CN115129046A - Automatic driving vehicle path tracking method based on sliding mode neural network control - Google Patents
Automatic driving vehicle path tracking method based on sliding mode neural network control Download PDFInfo
- Publication number
- CN115129046A CN115129046A CN202210582974.XA CN202210582974A CN115129046A CN 115129046 A CN115129046 A CN 115129046A CN 202210582974 A CN202210582974 A CN 202210582974A CN 115129046 A CN115129046 A CN 115129046A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- control
- sliding mode
- automatic driving
- neural network
- 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
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 51
- 238000000034 method Methods 0.000 title claims abstract description 36
- 125000004122 cyclic group Chemical group 0.000 claims abstract description 5
- 230000000306 recurrent effect Effects 0.000 claims description 17
- 238000006073 displacement reaction Methods 0.000 claims description 16
- 230000006870 function Effects 0.000 claims description 15
- 239000013598 vector Substances 0.000 claims description 15
- 230000000694 effects Effects 0.000 claims description 5
- 238000013459 approach Methods 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 230000001133 acceleration Effects 0.000 claims description 2
- 230000033001 locomotion Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 6
- 238000013461 design Methods 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 230000001276 controlling effect Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000002401 inhibitory effect Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000005251 gamma ray Effects 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 230000017105 transposition Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05D—SYSTEMS FOR CONTROLLING OR REGULATING NON-ELECTRIC VARIABLES
- G05D1/00—Control of position, course, altitude or attitude of land, water, air or space vehicles, e.g. using automatic pilots
- G05D1/02—Control of position or course in two dimensions
- G05D1/021—Control of position or course in two dimensions specially adapted to land vehicles
- G05D1/0212—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory
- G05D1/0221—Control of position or course in two dimensions specially adapted to land vehicles with means for defining a desired trajectory involving a learning process
Landscapes
- Engineering & Computer Science (AREA)
- Aviation & Aerospace Engineering (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Control Of Position, Course, Altitude, Or Attitude Of Moving Bodies (AREA)
Abstract
The invention provides an automatic driving vehicle path tracking method based on sliding mode neural network control, which comprises the following steps: firstly, obtaining the current vehicle position deviation and course angle deviation through vehicle dynamics modeling and path tracking kinematics modeling; secondly, determining a control target and a control quantity of the path tracking system of the automatic driving vehicle, and designing a sliding mode control rate of vehicle position deviation and course angle deviation by using a nonsingular terminal sliding mode control method; nonlinear sum in the on-line fitting control rate is an unmodeled dynamic part through a cyclic neural network method, and finally, the position deviation and the course angle deviation tend to be zero through the active steering of the front wheels, so that the accurate control of the path tracking of the automatic driving vehicle is realized, the limited time convergence of the system can be ensured while the shake is effectively inhibited, and the expected path is accurately tracked.
Description
Technical Field
The invention belongs to the field of automatic driving vehicle motion control, and particularly relates to an automatic driving vehicle path tracking method based on sliding mode neural network control.
Background
Automated driving technology has made great progress over the past decades due to the rapid development of automotive cyber-physical systems, artificial intelligence, and advanced control technology. Compared with the traditional ground vehicle, the automatic driving vehicle has the advantages of better road safety, outstanding performance of reducing transportation cost, riding comfort and the like. The automatic driving vehicle is a typical high-technology comprehensive body and integrates the functions of environmental perception, motion planning and decision making, motion control and the like. In the motion control of an automatic driving vehicle, path tracking control is one of key problems, and relates to a steering system of the vehicle, so that the vehicle can drive along a desired path given by a motion planning and decision layer. However, the complex and varied traffic conditions and the high dynamic non-linearity of vehicles make autonomous vehicle path tracking control a significant challenge. Therefore, the control method for the path tracking system with good performance is designed, and the method has important significance.
At present, in the aspect of path tracking control of an automatic driving vehicle, common control algorithms comprise model prediction control, linear quadratic optimal control, robust control and sliding mode control. The model prediction control has the capability of rolling optimization practice correction and has a good control effect, but the algorithm is mainly used for a linear system, and the real-time control requirement is difficult to meet. The linear quadratic optimal control and robust control have the advantages of simple controller design and good robustness, but are difficult to meet the complex and variable running condition requirements of the automatic driving vehicle. Most existing control algorithms simplify vehicle systems into linear models and ignore external disturbances and uncertainty factors.
In the "vehicle path tracking method based on improved Stanley control" disclosed in chinese patent publication CN114355941A by reqi et al, the path tracking method is first modeled according to a Stanley control algorithm based on a two-degree-of-freedom vehicle motion model, and then a fuzzy controller is designed to adjust the controller gain parameters. However, the Stanley control method has high parameter dependency, the fuzzy rule of the fuzzy controller designed for optimizing the parameters is relatively fixed, and the method cannot adapt to the complex working environment of the automatic driving vehicle in the actual working process, including the changed longitudinal speed of the vehicle, the lateral deflection rigidity of the tire, the lateral wind, the road surface disturbance and the like.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides the automatic driving vehicle path tracking method based on sliding mode neural network control, which can overcome the problems of uncertain parameters, nonlinearity, modeling dynamics and the like in the automatic driving vehicle path tracking model, can ensure the dynamic response speed block of the automatic driving vehicle path tracking, inhibits shake and finite time convergence and has high tracking precision.
In order to achieve the aim of the invention, the invention provides an automatic driving vehicle path tracking method based on sliding mode neural network control, which is realized by the following steps:
step 1: establishing an automatic driving vehicle path tracking control model, which comprises establishing a vehicle two-degree-of-freedom model and a vehicle path tracking kinematics model;
and 2, step: designing a path tracking controller aiming at the established automatic driving vehicle path tracking control model, wherein the method comprises the steps of determining a system control target, designing a proper sliding mode control rate by utilizing a nonsingular terminal sliding mode control algorithm, and converging the system in a limited time through the proper control rate, inhibiting output shake and meeting the Lyapunov stability judgment;
and 3, step 3: and (3) approximating the parameter uncertainty in the control rate obtained in the step (2) and the unknown part caused by unmodeled dynamics by adopting a cyclic neural network estimator for updating the weight value on line, and combining the sliding mode control rate to obtain the control rate of the wheel corner.
And 4, step 4: and calculating the corner of the front wheel according to the control rate of the corner of the wheel, and controlling the front wheel to steer actively through a steering system to realize the path tracking control of the automatic driving vehicle.
Further, in step 1, the vehicle two-degree-of-freedom model:
tire force can be expressed as:
vehicle path tracking kinematics model:
integrated, autonomous vehicle path tracking control model:
wherein v is x And v y Representing the longitudinal speed and the lateral speed of the autonomous vehicle, respectively; m is the autonomous vehicle mass; i is z Representing an autonomous vehicle yaw moment couple; f yf And F yr Front and rear axle tire lateral forces, respectively; r represents a vehicle yaw rate; l f And l r Respectively representing the distance between the center of mass of the vehicle and the front and rear axes; d 1 And d 2 Indicating an external disturbance. Delta f Indicating the corner of the front wheel of the automatic driving vehicle; c f And C r Indicating the total cornering stiffness of the front and rear wheels, the value of which varies as a function of the vehicle load, the tyre pressure and the vehicle driving conditions; delta. for the preparation of a coating f Indicating a front wheel turning angle; alpha is alpha f And alpha r Respectively, the front and rear wheel side slip angles. E representing the preview pointA lateral displacement error;representing a deviation between a vehicle heading angle and a desired path; ρ represents the path curvature of the preview point; l represents the pre-aiming distance; d 3 And d 4 Representing the modeling error. x is the number of 1 And x 2 And y is the system output.
In the formula, andw (t) represents unmodeled dynamics, system uncertainty, and external interference. u is δ f As a control input to the system.
Further, in step 2, the control objective of the automatic driving vehicle path tracking control system is to make the transverse displacement error and the course angle deviation between the automatic driving vehicle and the expected path zero, so as to realize that the automatic driving vehicle accurately tracks the expected path.
Further, in step 2, the path tracking control rate of the automatic driving vehicle is realized by a sliding mode control algorithm based on a nonsingular terminal, and the method comprises the following steps:
defining control objectivesSo that x 1 → 0, where ∈ denotes the lateral displacement error of the preview point;representing a deviation between a vehicle heading angle and a desired path; xi is an adjustable parameter.
According to a nonsingular terminal sliding mode control algorithm, the sliding mode surface of the path tracking system of the automatic driving vehicle is designed as follows:
wherein λ > 0,2 > p/q > 1 and p, q are positive odd numbers.
The derivative of the slip form surface is:
further, the air conditioner is provided with a fan,
in order to make the state of the control system quickly return to the nonlinear sliding mode surface and reduce shake, the switching control rate is defined as:
u s =-g -1 (k 1 s+k 2 tanh(s))
in the formula, k 1 > 0 and k 2 And more than 0 is an adjustable parameter.
Further, the control rate can be written as:
u=-g -1 (κ 2 +f+k 1 s+k 2 tanh(s))
according to the Lyapunov functionThe stability judgment of (2) makes the system approach to a nonsingular terminal sliding mode surface s to be 0 in a limited time, and realizes the state convergence of the system.
Further, the step 3 of designing the nonlinear unknown parts f and g in the recurrent neural network online estimation control rate specifically comprises the following steps:
the adopted neural network structure belongs to a recurrent neural network and comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a feedback unit. In order not to generate interference, two independent recurrent neural networks are used to estimate f and g, respectively. Its algorithm can be expressed as:
x=[x 1 x 2 ] T
the output of the neural network is:
wherein,
in the formula h f,g =[h j ] T (j ═ 1, 2, …, n) represents a gaussian output function; x is the number of i (i ═ 1, 2, …, n) is the input to the neural network; exp is an exponential function with a constant e as base. c. C ij A vertex representing a Gaussian function; b ij Is the width of a Gaussian function; q j Is the cycle weight; w * And V * Are the connection weights between the connection hidden layer and the output layer. W * ,V * ,Andthe optimal vector of f and g can be estimated; whereinCan be expressed asAndε f and ε g Respectively representing the error between the actual value and the optimal value;andis an estimate of the optimal vector. And is provided withm and n respectively represent the number of input layers and the number of hidden layers, and design selection can be carried out according to an actual system.
Further, the control rate obtained in step 4 according to steps 2 and 3 is:
the autonomous vehicle applies the front wheel steering angle calculated by the above control rate through a steering mechanism to ensure that the autonomous vehicle tracks the desired path.
Compared with the prior art, the invention has the beneficial effects that at least:
according to the automatic driving vehicle path tracking method based on sliding mode neural network control, shaking is effectively inhibited through a terminal sliding mode control algorithm and a Lyapunov stability principle, meanwhile, limited time convergence of a system can be guaranteed, and an expected path can be accurately tracked; in addition, the method estimates the uncertainty of system parameters and unknown parts caused by unmodeled dynamics through the recurrent neural network, and has the advantages of high precision and quick response.
Drawings
Fig. 1 is a flowchart of an automatic driving vehicle path tracking method based on sliding mode neural network control according to an embodiment of the present invention.
FIG. 2 is a schematic view of a lateral two-degree-of-freedom model of an autonomous vehicle in an embodiment of the invention.
FIG. 3 is a schematic diagram of a lateral kinematics model for path tracking of an autonomous vehicle in an embodiment of the present invention.
FIG. 4 is a lateral error diagram of an autonomous vehicle under a double lane change condition, according to an embodiment of the present invention.
FIG. 5 is a control input diagram of an autonomous vehicle under a dual lane change condition, according to an embodiment of the present invention.
FIG. 6 is a comparison graph of the actual movement path and the expected path of the autonomous vehicle under the condition of double lane changing according to the embodiment of the invention.
Detailed Description
The invention is further described in connection with the accompanying drawings and specific embodiments so that those skilled in the art may better understand the invention and practice it, but the examples are not intended to be limiting.
Referring to fig. 1, the method for tracking a path of an autonomous vehicle based on sliding mode neural network control according to the present invention includes the following steps:
step 1, establishing an automatic driving vehicle path tracking control model, which comprises the steps of respectively establishing a vehicle two-degree-of-freedom model and a vehicle path tracking kinematics model, considering vehicle dynamics behaviors and kinematics behaviors in the modeling of a vehicle system, and forming a closed loop system with road curvature information.
In some embodiments of the invention, step 1 comprises:
1) as shown in fig. 2, x-y is a coordinate system fixed to the vehicle, where x represents the longitudinal direction of the vehicle and y represents the lateral direction of the vehicle, with the origin of the coordinate system located at the center of mass of the vehicle. The embodiment of the invention focuses on the lateral dynamics and the stability of the vehicle, neglects the longitudinal force behavior of the vehicle, and establishes a two-degree-of-freedom model of the vehicle under the assumption of a small corner:
where tire force can be expressed as:
in the formula,representing the lateral acceleration, v, of the vehicle y And v x Representing the lateral and longitudinal speed of the vehicle; m represents vehicle mass; f yf Indicating the total cornering power of the front wheel, F yr Indicating total cornering power of the rear wheel, singlyThe bits are: n; r andrespectively representing the yaw angle and the yaw rate of the vehicle; i is z Representing the yaw moment of inertia of the vehicle; l f And l r Respectively refer to the distance from the center of mass of the vehicle to the front axle and the rear axle; delta f Indicating a front wheel steering angle; c f And C r The cornering powers of the front and rear tires of the autonomous vehicle, respectively, whose values vary with the vehicle load, the tire pressure and the vehicle running condition, are given by: n/rad; alpha (alpha) ("alpha") f Indicating front wheel slip angle, α r Represents the rear wheel side slip angle, and the unit is: rad; d 1 And d 2 Indicating an external disturbance.
Further, the two-degree-of-freedom model of the vehicle can be expressed as:
2) as shown in fig. 3, a vehicle path tracking kinematics model is established:
wherein e represents the transverse displacement error of the preview point, and the unit is as follows: m;representing a heading angle deviation between a vehicle heading angle and a desired path in units of: rad; ρ represents the path curvature of the preview point in units of: 1/m; l represents the pre-aiming distance and,the unit is as follows: m; d is a radical of 3 And d 4 Representing the modeling error of each formula;representing the heading angular velocity.
3) Establishing an automatic driving vehicle path tracking control model:
further obtaining:
in the formula, x 1 Representing a linear combination of lateral displacement error and course angle error, x 2 Representing the derivative, x, of the linear combination 1 And x 2 And y is the system output.
Wherein:
ξ 1 indicating the magnitude of the effect of the vehicle heading angle error on the path tracking error as compared to the lateral position error, ξ 2 Indicating the magnitude of the effect of the vehicle heading angle error derivative compared to the lateral position error derivative on the path tracking error change.
Where w (t) represents unmodeled dynamics, system uncertainty, and external interference. u is δ f As a control input to the system.And withRespectively representing the derivatives of the two system state quantities, and deriving the derivatives from a defined system state equation;andthe first order and the second order derivatives of the vehicle transverse error and the vehicle course angle error are represented, and the control rate can be designed more flexibly by introducing the high order derivatives of the error, so that the path tracking is more accurate; because the error of the lateral displacement of the vehicle and the error of the course angle are controlled simultaneously, xi is utilized 1 To regulate the ratio of two errors, xi 2 According to xi 1 The formula is derived; in order to simplify a complex formula, two symbols f and g are used for replacing a part of formula term numbers, f represents a part of unknown nonlinearity and modeling disturbance in a system equation, g is a coefficient of system input and can be directly obtained according to vehicle parameters, and xi is an adjustable parameter.
And 2, determining a control target and designing a control rate. The method aims at designing an automatic driving vehicle path tracking controller for the established automatic driving vehicle path tracking control model, comprises the steps of determining a system control target, determining a sliding mode control rate by utilizing a nonsingular terminal sliding mode control algorithm, and enabling the system to be in finite time convergence, inhibiting output shake and meeting the requirements of Liya Punuff stability judgment through the sliding mode control rate
In some of the embodiments of the present invention, the control objective of the autonomous vehicle path tracking controller is to make the lateral displacement error and the heading angle deviation between the autonomous vehicle and the desired path zero, enabling the autonomous vehicle to accurately track the desired path. Defining control objectivesSo that x 1 → 0, where ∈ denotes the lateral displacement error of the preview point;representing a deviation between a vehicle heading angle and a desired path; xi is an adjustable parameter.
In some embodiments of the present invention, the sliding mode control rate is determined by:
according to a nonsingular terminal sliding mode control algorithm, a sliding mode surface s of the path tracking system of the automatic driving vehicle is designed as follows:
for the convenience of sliding mode surface design, lambda is larger than 0,2 is larger than p/q is larger than 1, and p and q are positive odd numbers. λ represents the noise immunity of the slip form surface, and p and q represent the convergence rate of the slip form surface.
The derivative of the slip form surface is:
further, the air conditioner is provided with a fan,
for simplification of the formula, use is made of κ 1 、κ 2 Instead of a part of the number of terms,
in order to make the state of the control system quickly return to the nonlinear sliding mode surface and reduce the shake, the switching rate is defined as follows:
u s =-g -1 (k 1 s+k 2 tanhs))
in the formula, k 1 > 0 and k 2 And more than 0 is an adjustable parameter. k is a radical of 1 、k 2 The initial value is obtained from the defined parameters of the sliding mode surface and the initial state of the system and can be adaptively changed along with the movement of the system.
Further, by introducing the above switching rate, the control rate of the system can be obtained as follows:
u=-g -1 (κ 2 +f+k 1 s+k 2 tanh(s))
the control rate of the front wheel steering angle of the vehicle obtained by the nonsingular terminal sliding mode control method can enable the transverse position deviation and course angle deviation of the actual running path and the reference path of the vehicle to return to 0, so that the purpose of path tracking is achieved.
According to the Lyapunov functionThe stability judgment of (2) makes the system approach to a nonsingular terminal sliding mode surface s to be 0 in a limited time, and realizes the state convergence of the system.
And 3, approximating an unknown part caused by parameter uncertainty and unmodeled dynamics in the control rate by adopting a cyclic neural network estimator for updating the weight value on line, and combining the sliding mode control rate to obtain the control rate of the wheel corner.
Because an unknown nonlinear and modeling disturbance part exists in the sliding mode control rate and cannot be obtained through direct calculation, the numerical value of the part can be fitted in real time by utilizing a recurrent neural network, and the numerical value is substituted into a sliding mode control rate calculation formula to obtain the wheel rotation angle of the vehicle.
The adopted neural network structure belongs to a recurrent neural network and comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a feedback unit. In order not to generate interference, two independent recurrent neural networks are used to estimate f and g, respectively. Its algorithm can be expressed as:
x=[x 1 x 2 ] T
the output of the neural network is:
wherein,
in the formula h j,k =[h j ] T (j ═ 1, 2, …, n) denotes the gaussian output function of each hidden layer node, h j Representing the output of the jth hidden layer node of the recurrent neural network, k representing the node position, exp being an exponential function with a constant e as the base, x i (i ═ 1, 2, … m) is the input to the neural network, i represents the ith node number of the neural network; c. C ij A vertex representing a Gaussian function; q is a cyclic weight; b ij Is the width of a Gaussian function; x denotes the state of the path tracking system, x 1 Representing a linear combination of lateral displacement error and course angle error, x 2 To representDerivative of the linear combination, T denotes transposition, W * And V * Are the connection weights between the connection hidden layer and the output layer. W * ,V * ,Andthe optimal vector of f and g can be estimated; whereinCan be expressed asAndb f ,c f width vector and center point vector values representing the nodes of the hidden layer of the recurrent neural network for f, b g ,c g Representing width vectors and central point vector values of the nodes of the hidden layer of the recurrent neural network aiming at g; epsilon f And ε g Respectively representing the error between the actual value and the optimal value;andis an estimate of the optimal vector. And is Representing the weight error from the hidden layer to the output layer,representing the error of the estimated vector for f,is the loop weight estimation error for f and g; gamma ray 1 、γ 2 、γ 3 、γ 4 Expressing Lyapunov parameters, and obtaining the Lyapunov parameters through artificial regulation; m and n respectively represent the number of input layers and the number of hidden layers, and can be selected according to the actual system, and in some embodiments of the invention, m is 2 and n is 5.
The control rate expression of the obtained wheel rotation angle is:
Using the obtained path tracking controller to perform on-line control of the autonomous vehicle at a control rate of The global system meets the Lyapunov stability principle, and has the advantages of limited time convergence and small input shake.
And 4, calculating the corner of the front wheel according to the control rate of the corner of the wheel, and controlling the front wheel to actively steer through a steering system to realize the path tracking control of the automatic driving vehicle.
The control rate of the front wheel steering angle of the automatic driving vehicle can be calculated by a control rate expression of the upper wheel steering angle, the calculation result of the control rate is the actual input value of the system, the expected path can be ensured to be tracked by the vehicle in real time by automatically regulating and controlling the front wheel steering angle, and the automatic driving vehicle has the characteristics of strong anti-interference capability, quick convergence and small shaking.
In some embodiments of the present invention, a storage medium, which may be a storage medium such as a ROM, a RAM, a magnetic disk, an optical disk, or the like, is further provided, and the storage medium stores one or more programs, and when the programs are executed by a processor, the method for tracking a vehicle path of an autonomous driving vehicle based on sliding mode neural network control provided in the foregoing embodiments is implemented.
In some embodiments of the present invention, an apparatus is further provided, where the computing apparatus may be a desktop computer, a notebook computer, a smart phone, a PDA handheld terminal, a tablet computer, or other terminal apparatus with a display function, the computing apparatus includes a processor and a memory, the memory stores one or more programs, and when the processor executes the programs stored in the memory, the processor implements the sliding mode neural network control-based automatic driving vehicle path tracking method provided in the foregoing embodiments.
In some of the embodiments of the present invention, the main technical performance indicators and equipment parameters used are: m 1830kg, I z =3234kg·m 2 ,l f =1.4m,l r =1.65m,C f =2*41428N/rad,C r =2*36443N/rad, ξ=0.5,λ=0.1,p=9,q=7,k 1 =180,k 2 =100,γ 1 =γ 3 =8,γ 2 =γ 4 2. FIG. 4 shows a diagram of lateral displacement error of the path of the autonomous vehicle under the condition that the embodiments relate to double lane changing, FIG. 5 is a diagram of control input of the autonomous vehicle under the condition that the embodiments relate to double lane changing, and FIG. 6 is a diagram comparing the actual moving path with the expected path of the vehicle. As can be seen from the figure, the method can effectively control the vehicle to track the expected path, ensure that the transverse displacement deviation and the course angle deviation are in a very small range, simultaneously control the input to inhibit shake, and effectively improve the vehicle tracking performance and the operation stability.
The sliding mode variable structure control adopted by the embodiment of the invention is a control algorithm with strong robustness and strong anti-interference capability, and the shake problem can be solved by switching the control rate, the approach rate and the like. In addition, the adopted neural network has the capability of fitting any nonlinearity, and the control performance can be further improved by fitting the unmodeled dynamics and nonlinearity of the automatic driving vehicle through the neural network.
The above embodiments are merely illustrative of the technical ideas and features of the present invention and are intended to enable those skilled in the art to better understand and implement the same. The protection scope of the present invention is not limited to the above embodiments, and all equivalent changes and modifications made according to the principles and design ideas disclosed by the present invention are within the protection scope of the present invention.
Claims (10)
1. An automatic driving vehicle path tracking method based on sliding mode neural network control is characterized by comprising the following steps:
establishing an automatic driving vehicle path tracking control model;
designing a path tracking controller aiming at the established automatic driving vehicle path tracking control model, wherein the path tracking controller comprises a system control target and a sliding mode control rate determined by a nonsingular terminal sliding mode control algorithm;
a cyclic neural network estimator for updating the weight value on line is adopted to approximate the parameter uncertainty in the control rate and the unknown part caused by unmodeled dynamics, and the control rate of the wheel corner is obtained by combining the sliding mode control rate;
and calculating the corner of the front wheel according to the control rate of the corner of the wheel, and controlling the front wheel to steer actively through a steering system to realize the path tracking control of the automatic driving vehicle.
2. The method for tracking the path of the automatic driving vehicle based on the sliding mode neural network control according to claim 1, wherein the establishing of the automatic driving vehicle path tracking control model comprises the following steps:
establishing a two-degree-of-freedom model of the vehicle;
establishing a vehicle path tracking kinematic model;
establishing an automatic driving vehicle path tracking control model based on the vehicle two-degree-of-freedom model and the vehicle path tracking kinematic model, wherein the automatic driving vehicle path tracking control model is as follows:
further obtaining:
wherein:
in the formula, x 1 Representing a linear combination of lateral displacement error and course angle error, x 2 Representing the derivative of the linear combination, y being the system output,andare respectively x 1 、x 2 The derivative of (a) is determined,which represents the angular speed of the heading,andthe first and second derivatives of the vehicle lateral error and the second derivative of the vehicle heading angle error are shown, w (t) represents unmodeled dynamics, system uncertainty and external disturbance, and u is delta f As a control input to the system, δ f Indicates the front wheel steering angle, f denotes an unknown non-lineSex and modeling disturbance, g is the coefficient of the system input, v y And v x Representing the lateral and longitudinal speed of the vehicle, m representing the vehicle mass, r representing the yaw angle of the vehicle, l f And l r Respectively the distance of the centre of mass of the vehicle to the front axle and the rear axle, C f And C r Yaw stiffness, I, of front and rear tires, respectively, of an autonomous vehicle z Representing the yaw moment of inertia of the vehicle, L representing the pre-aiming distance, xi being an adjustable parameter, xi 1 Indicating the magnitude of the effect of the vehicle heading angle error on the path tracking error as compared to the lateral position error, ξ 2 Indicating the magnitude of the effect of the vehicle heading angle error derivative compared to the lateral position error derivative on the path tracking error change.
3. The automatic driving vehicle path tracking method based on sliding mode neural network control of claim 2, characterized in that the vehicle two-degree-of-freedom model is as follows:
5. The method for tracking the path of the automatic driving vehicle based on the sliding-mode neural network control according to claim 1, wherein the control target of the path tracking controller of the automatic driving vehicle is to make the transverse displacement error and the course angle deviation between the automatic driving vehicle and the expected path be zero, realize the accurate tracking of the expected path by the automatic driving vehicle and define the control targetSo that x 1 → 0, where ∈ denotes the lateral displacement error of the preview point;representing a deviation between a vehicle heading angle and a desired path; and xi is an adjustable parameter.
6. The method for tracking the path of the automatic driving vehicle based on the sliding mode neural network control according to claim 1, wherein the determining the sliding mode control rate by using the nonsingular terminal sliding mode control algorithm comprises the following steps:
according to a nonsingular terminal sliding mode control algorithm, the sliding mode surface of the path tracking system of the automatic driving vehicle is designed as follows:
in the formula, x 1 Representing a linear combination of lateral displacement error and course angle error, x 2 The derivative of the linear combination is expressed, lambda represents the anti-interference capability of the sliding mode surface, and p and q represent the convergence speed of the sliding mode surface;
the derivative of the slip form surface is:
further, the air conditioner is provided with a fan,
where f represents the unknown non-linearity and modeling perturbation, g is the coefficient of the system input, u is the control input to the system,
defining a handover control rate u s Comprises the following steps:
u s =-g -1 (k 1 s+k 2 tanh(s))
in the formula, k 1 > 0 and k 2 More than 0 is an adjustable parameter;
the sliding mode control rate is:
u=-g -1 (κ 2 +f+k 1 s+k 2 tanh(s))
7. The method for tracking the path of the automatic driving vehicle based on the sliding mode neural network control is characterized in that the recurrent neural network estimator comprises an input layer, a hidden layer and an output layer, wherein the hidden layer comprises a feedback unit, and in order not to generate interference, two independent recurrent neural networks are adopted to estimate f and g respectively, and the algorithm is represented as follows:
x=[x 1 x 2 ] T
the output of the neural network is:
wherein,
in the formula, h j,k =[h j ] T (j ═ 1, 2, …, n) denotes the gaussian output function of each hidden layer node, h j Representing the output of the jth hidden layer node of the recurrent neural network, and k representing the node position; x denotes the state of the path tracking system, x i (i ═ 1, 2, … m) is the input to the neural network; exp is an exponential function based on a constant e, c ij A vertex representing a Gaussian function; b ij Is the width of a Gaussian function; q j Is the cycle weight; w * And V * Is the connection weight, W, between the connection hidden layer and the output layer * ,V * ,Andthe optimal vector of f and g can be estimated; whereinCan be expressed as Andε f and ε g Respectively representing the error between the actual and the optimum value, b f ,c f Width vector and center point vector values representing the nodes of the hidden layer of the recurrent neural network for f, b g ,c g Representing width vectors and central point vector values of the nodes of the hidden layer of the recurrent neural network aiming at g;andis an estimate of the optimal vector, andm and n represent the number of inputs and the number of hidden layers, respectively.
8. The automatic driving vehicle path tracking method based on sliding-mode neural network control according to any one of claims 1 to 7, wherein the control rate of the wheel turning angle isu is the control input of the system, k 1 > 0 and k 2 More than 0 is an adjustable parameter, s is a sliding mode surface,represents the estimation value of the recurrent neural network to f, f represents the unknown nonlinearity and modeling disturbance,x 2 and the derivative of the linear combination of the transverse displacement error and the course angle error is shown, lambda represents the disturbance rejection capability of the sliding mode surface, and p and q represent the convergence speed of the sliding mode surface.
9. A storage medium storing a program which when executed by a processor implements the method of any one of claims 1 to 8.
10. An apparatus comprising a processor and a memory for storing a program executable by the processor, wherein the processor, when executing the program stored in the memory, performs the method of any of claims 1 to 8.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210582974.XA CN115129046A (en) | 2022-05-26 | 2022-05-26 | Automatic driving vehicle path tracking method based on sliding mode neural network control |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210582974.XA CN115129046A (en) | 2022-05-26 | 2022-05-26 | Automatic driving vehicle path tracking method based on sliding mode neural network control |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115129046A true CN115129046A (en) | 2022-09-30 |
Family
ID=83376318
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210582974.XA Pending CN115129046A (en) | 2022-05-26 | 2022-05-26 | Automatic driving vehicle path tracking method based on sliding mode neural network control |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115129046A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140074361A1 (en) * | 2012-09-12 | 2014-03-13 | Kabushiki Kaisha Topcon | Construction Machine Control Method And Construction Machine Control System |
WO2018023201A1 (en) * | 2016-08-03 | 2018-02-08 | 孟强 | Adaptive terminal sliding mode control method |
CN107831761A (en) * | 2017-10-16 | 2018-03-23 | 中国科学院电工研究所 | A kind of path tracking control method of intelligent vehicle |
CN113110512A (en) * | 2021-05-19 | 2021-07-13 | 哈尔滨工程大学 | Benthonic AUV self-adaptive trajectory tracking control method for weakening unknown interference and buffeting influence |
-
2022
- 2022-05-26 CN CN202210582974.XA patent/CN115129046A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20140074361A1 (en) * | 2012-09-12 | 2014-03-13 | Kabushiki Kaisha Topcon | Construction Machine Control Method And Construction Machine Control System |
WO2018023201A1 (en) * | 2016-08-03 | 2018-02-08 | 孟强 | Adaptive terminal sliding mode control method |
CN107831761A (en) * | 2017-10-16 | 2018-03-23 | 中国科学院电工研究所 | A kind of path tracking control method of intelligent vehicle |
CN113110512A (en) * | 2021-05-19 | 2021-07-13 | 哈尔滨工程大学 | Benthonic AUV self-adaptive trajectory tracking control method for weakening unknown interference and buffeting influence |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107561942B (en) | Intelligent vehicle trajectory tracking model prediction control method based on model compensation | |
CN110780594B (en) | Path tracking method and system of intelligent vehicle | |
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 | |
CN113126623B (en) | Adaptive dynamic sliding mode automatic driving vehicle path tracking control method considering input saturation | |
CN114967475B (en) | Unmanned vehicle trajectory tracking and stability robust control method and system | |
CN111258218B (en) | Intelligent vehicle path tracking method based on maximum correlation entropy criterion | |
CN114047722A (en) | MPC-based path tracking control method for adaptive attitude correction | |
CN114572251A (en) | High-speed automatic driving automobile track tracking method based on predictive control | |
CN113671962A (en) | Transverse control method of unmanned articulated sweeper | |
CN113183957A (en) | Vehicle control method, device and equipment and automatic driving vehicle | |
CN112606843A (en) | Intelligent vehicle path tracking control method based on Lyapunov-MPC technology | |
CN113031443A (en) | Vehicle transverse motion control method with active safety and self-adaptive preview | |
Qinpeng et al. | Path tracking control of wheeled mobile robot based on improved pure pursuit algorithm | |
CN116560371A (en) | Self-adaptive model predictive control-based automatic driving vehicle path tracking method | |
Fu et al. | Nmpc-based path tracking control strategy for autonomous vehicles with stable limit handling | |
Xing et al. | Recurrent neural network non‐singular terminal sliding mode control for path following of autonomous ground vehicles with parametric uncertainties | |
CN111736598B (en) | Harvester path tracking control method and system based on adaptive neural network | |
He et al. | Composite deep learning control for autonomous bicycles by using deep deterministic policy gradient | |
CN115129046A (en) | Automatic driving vehicle path tracking method based on sliding mode neural network control | |
CN112764347B (en) | Intelligent vehicle path tracking method based on maximum correlation entropy criterion | |
CN114030526B (en) | Vehicle active steering control method and system | |
CN115167135A (en) | Feedback and model feedforward cascade unmanned vehicle self-tendency optimal position and posture control system | |
CN114802200A (en) | Method for tracking and controlling stability of tracks of intelligent automobile under extreme working conditions | |
CN114740845A (en) | Vehicle tracking control method based on immersion and invariant manifold |
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 |