P1

2) Acceleration prediction model

The longitudinal dynamic model of the vehicle is analyzed, and the expected acceleration is obtained by using the model predictive control principle. Firstly, the first-order inertial system is used to express the longitudinal control of the vehicle, which can be obtained as follows:

Among them, K is the system gain; τ _d is the time constant; a is the current acceleration of the vehicle; a _des is the desired acceleration.

Transform the above model into a state-space expression:

in, The state quantity is x=[va] ^T ; the acceleration control quantity is u ₃ =a _des ; the speed v is the system output.

Then use the forward Euler method to discretize the model, and the discrete state space expression can be obtained:

x(k+1)＝A ₄ x(k)+B ₄ u ₃ (k)

in,

Then the model output is:

y(k|t)=C ₂ x(k|t)

Finally, according to the above model, according to the state information of the vehicle, the output parameter Y ₂ of the acceleration prediction model can be calculated by using the acceleration prediction time domain N _P2

The optimization solver includes an acceleration optimization function and a rotation angle optimization function, which can control the amount of rotation angle according to the expected trajectory, expected vehicle speed, model output parameters Y ₁ and Y ₂ of the system model library, and the predicted time domain N _P1 and N _P2 output by the adapter The solution of u ₂ and acceleration control variable u ₃ , and send u ₂ and u ₃ to the system control module, including:

1) Solve the angle control quantity

The optimization solver can solve the optimal control angle under the constraints by using the angle prediction time domain _NP1 according to the expected trajectory and the angle optimization function.

According to the principle of model predictive control, the corner optimization function can be obtained as:

Matrix Q ₁ is the weight matrix for tracking deviation; matrix R ₁ is the weight matrix for controlling the increment.

Referring to the expected trajectory, according to the output parameter Y ₁ of the rotation angle prediction model and the prediction time domain N _P1 , a series of optimal rotation angle increments ΔU ₂ (t) under the system constraints can be solved, and the first rotation angle of the series is taken Increment Δu ₂ (k|t), plus the angle control amount at the previous moment, can get the current angle control amount u ₂ .

2) Solution of acceleration control quantity

The optimization solver can use the acceleration prediction time domain N _P2 to solve the optimal control acceleration under the constraints according to the expected vehicle speed and acceleration optimization function.

According to the principle of model predictive control, the acceleration optimization function can be obtained as

Referring to the expected vehicle speed, according to the output parameter Y ₂ of the acceleration prediction model and the prediction time domain N _P2 , a series of optimal acceleration increments ΔU ₃ (t) under system constraints can be solved, and the first rotation angle of the series is taken Increment Δu ₃ (k|t), plus the acceleration control amount at the previous moment, can get the current acceleration control amount u ₃ .

After the system control module receives the angle control quantity u ₂ and the acceleration control quantity u ₃ , it cannot be directly used for vehicle control, and the acceleration control quantity u ₃ needs to be converted into the accelerator/brake control quantity before the automatic driving vehicle can be controlled. control.

The system control module contains a logic converter, which can convert the acceleration signal into an accelerator/brake signal, and then generate a corresponding control instruction through the instruction generator through the rotation angle signal and the accelerator/brake signal, so as to control the self-driving vehicle to follow the desired trajectory and desired Speed up or apply emergency brakes. The system control module includes two working modes. The specific principles are as follows:

1) Working mode 1

After the logic converter receives the emergency braking acceleration control value u ₁ , it starts working mode 1, and its working process is shown in Figure 4:

S410: firstly determine whether the driver performs a takeover operation after the siren alarms. When the driver takes over the operation, the logic converter has no output and does not perform any operation. When the driver does not perform the takeover operation, compare u ₁ with the maximum limit value r ₀ , and the obstacle distance a with the safety distance a ₀ ;

S420: When u ₁ <-r ₀ or a<a ₀ , it means that the emergency braking acceleration control value u ₁ has exceeded the maximum limit value, or the vehicle is too close to the obstacle, then brake according to the maximum limit value r ₀ Operation, output braking control quantity k ₁ r ₀ ;

S430: Otherwise, perform the braking operation according to the acceleration control value u ₁ , and output the braking control value k ₁ u ₁ . k ₁ is the braking coefficient.

Working mode 1 can not only avoid dangerous situations caused by excessive braking acceleration, but also prevent the vehicle from being unable to stop due to insufficient braking force when the distance is too short.

2) Working mode 2

After the logic converter receives the acceleration control value u ₃ , it starts working mode 2, and its working process is shown in Figure 5. :

S510: Comparing the acceleration control amount u ₃ with the control adjustment coefficient r ₁ ;

S520: when u ₃ <-r ₁ , perform a braking operation and output the braking control quantity k ₁ u ₃ ;

S530: when u ₃ >r ₁ , perform driving operation, and output throttle control value k ₂ u ₃ ;

Where _k2 is the driving coefficient;

S540: When -r ₁ <<u ₃ <<r ₁ , the logic converter has no output and does not perform any operation.

On the one hand, working mode 2 can avoid frequent switching of the accelerator pedal/brake pedal as far as possible, which can not only improve the ride comfort, but also reduce the loss of parts; on the other hand, it can also avoid operating the accelerator pedal and the brake pedal at the same time , Improved driving safety.

After _{the instruction} generator receives the parameters output by _the emergency _{braking module} and the _optimization solver, it can generate the _{corresponding} The control command is sent to the self-driving vehicle.

After receiving the control signal, the self-driving vehicle performs corresponding deflection and acceleration/deceleration operations, so that the vehicle can travel according to the desired trajectory and speed, realize trajectory tracking control, or perform emergency braking to avoid danger. Then the relevant measurement data is obtained in real time through the on-board sensor components, and the data is input to the state estimation module. The cycle goes on and on, and the vertical and horizontal speed change control of the automatic driving vehicle is finally realized.

The disclosed embodiments of the present invention also disclose a track tracking control method for an automatic driving vehicle. The specific flow of the method is shown in FIG. 6 , and the specific implementation is as follows:

S610: Vehicle status acquisition and obstacle judgment

When the signal processing subsystem obtains the expected vehicle speed and expected trajectory of the current self-driving vehicle, the state estimation module estimates the state information of the vehicle based on the acquired measurement data, and detects the driving environment ahead to obtain the road ahead Curvature and obstacle information; if there is an obstacle ahead, start the emergency braking module, and if there is no obstacle ahead, start the parameter adaptation module;

S620: Emergency Brake Alert

Use the trained acceleration neural network to generate an output value in real time based on the current vehicle speed and the distance a between the vehicle and the obstacle, and then denormalize the output value to obtain the real-time control value u ₁ corresponding to the acceleration and send it to the system control module, the alarm is alarmed by judging the magnitude of the value between a and _a0 , where _a0 is the current safe distance between the vehicle and the obstacle;

S630: Generate prediction time domain

Use the trained prediction time-domain neural network to generate output values in real time based on the current vehicle speed, expected vehicle speed and curvature, and then denormalize the output values to obtain the corresponding rotation angle prediction time-domain parameters P ₁ and acceleration prediction time-domain parameters P ₂ and send it to the adapter, and the adapter selects the corresponding rotation angle prediction time domain N _P1 and acceleration prediction time domain N _P2 within the range based on the size of P ₁ and P ₂ and sends them to the system model library and optimization solver;

S640: Updating the prediction model parameters in real time

By the acceleration prediction model and the rotation angle prediction model, according to the status information, _NP1 and _NP2 , the output parameter _Y1 of the rotation angle prediction model and the output parameter _Y2 of the acceleration prediction model are updated in real time;

S650: Real-time optimization of vehicle trajectory

From the acceleration optimization function and the rotation angle optimization function, calculate according to the desired trajectory, desired vehicle speed, Y ₁ , Y ₂ , N _P1 , N _P2 to obtain the rotation angle control value u ₂ and the acceleration control value u ₃ , and send u ₂ and u ₃ to to the system control module;

S660: Convert control parameters into control vehicles and send them to self-driving vehicles

After receiving u ₁ , u ₂ and u _{3 ,} the logic converter generates corresponding control commands through the command generator to control the current vehicle to perform corresponding deflection and acceleration and deceleration operations according to u ₁ , u ₂ and u ₃ .

The specific embodiments described above have further described the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above descriptions are only specific embodiments of the present invention and are not intended to limit the scope of the present invention. Protection scope, any modification, equivalent replacement, improvement, etc. made on the basis of the technical solution of the present invention shall be included in the protection scope of the present invention.

Claims (9)

1. An automatic driving vehicle trajectory tracking control system, characterized in that the system comprises: a signal processing subsystem for determining the curvature of the road ahead and state information of the vehicle; A parameter adaptation module, the parameter adaptation module includes a prediction time-domain neural network and an adapter, and the prediction time-domain neural network is used for processing according to the current vehicle speed, the expected vehicle speed and the curvature of the road ahead to obtain the corner prediction time-domain parameters and acceleration Prediction time domain parameters; the adapter is used to select the corresponding rotation angle prediction time domain and acceleration prediction time domain according to the size of the rotation angle prediction time domain parameters and acceleration prediction time domain parameters in the custom corner interval group and acceleration interval group; A system model library, the system model library is used to obtain the output parameters of the rotation angle prediction model and the output parameters of the acceleration prediction model according to the state information of the vehicle, the rotation angle prediction time domain and the acceleration prediction time domain; An optimization solver, the optimization solver is used to obtain the rotation angle control amount and the acceleration control amount according to the expected trajectory, the desired vehicle speed, the output parameters of the rotation angle prediction model and the rotation angle prediction time domain, the acceleration prediction model output parameters and the acceleration prediction time domain; A system control module, the system control module is used to receive the control amount of rotation angle and the acceleration control amount and generate corresponding control commands to control the vehicle to perform corresponding deflection and acceleration and deceleration operations. 2. The system according to claim 1, wherein the system further comprises an emergency braking module, comprising: Acceleration neural network, the acceleration neural network is used to generate the corresponding emergency braking acceleration control amount according to the current vehicle speed and the distance between the vehicle and the obstacle obtained by the object detection module in the signal processing subsystem and send it to the system control module; then the system control module generates a corresponding control instruction after receiving the emergency braking acceleration control amount, and controls the emergency braking of the vehicle; an alarm, and the alarm performs an alarm by comparing the distance from the vehicle to the obstacle and the preset safety distance. 3. The system according to claim 1, wherein the system model storehouse includes a corner prediction model, and the mathematical expression of the corner prediction model is: In the expression, the state quantity is The angle control quantity is u ₂ ＝δ _f ; the output quantity is Among them, m is the vehicle mass of the vehicle; a and b are the distances from the center of mass of the vehicle to the front and rear axles respectively; φ is the yaw angle of the center of mass; ^{φ is} the yaw rate of the center of mass; Swing angular acceleration; x ^{and y} are the vehicle longitudinal velocity and lateral velocity respectively; x ^{and y} are longitudinal acceleration and lateral acceleration respectively; _Iz is the moment of inertia of the vehicle around the z axis; δ _f is the rotation angle of the front wheel; C _cf and C _cr are the cornering stiffness of the front and rear wheels respectively; C _1f and C _1r are the longitudinal stiffness of the front and rear wheels respectively; s _f and s _r are the front and rear wheels respectively slip rate; X and Y are the lateral and longitudinal displacements of the vehicle in the inertial coordinate system; among them, m, a, b, I _z , C _cf , C _cr , C _1f , C _1r , s _f , s _r is a known value. 4. The system according to claim 3, characterized in that, the mathematical expression of the angle prediction model uses the model predictive control principle to perform linearization and discretization to obtain a discrete state space expression: ξ ₁ (k+1)=A ₁ (k)ξ ₁ (k)+B ₁ (k)u ₂ (k) Wherein, A ₁ (k)=I+TA ₁ (t); B ₁ (k)=TB ₁ (t); k is the current sampling moment, k+1 is the next sampling moment; T is the sampling period; Select the rotation angle increment Δu ₂ as the control quantity, solve the control increment Δu ₂ (k) at the current moment, and add the control quantity known at the previous moment to obtain the control quantity u ₂ (k) at the current moment; set Certainly: This leads to a new state-space expression: ξ(k+1|t)＝A ₂ ξ(k|t)+B ₂ Δu ₂ (k|t) Let the model output be: η(k|t)=C ₁ ξ(k|t) Set the prediction time domain of this model as the rotation angle prediction time domain N _P1 ; the control time domain is N _C1 , which is known, and N _C1 < N _P1 ; then the output at the future N _P1 moment is: From the model and according to the state information of the vehicle, the time domain N _P1 is predicted by using the corner angle, using the formula: After solving, the output parameter Y ₁ of the rotation angle prediction model of the rotation angle prediction model is obtained. 5. The system according to claim 1, wherein the system model library includes an acceleration prediction model, and the mathematical expression of the acceleration prediction model is: Among them, K is the system gain; τ _d is the time constant; a is the current acceleration of the vehicle; a _des is the desired acceleration; The mathematical expression converted to a spatial expression is: in, The state quantity is x=[va] ^T ; the acceleration control quantity is u ₃ ＝a _des ; the speed v is the system output; The model is discretized using the forward Euler method to obtain a discrete state space expression: x(k+1)＝A ₄ x(k)+B ₄ u ₃ (k) in, Then the model output is: y(k|t)=C ₂ x(k|t) Analyze the longitudinal dynamics model of the vehicle, and use the acceleration prediction time domain N _P2 according to the state information of the vehicle, using the formula: After solving, the output parameter Y ₂ of the acceleration prediction model is obtained. 6. The system according to claim 1, characterized in that, the control angle optimization function is: Among them, matrix Q ₁ is the weight matrix of tracking deviation, and matrix R ₁ is the weight matrix of control increment; From the expected trajectory, the output parameter Y ₁ of the rotation angle prediction model and the rotation angle prediction time domain N _P1 , a series of optimal rotation angle increments ΔU ₂ (t) under the system constraints are solved, and the first rotation angle increment of the series is taken Δu ₂ (k|t) is added to the angle control amount at the previous moment to obtain the current angle control amount u ₂ . 7. The system according to claim 1, wherein the optimization function of the acceleration control amount is: A series of optimal acceleration increments ΔU ₃ (t) under system constraints are obtained from the desired vehicle speed, the output parameter Y ₂ of the acceleration prediction model and the acceleration prediction time domain N _P2 , and the first rotation angle increment Δu of the series is taken ₃ (k|t), plus the acceleration control amount at the previous moment, to obtain the current acceleration control amount u ₃ . 8. The system according to claim 1, wherein the system control module comprises a logic converter, and the logic converter has two operating modes: After the logic converter receives the emergency braking acceleration control amount, it turns on mode 1. According to whether there is a driver to take over after the siren alarm, if there is a driver to take over, the logic converter has no output and does not perform any operation; if there is no driver to take over, execute: When u ₁ <-r ₀ or a<a ₀ , perform braking operation according to r ₀ and output braking control quantity k ₁ r ₀ ; Otherwise, the braking operation is performed according to the emergency braking acceleration control quantity u1, and the braking control quantity k ₁ u ₁ is output; wherein, r ₀ is the maximum limit value of the vehicle acceleration, and k ₁ is the braking coefficient; After receiving the acceleration control value u ₃ , the logic converter turns on mode 2. When u ₃ <-r ₁ , the brake operation is performed and the brake control value k ₂ u ₃ is output; when -r ₁ <<u ₃ << When r _{is 1} , the logic converter has no output and does not perform any operation; When u ₃ >r ₁ , the driving operation is performed, and the throttle control value k ₂ u ₃ is output; where, r ₁ is the control adjustment coefficient, and k ₂ is the driving coefficient. 9. An automatic driving vehicle trajectory tracking control method, characterized in that the method comprises: Detecting the road environment in front of the vehicle according to the object detection module in the signal processing subsystem, and judging whether there is an obstacle in the road ahead; Generate corresponding corner prediction time domain parameters and acceleration prediction time domain parameters based on the current vehicle speed, expected vehicle speed and curvature of the road ahead via the prediction time domain neural network in the parameter adaptation module; Then the adapter selects the corresponding rotation angle prediction time domain and acceleration prediction time domain within the range based on the size of the rotation angle prediction time domain parameter and the acceleration prediction time domain parameter; The steering angle prediction model in the system model library obtains the output parameters of the steering angle prediction model according to the vehicle state information obtained by the state estimation module and the steering angle prediction time domain; The acceleration prediction model in the system model library obtains the output parameters of the acceleration prediction model according to the vehicle state information and the acceleration prediction time domain; The corner optimization function in the optimization solver is based on the expected trajectory, according to the output parameters of the corner prediction model and the corner control amount obtained in the time domain of the corner prediction; The acceleration optimization function in the optimization solver is based on the expected vehicle speed, and the acceleration control amount is obtained according to the output parameters of the speed prediction model and the acceleration prediction time domain; After the logic converter in the system control module receives the angle control amount and the acceleration control amount, the instruction generator in the system control module generates corresponding control instructions to control the vehicle to perform corresponding deflection and acceleration and deceleration operations.