CN116061945A - Self-adaptive cruise control method for commercial vehicle based on lane keeping - Google Patents

Self-adaptive cruise control method for commercial vehicle based on lane keeping Download PDF

Info

Publication number
CN116061945A
CN116061945A CN202310120341.1A CN202310120341A CN116061945A CN 116061945 A CN116061945 A CN 116061945A CN 202310120341 A CN202310120341 A CN 202310120341A CN 116061945 A CN116061945 A CN 116061945A
Authority
CN
China
Prior art keywords
commercial vehicle
vehicle
adaptive cruise
road
control method
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310120341.1A
Other languages
Chinese (zh)
Inventor
于树友
刘郡亭
常欢
李文博
陈虹
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Jilin University
Original Assignee
Jilin University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Jilin University filed Critical Jilin University
Priority to CN202310120341.1A priority Critical patent/CN116061945A/en
Publication of CN116061945A publication Critical patent/CN116061945A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W40/00—Estimation or calculation of non-directly measurable driving parameters for road vehicle drive control systems not related to the control of a particular sub unit, e.g. by using mathematical models
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units, or advanced driver assistance systems for ensuring comfort, stability and safety or drive control systems for propelling or retarding the vehicle
    • B60W30/10—Path keeping
    • B60W30/12—Lane keeping
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60W—CONJOINT CONTROL OF VEHICLE SUB-UNITS OF DIFFERENT TYPE OR DIFFERENT FUNCTION; CONTROL SYSTEMS SPECIALLY ADAPTED FOR HYBRID VEHICLES; ROAD VEHICLE DRIVE CONTROL SYSTEMS FOR PURPOSES NOT RELATED TO THE CONTROL OF A PARTICULAR SUB-UNIT
    • B60W30/00—Purposes of road vehicle drive control systems not related to the control of a particular sub-unit, e.g. of systems using conjoint control of vehicle sub-units, or advanced driver assistance systems for ensuring comfort, stability and safety or drive control systems for propelling or retarding the vehicle
    • B60W30/14—Adaptive cruise control
    • B60W30/16—Control of distance between vehicles, e.g. keeping a distance to preceding vehicle
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
    • Y02T10/00—Road transport of goods or passengers
    • Y02T10/10—Internal combustion engine [ICE] based vehicles
    • Y02T10/40—Engine management systems

Abstract

The invention belongs to the technical field of automatic control of vehicles, and discloses a self-adaptive cruise control method of a commercial vehicle based on lane keeping, which comprises the following steps: based on a Koopman operator theory, constructing a commercial vehicle linearization model based on a neural network; collecting running information and front road information of a commercial vehicle, and constructing a vehicle-road relative position model; acquiring state quantity related to turning of a curve of the commercial vehicle through the vehicle-road relative position model; and designing a predictive controller based on the linearization model of the commercial vehicle, inputting the state quantity into the predictive controller, calculating to obtain a control quantity related to the turning of the curve of the commercial vehicle, and acting the control quantity on a control system of the commercial vehicle. In conclusion, the method uses the Koopman operator theory and the neural network as the basis to fit and construct the linearization model of the commercial vehicle, and combines the linearization model of the commercial vehicle with the vehicle-road relative position model, so that the commercial vehicle has better track tracking performance when steering and driving.

Description

Self-adaptive cruise control method for commercial vehicle based on lane keeping
Technical Field
The invention belongs to the technical field of automatic control of vehicles, and particularly relates to a self-adaptive cruise control method of a commercial vehicle based on lane keeping.
Background
An adaptive cruise control system (ACC) is a vehicle safety-assisted driving system developed at the end of the 70 s of the 20 th century. The self-adaptive cruise control system is added with a safe vehicle distance maintaining system based on the traditional constant-speed cruise control, measures the front driving environment, such as the front vehicle distance, the relative speed with the front vehicle and the like by utilizing the vehicle-mounted radar sensor, and then actively adjusts the vehicle speed based on the throttle valve and the braking device, thereby ensuring the safe running of the vehicle.
At present, with the development of an adaptive cruise control system (ACC), students at home and abroad change a research target from a straight driving condition to a curve driving condition. Compared with other vehicles, the commercial vehicle has the characteristics of higher mass center of the whole vehicle, long longitudinal dimension and the like when running in full load, and rollover accidents are easy to occur, so that the research on the lateral stability of the full-load commercial vehicle has important significance for ensuring that the commercial vehicle can safely steer at high speed.
In addition, the existing research shows that the prediction control can realize control targets such as minimum safety distance to a front vehicle in the adaptive cruise control system (ACC) through explicit processing constraint, and the non-convex optimization problem needs to be solved on line by a correspondingly designed non-linear model prediction controller because the commercial vehicle has strong non-linear characteristics, so that the problem that the calculation burden is heavy and the real-time requirement in the actual running control of the commercial vehicle is difficult to meet is solved.
Disclosure of Invention
In view of the above, it is an object of the present invention to provide a method for adaptive cruise control of a commercial vehicle based on lane keeping, in order to solve the above-mentioned problems.
In order to achieve the above purpose, the present invention provides the following technical solutions:
a lane keeping based adaptive cruise control method for a commercial vehicle, comprising:
based on a Koopman operator theory, constructing a commercial vehicle linearization model based on a neural network;
collecting running information and front road information of a commercial vehicle, and constructing a vehicle-road relative position model;
acquiring state quantity related to turning of a curve of the commercial vehicle through the vehicle-road relative position model;
and designing a predictive controller based on the linearization model of the commercial vehicle, inputting the state quantity into the predictive controller, calculating to obtain a control quantity related to the turning of the curve of the commercial vehicle, and acting the control quantity on a control system of the commercial vehicle.
Preferably, the state quantity includes a lateral error between a current running position of the commercial vehicle and a target running position, and an angle error between a current running direction of the commercial vehicle and a tangential direction of the target running position.
Preferably, the state quantity further includes a distance error between an actual distance between the commercial vehicle and the pilot vehicle and a desired distance between the commercial vehicle and the pilot vehicle.
Preferably, U (k) = { U (k|k), U (k+ 1|k) … U (k+n) p -1|k) is a control sequence of the predictive controller, and the predictive expression of the predictive controller is:
Figure BDA0004079745040000021
/>
Figure BDA0004079745040000022
Figure BDA0004079745040000023
wherein X is min ,X max To limit the limit value of the constraint range of the state quantity, u min ,u max For the limit value of the constraint range of the control quantity, Q, R is a positive weighting coefficient, Y (k+i|k) is the prediction horizon [0N ] p -1]The prediction output of the intra-frame is,
Figure BDA0004079745040000024
to predict the penalty term between output and expected output, ζ (k) = [ v ] ref (k)ω ref (k)] T For measurable disturbance, ++>
Figure BDA0004079745040000025
v x (k) Is a constant value;
Figure BDA0004079745040000031
preferably, the control amount is a first control element in the optimal control sequence predicted and calculated by the prediction controller.
Preferably, the control amount includes a longitudinal force and a front wheel turning angle when the commercial vehicle turns.
Preferably, the constraint of the state quantity includes a constraint of a lateral speed, a yaw rate, and a longitudinal speed; the control amount constraint includes a longitudinal force and a front wheel rotation angle constraint.
Preferably, the Koopman operator theory is to complete the evolution of the nonlinear system in a linear manner in a state space of infinity, and the evolution expression is as follows:
Figure BDA0004079745040000032
wherein u is k For controlling quantity, ++>
Figure BDA0004079745040000033
Extracting a matrix for a state->
Figure BDA0004079745040000038
The method is used for estimating the state quantity of the original nonlinear system, and A and B are parameter matrixes.
Preferably, the expression of the linearization model of the commercial vehicle is:
Figure BDA0004079745040000034
wherein u is k For controlling quantity, ++>
Figure BDA0004079745040000035
Extracting a matrix for a state->
Figure BDA0004079745040000036
For the estimated value of the state quantity of the original nonlinear system at the moment k, A DNN ,B DNN Is an optimal parameter matrix obtained through neural network training.
Preferably, the expression of the vehicle-road relative position model is:
Figure BDA0004079745040000037
in the formula, v ref /v x Longitudinal speeds e of pilot and commercial vehicles, respectively x E is the distance error between the actual distance between the commercial vehicle and the pilot vehicle and the expected distance between the commercial vehicle and the pilot vehicle y For the current driving position and the front D of the commercial vehicle L A lateral error in the road center position e a For the current running direction and the front D of the commercial vehicle L Angle error in tangential direction of road center, D L To detect distance, K L =1/d is the road curvature of the vehicle centerline, d is the road radius.
Compared with the prior art, the invention has the following beneficial effects:
according to the lane-keeping-based commercial vehicle self-adaptive cruise control method, a commercial vehicle linearization model is constructed by combining the Koopman operator theory and the neural network fitting, so that the nonlinear self-adaptive cruise control problem is described as a constrained optimization problem, and the model has a good fitting effect on the dynamics of the transverse speed, the longitudinal speed, the yaw rate and the like of the vehicle, so that the accuracy of predictive control is effectively ensured.
According to the lane-keeping-based self-adaptive cruise control method for the commercial vehicle, the linearization model of the commercial vehicle is combined with the vehicle-road relative position model, so that the commercial vehicle has better track tracking performance during steering running, the solving time is greatly shortened, meanwhile, the stability during steering is ensured by utilizing constraint, and the running safety of the commercial vehicle and the real-time requirement in practical application are further effectively met.
Drawings
FIG. 1 is a functional block diagram of a lane keeping based adaptive cruise control method for a commercial vehicle of the present invention;
FIG. 2 is a block diagram of a neural network;
FIG. 3 is a graph of the geometry of a vehicle/road in a vehicle-road relative position model;
FIG. 4 is a schematic illustration of relative vehicle position;
fig. 5-7 are state quantity comparison diagrams of a commercial vehicle linearization model and a TruckSim simulation vehicle.
Fig. 8-15 are graphs comparing the simulation results of the standard highway driving conditions.
Fig. 16 to 24 are graphs showing comparison between simulation results of the driving-limit conditions.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. 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.
As shown in a schematic diagram in fig. 1, the adaptive cruise control method for a commercial vehicle based on lane keeping provided by the invention specifically includes:
s1, constructing a commercial vehicle linearization model based on a neural network based on a Koopman operator theory.
The core idea of Koopman operator theory is to accomplish the evolution of nonlinear systems in a linear fashion in an infinite-dimensional state space.
Classical Koopman operator theory applies to autonomous systems, assuming an n-dimensional nonlinear discrete autonomous system: x is x k+1 ＝h(x k ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is k ∈R n The state quantity of the nonlinear system at the moment k; h is the mapping relation from the nonlinear system to the linear evolution.
The operator K of the Koopman operator theory is defined as a linear operator acting on a lifting function g (x), and the evolution mode based on the Koopman operator theory is that a nonlinear discrete autonomous system is lifted to an infinite dimensional space through the lifting function g (x), so that a linear system is obtained:
Figure BDA0004079745040000051
because the infinite dimensional state space is difficult to realize in practical application and the actual vehicle system is a nonlinear controlled system, a finite dimensional space is selected to approximate an infinite dimensional linear space, and classical Koopman operator theory is applied to a commercial vehicle system.
Based on the theory:
the commercial vehicle discrete nonlinear controlled system is as follows: x is x k+1 ＝L(x k ,u k ) The method comprises the steps of carrying out a first treatment on the surface of the Wherein x is k ∈R n For the nonlinear controlled system state quantity at k moment, u k ∈R n And L is the mapping relation from the nonlinear system to the linear evolution, wherein the control quantity is the nonlinear controlled system control quantity at the moment k.
Let the lifting function be
Figure BDA0004079745040000052
In (1) the->
Figure BDA0004079745040000053
Is a high-dimensional state space dimension.
By finding a linear matrix
Figure BDA0004079745040000054
Make->
Figure BDA0004079745040000055
Thereby enabling the commercial vehicle discrete nonlinear controlled system to be lifted into a limited high-dimensional state space.
Adding original information into a high-dimensional state space to construct a new space state quantity
Figure BDA0004079745040000056
And is also provided with
Figure BDA0004079745040000061
Assuming that the control amount in the high-dimensional state space is the same as the control amount in the original state space, the evolution of the commercial vehicle discrete nonlinear controlled system can be expressed as:
Figure BDA0004079745040000062
wherein the control amount u k The evolution of (c) is given externally without prediction, whereby an evolution expression based on Koopman operator theory is obtained as: />
Figure BDA0004079745040000063
Wherein u is k For controlling quantity, ++>
Figure BDA0004079745040000064
Extracting a matrix for a state->
Figure BDA0004079745040000065
The method is used for estimating the state quantity of the original nonlinear system, and A and B are parameter matrixes.
From the above, the state quantity at time k of the nonlinear vehicle system is defined as x V (k)＝[v x (k)v y (k)ω(k)] T And based on the neural network structure shown in FIG. 2 to
Figure BDA0004079745040000066
As an output of the neural network. Since it is assumed that the control amount in the high-dimensional state space is the same as that in the original state space, the up-scaling process is not performed on u (k).
Constructing a spatial state quantity by combining the neural network
Figure BDA0004079745040000067
And optimizing and training the parameter matrixes A and B through a neural network so as to obtain an optimal parameter matrix A DNN ,B DNN 。
To more accurately approximate the vehicle dynamics system, a minimization of multi-step prediction error is performed, where the p-step state prediction is:
Figure BDA0004079745040000068
to verify neural network fitting ability, a linear error L is defined xp The average of the sum of p-step prediction errors:
Figure BDA0004079745040000069
in (1) the->
Figure BDA00040797450400000610
Figure BDA00040797450400000611
The number of the lifting functions.
To verify the neural network fitting ability, define the linearity error in the high-dimensional state space:
Figure BDA0004079745040000071
/>
to sum up, a loss function is obtained for the neural network:
Figure BDA0004079745040000072
in (1) the->
Figure BDA0004079745040000073
Regularization term added to avoid neural network overfitting, alpha 1 、α 2 And alpha 3 The weight coefficients of the corresponding terms are respectively obtained, and the linearization model of the commercial vehicle is obtained based on the weight coefficients:
Figure BDA0004079745040000074
wherein u is k For controlling quantity, ++>
Figure BDA0004079745040000075
Extracting a matrix for a state->
Figure BDA0004079745040000076
For the estimated value of the state quantity of the original nonlinear system at the moment k, A DNN ,B DNN Is an optimal parameter matrix obtained through neural network training.
S2, acquiring running information and front road information of the commercial vehicle through a sensor, and constructing a vehicle-road relative position model.
S3, acquiring state quantity related to turning of the curve of the commercial vehicle through a vehicle-road relative position model.
Specifically, the state quantity includes a lateral error between the current running position of the commercial vehicle and the target running position, and an angle error between the current running direction of the commercial vehicle and the tangential direction of the target running position. FIG. 3 is a graph of vehicle/road geometry in a vehicle-road relative position model, wherein: d (D) L To detect distance e y For the current driving position and the front D of the commercial vehicle L A lateral error in the road center position e a For the current running direction and the front D of the commercial vehicle L Angle error in tangential direction of road center, K L =1/d is the road curvature of the vehicle centerline, d is the road radius, based on which the relative vehicle-road position model can be expressed as:
Figure BDA0004079745040000077
in addition, the state quantity also comprises a distance error between the actual distance between the commercial vehicle and the pilot vehicle and the expected distance between the commercial vehicle and the pilot vehicle. Fig. 4 shows a relative position diagram of a commercial vehicle and a pilot vehicle, regardless of the length of the vehicle body, ensuring that the distance between the two vehicles is not less than the minimum safe vehicle distance: e, e x ＝x f -x h -d ref The method comprises the steps of carrying out a first treatment on the surface of the In the formula e x For the actual distance between two vehicles and the expected distance d ref Spacing error, x f ,x h The current longitudinal positions of the commercial vehicle and the pilot vehicle respectively. Further, obtain
Figure BDA0004079745040000081
In the formula, v ref /v x Longitudinal speeds of the pilot vehicle and the commercial vehicle respectively.
S4, designing a predictive controller based on a linearization model of the commercial vehicle, inputting the state quantity into the predictive controller, calculating to obtain the control quantity related to the turning of the curve of the commercial vehicle, and acting the control quantity on a control system of the commercial vehicle.
Combining the linear model of the commercial vehicle and the relative position model of the vehicle and the road, wherein the linear model of the commercial vehicle is expressed as X (k) = [ z ] V (k) T e x (k)e y (k)e a (k)] T To expand the state quantity, u k ＝[F x (k),δ(k)] T For the control amount, ζ (k) = [ v ref (k)ω ref (k)] T For a measurable disturbance, a linear kinetic equation is obtained for the steering of the commercial vehicle:
Figure BDA0004079745040000082
in the method, in the process of the invention,
Figure BDA0004079745040000083
v x (k) Is a constant value; />
Figure BDA0004079745040000084
Figure BDA0004079745040000089
/>
Definition U (k) = { U (k|k), U (k+ 1|k) … U (k+n) p -1|k) is the control sequence of the predictive controller, in the predictive time domain [0N ] p -1]Taking Y (k+i|k) as a prediction output, obtaining the following prediction controllers and optimizing and solving:
Figure BDA0004079745040000087
Figure BDA0004079745040000088
Figure BDA0004079745040000091
wherein X is min ,X max To limit the limit value of the constraint range of the state quantity, u min ,u max In order to limit the limit value of the constraint range of the control quantity, Q, R is a positive weight coefficient,
Figure BDA0004079745040000092
a penalty term between the predicted output and the desired output.
To sum up
Regarding the built linearization model of the commercial vehicle, verification was performed by TruckSim:
to identify the dynamics of the commercial vehicle, a total of 70 sets of data were collected. 60 sets of data are selected as a neural network training set, 5 sets of data are selected as a model verification set, 5 sets of data are selected as an experimental test set, and each set of data contains 6200 time step data. The neural network is selected as a network structure with the hidden layer number of 4, and 80 neurons are selected for each layer. Number of lifting functions N p ＝30。
The learning rate is an important super-parameter in deep learning, determining whether and when the objective function can converge to a local minimum. The proper learning rate can enable the objective function to converge to the local minimum value in proper time, and the learning rate is set to be 10 according to experiments -3 The weight is selected as alpha 1 ＝0.1,α 2 ＝1,α 3 ＝10 -6 。
And verifying a linear model of the commercial vehicle obtained by fitting under the combined simulation of TruckSim and Simulink. Defining average root mean square error as objective evaluation index of linear model prediction effect, and the expression is:
Figure BDA0004079745040000093
wherein x is true (k) And x pred (k) The actual value of the actual vehicle system state quantity at the k moment and the estimated value of the linear system state quantity are respectively.
The initial value of the state quantity is selected as x 0 ＝[25 0 0] T The control quantity is selected as F x As can be seen from the figures, the commercial vehicle linearization model using the neural network fitting has good fitting effects on the longitudinal speed, the lateral speed and the yaw rate. Meanwhile, the root mean square error of the linear model of the commercial vehicle, which is fitted by using the deep neural network, can be calculated to be 1.11%, namely the linear model of the commercial vehicle has higher fitting accuracy.
Regarding prediction by the predictive controller, by TruckSim simulation:
selecting a truck with a load of 5760kg, 12240kg and a total weight of 18000kg for commercial modeling, and selecting a longitudinal force F according to the control amount x And front wheel steering angle delta, vehicle state quantity selects longitudinal speed v x Transverse velocity v y Yaw rate ω. Sampling step length is T s =10 ms; in the process of collecting data, the sampling range of the longitudinal force is [ -6000,6000]N, the front wheel angle is [ -10,10]deg, corresponding v x ,v y Omega ranges from [16,25]m/s、[-0.6,0.6]m/s、[-6,6]deg/s. In addition, the weight matrix is:
Figure BDA0004079745040000101
Figure BDA0004079745040000102
(1) Standard highway driving condition simulation
The maximum road curvature of the standard expressway of China is 0.0025, the road width is 3.7m, the road adhesion coefficient is mu=0.85, the pilot vehicle passes through the curve at a constant speed of 90km/h, and the vehicle width is 2.5m.
Setting the pilot vehicle to always run along the central line of the road in TruckS im, namely, the running track of the pilot vehicle is the expected running track, the following vehicle (commercial vehicle) tracks the pilot vehicle, and the tracks of the two vehicles are shown in fig. 8-15 (the control comparison result of the predictive control method and the five-degree-of-freedom vehicle model in the figure is shown in the figure):
fig. 8 is a graph of longitudinal speed versus time for two control systems. As can be seen from the figure, both control systems can track the longitudinal speed of the pilot vehicle, and within the first 5 seconds, the error of the 5dof-ACC is large, and the maximum error is 0.18km/h.
FIG. 9 is a graph of the lateral speed of two control systems tracking the desired lateral speed 0 during straight-through travel and generating lateral speed during steering to ensure that the vehicle is properly traveling through a curve, the lateral speed is maintained at 0 again after reentering the straight-through, and the overshoot of the 5dof-ACC lateral speed is greater than Koopman-ACC after entry into the curve.
FIG. 10 is a trace of yaw rate, showing that the yaw rate deviation and overshoot of 5dof-ACC is greater and oscillations are generated, compared to Koopman-ACC, indicating better stability of Koopman-ACC in steering.
FIG. 11 is a graph of longitudinal distance errors, each of which is no more than 0.05m; because various losses and errors exist in the vehicle, the errors are not considered in the five-degree-of-freedom model in modeling, so that the 5dof-ACC error is finally stabilized at 0.028m, the losses are included in the modeling sampling by using the Koopman operator theory, and the result also shows that the vehicle model based on the Koopman operator theory is closer to a TruckSim simulation vehicle.
Fig. 12 shows the lateral distance error, the maximum value of the lateral distance error of both systems is not more than 0.1m, and the lateral distance error of the two systems is not more than the boundary line of the road 0.5m from the center line of the road.
The lateral acceleration of both control systems is shown in FIG. 13, the maximum of which is not more than 0.2m/s 2 Far less than 4m/s of rollover threshold of full-load commercial vehicle 2 。
FIG. 14 is a diagram of a travel path showing that both control systems can track a desired path and that the Koopman-ACC is more closely adjacent to the pilot travel path. And also to indicate that the following vehicle (commercial vehicle) is in a safe and stable state when turning.
FIG. 15 shows that the solving time of each step of the controller is not more than 1.5ms, which is far less than the sampling time of the system and can meet the real-time requirement of the system application, and the 5dof-ACC adopts a nonlinear model predictive controller, the average solving time is 93.26ms, which is far higher than the sampling time of the system and the solving time of the Koopman-ACC.
In summary, the predictive control method has the advantages of smaller longitudinal and transverse errors and very short solving time, so that the running safety and real-time requirements in the practical system application can be ensured.
(2) Standard highway driving condition simulation:
a limit driving condition is designed for verifying the fitting degree of the Koopman linear model to the tire working in the nonlinear region. The road curvature information is: the maximum road curvature is 0.01, the road width is 3.7m, the road adhesion coefficient is mu=0.85, the pilot vehicle passes through the curve at a constant speed of 90km/h, and the vehicle width is 2.5m.
The track following results are shown in fig. 16-24 (which show the control comparison result of the predictive control method of the present invention and the five-degree-of-freedom vehicle model):
FIG. 16 is a graph of longitudinal speed versus time for two systems, with Koopman-ACC tracking significantly better than 5dof-ACC, which produces a large oscillation after entering a curve, and which adjusts to the desired speed at 17 s.
FIGS. 17 and 18 show the lateral and yaw rates of the two control systems, with 5dof-ACC producing greater lateral and yaw rates and oscillations all the way through, and Koopman-ACC having better lateral stability.
FIG. 19 shows that the absolute value of the maximum value of the longitudinal distance error is not more than 0.05m, and the Koopman-ACC has better tracking effect than the 5 dof-ACC.
FIG. 20 shows that the maximum value of the Koopman-ACC is not more than 0.4m and not more than the road boundary line 0.5m from the center line of the road, but the maximum value of the 5dof-ACC is 0.78m when passing through the curve and the road boundary line is already exceeded.
Fig. 21 shows the tire slip angle of the front wheel of the following vehicle (commercial vehicle), and it can be seen that the tire has been operated in the nonlinear region within 5.2-6.6s of the front wheel slip angle of the following vehicle exceeding 5 °.
The lateral acceleration of both control systems is shown in FIG. 22, the maximum of which is not more than 0.8m/s 2 Far less than 4m/s of rollover threshold of full-load commercial vehicle 2 However, the lateral acceleration of the 5dof-ACC is always in an oscillating state after the curve is ended and enters the straight road.
FIG. 23 is a diagram of the travel path of two systems, it being seen that the Koopman-ACC can track the desired path; although the 5dof-ACC may also track the desired trajectory, it is illustrated by fig. 17 and 18 that the lateral stability of the vehicle is worse. FIG. 23 also illustrates that the Koopman-ACC is in a safe and steady state during steering, while the 5dof-ACC is always in an oscillating state after passing a curve.
FIG. 24 shows that the solving time of each step of the controller is not more than 1.5ms, and the real-time requirement is met; the average solution time of 5dof-ACC is 130.22ms, which is far longer than the solution time of Koopman-ACC.
Both working conditions show that compared with the five-degree-of-freedom vehicle model, the linear model of the commercial vehicle can more comprehensively represent the nonlinear characteristics of the commercial vehicle and contains the nonlinear information of the tire. The self-adaptive cruise controller designed based on the model not only can ensure the running safety and the transverse stability of the commercial vehicle under various working conditions, but also can greatly reduce the solving time and ensure the real-time requirement in the application of an actual system.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.

Claims (10)

1. A lane keeping-based adaptive cruise control method for a commercial vehicle, comprising:
based on a Koopman operator theory, constructing a commercial vehicle linearization model based on a neural network;
collecting running information and front road information of a commercial vehicle, and constructing a vehicle-road relative position model;
acquiring state quantity related to turning of a curve of the commercial vehicle through the vehicle-road relative position model;
and designing a predictive controller based on the linearization model of the commercial vehicle, inputting the state quantity into the predictive controller, calculating to obtain a control quantity related to the turning of the curve of the commercial vehicle, and acting the control quantity on a control system of the commercial vehicle.
2. The lane keeping based commercial vehicle adaptive cruise control method according to claim 1, wherein: the state quantity comprises a transverse error between the current running position of the commercial vehicle and the target running position and an angle error between the current running direction of the commercial vehicle and the tangential direction of the target running position.
3. The lane keeping based commercial vehicle adaptive cruise control method according to claim 1, wherein: the state quantity also comprises a spacing error between the actual vehicle spacing and the expected vehicle spacing of the commercial vehicle and the pilot vehicle.
4. A lane keeping based adaptive cruise control method for a commercial vehicle according to claim 3, characterized in that U (k) = { U (k|k), U (k+ 1|k) … U (k+n) p -1|k) is a control sequence of the predictive controller, and the predictive expression of the predictive controller is:
Figure FDA0004079745030000011
Figure FDA0004079745030000012
Figure FDA0004079745030000013
wherein X is min ,X max To limit the limit value of the constraint range of the state quantity, u min ,u max For the limit value of the constraint range of the control quantity, Q, R is a positive weighting coefficient, Y (k+i|k) is the prediction horizon [0N ] p -1]The prediction output of the intra-frame is,
Figure FDA0004079745030000021
to predict the penalty term between output and expected output, ζ (k) = [ v ] ref (k)ω ref (k)] T In order to be able to measure the disturbance,
Figure FDA0004079745030000022
v x (k) Is a constant value;
Figure FDA0004079745030000023
5. the lane keeping based adaptive cruise control method for a commercial vehicle according to claim 4, wherein: the control quantity is the first control element in the optimal control sequence predicted and calculated by the prediction controller.
6. The lane keeping based adaptive cruise control method for a commercial vehicle according to claim 5, wherein: the control quantity comprises longitudinal force and front wheel rotation angle when the commercial vehicle turns around a curve.
7. The lane keeping based adaptive cruise control method for a commercial vehicle according to claim 6, wherein:
the constraints of the state quantity include constraints of lateral speed, yaw rate and longitudinal speed;
the control amount constraint includes a longitudinal force and a front wheel rotation angle constraint.
8. The lane keeping based commercial vehicle adaptive cruise control method according to claim 1, wherein: the Koopman operator theory is used for completing the evolution of a nonlinear system in a linear mode in a state space of infinite dimension, and an evolution expression is as follows:
Figure FDA0004079745030000024
wherein u is k For controlling quantity, ++>
Figure FDA0004079745030000025
Extracting a matrix for a state->
Figure FDA0004079745030000026
The method is used for estimating the state quantity of the original nonlinear system, and A and B are parameter matrixes.
9. The lane keeping-based adaptive cruise control method for a commercial vehicle according to claim 8, wherein the expression of the linearization model for the commercial vehicle is:
Figure FDA0004079745030000031
wherein u is k For controlling quantity, ++>
Figure FDA0004079745030000032
Extracting a matrix for a state->
Figure FDA0004079745030000033
For the estimated value of the state quantity of the original nonlinear system at the moment k, A DNN ,B DNN Is an optimal parameter matrix obtained through neural network training.
10. The lane-keeping based adaptive cruise control method for a commercial vehicle according to claim 1, wherein the expression of the vehicle-road relative position model is:
Figure FDA0004079745030000034
in the formula, v ref /v x Longitudinal speeds e of pilot and commercial vehicles, respectively x E is the distance error between the actual distance between the commercial vehicle and the pilot vehicle and the expected distance between the commercial vehicle and the pilot vehicle y For the current driving position and the front D of the commercial vehicle L A lateral error in the road center position e a For the current running direction and the front D of the commercial vehicle L Angle error in tangential direction of road center, D L To detect distance, K L =1/d is the road curvature of the vehicle centerline, d is the road radius. />
CN202310120341.1A 2023-02-16 2023-02-16 Self-adaptive cruise control method for commercial vehicle based on lane keeping Pending CN116061945A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310120341.1A CN116061945A (en) 2023-02-16 2023-02-16 Self-adaptive cruise control method for commercial vehicle based on lane keeping

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310120341.1A CN116061945A (en) 2023-02-16 2023-02-16 Self-adaptive cruise control method for commercial vehicle based on lane keeping

Publications (1)

Publication Number Publication Date
CN116061945A true CN116061945A (en) 2023-05-05

Family

ID=86178378

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310120341.1A Pending CN116061945A (en) 2023-02-16 2023-02-16 Self-adaptive cruise control method for commercial vehicle based on lane keeping

Country Status (1)

Country Link
CN (1) CN116061945A (en)

Similar Documents

Publication Publication Date Title
CN107561942B (en) Intelligent vehicle trajectory tracking model prediction control method based on model compensation
CN107415939B (en) Steering stability control method for distributed driving electric automobile
Cai et al. Implementation and development of a trajectory tracking control system for intelligent vehicle
CN112622903B (en) Longitudinal and transverse control method for autonomous vehicle in vehicle following driving environment
CN112733270A (en) System and method for predicting vehicle running track and evaluating risk degree of track deviation
CN111959527B (en) Automobile path tracking control method based on corner optimization sequence
CN110262229B (en) Vehicle self-adaptive path tracking method based on MPC
CN107585207B (en) A kind of vehicle line traffic control four-wheel steering system and its control method
Liu et al. Handling and stability integrated control of AFS and DYC for distributed drive electric vehicles based on risk assessment and prediction
CN114942642A (en) Unmanned automobile track planning method
CN114572251A (en) High-speed automatic driving automobile track tracking method based on predictive control
CN113581201A (en) Potential field model-based collision avoidance control method and system for unmanned automobile
CN112606843A (en) Intelligent vehicle path tracking control method based on Lyapunov-MPC technology
CN113341994B (en) Intelligent automobile path tracking control method based on optimal control of piecewise affine system
CN116560371A (en) Self-adaptive model predictive control-based automatic driving vehicle path tracking method
CN116834754A (en) Transverse and longitudinal cooperative control method for self-adaptive speed regulation of automatic driving vehicle
Li et al. Path tracking control based on the prediction of tire state stiffness using the optimized steering sequence
CN116061945A (en) Self-adaptive cruise control method for commercial vehicle based on lane keeping
CN115817509A (en) Multi-axis distributed driving vehicle steering auxiliary track tracking method based on AMPC
CN115933662A (en) Intelligent automobile trajectory tracking and stability control system and method based on adaptive model prediction control
CN116048081A (en) Automatic driving vehicle decision and regulation method considering safety boundary constraint
Zhao et al. Integrated longitudinal and lateral control system design and case study on an electric vehicle
CN114355941A (en) Vehicle path tracking method based on improved Stanley control
CN116198522B (en) Unmanned mining card transverse and vertical coupling hierarchical control method for complex mining area working conditions
Xie et al. Research on longitudinal control algorithm for intelligent automatic driving

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination