CN114379583B - Automatic driving vehicle track tracking system and method based on neural network dynamics model - Google Patents
Automatic driving vehicle track tracking system and method based on neural network dynamics model Download PDFInfo
- Publication number
- CN114379583B CN114379583B CN202111508163.7A CN202111508163A CN114379583B CN 114379583 B CN114379583 B CN 114379583B CN 202111508163 A CN202111508163 A CN 202111508163A CN 114379583 B CN114379583 B CN 114379583B
- Authority
- CN
- China
- Prior art keywords
- vehicle
- model
- neural network
- layer
- lateral
- 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.)
- Active
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 72
- 238000000034 method Methods 0.000 title claims abstract description 34
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 24
- 230000008569 process Effects 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000003062 neural network model Methods 0.000 claims abstract description 11
- 238000004088 simulation Methods 0.000 claims abstract description 11
- 238000013461 design Methods 0.000 claims abstract description 10
- 238000006073 displacement reaction Methods 0.000 claims description 42
- 230000006870 function Effects 0.000 claims description 40
- 238000005070 sampling Methods 0.000 claims description 40
- 230000008859 change Effects 0.000 claims description 26
- 230000004913 activation Effects 0.000 claims description 22
- 238000005457 optimization Methods 0.000 claims description 16
- 230000001133 acceleration Effects 0.000 claims description 12
- 241000282414 Homo sapiens Species 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 9
- 230000006399 behavior Effects 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 5
- 238000013135 deep learning Methods 0.000 claims description 5
- 238000005096 rolling process Methods 0.000 claims description 5
- 238000012795 verification Methods 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 4
- 230000010354 integration Effects 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000012938 design process Methods 0.000 claims description 2
- 239000000725 suspension Substances 0.000 description 6
- 230000008878 coupling Effects 0.000 description 5
- 238000010168 coupling process Methods 0.000 description 5
- 238000005859 coupling reaction Methods 0.000 description 5
- 230000015556 catabolic process Effects 0.000 description 2
- 238000006731 degradation reaction Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 206010039203 Road traffic accident Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000006880 cross-coupling reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
- B60W60/00—Drive control systems specially adapted for autonomous road vehicles
- B60W60/001—Planning or execution of driving tasks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/10—Geometric CAD
- G06F30/15—Vehicle, aircraft or watercraft design
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2111/00—Details relating to CAD techniques
- G06F2111/04—Constraint-based CAD
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Geometry (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Computer Hardware Design (AREA)
- Automation & Control Theory (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Human Computer Interaction (AREA)
- Medical Informatics (AREA)
- Mechanical Engineering (AREA)
- Transportation (AREA)
- Aviation & Aerospace Engineering (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Steering Control In Accordance With Driving Conditions (AREA)
- Control Of Driving Devices And Active Controlling Of Vehicle (AREA)
Abstract
The invention discloses an automatic driving vehicle track tracking system and method based on a neural network dynamics model, comprising a neural network vehicle dynamics model part, a vehicle dynamics data acquisition part (comprising a driving simulator and virtual simulation platform CarSim based simulation data acquisition process and real world automatic driving vehicle data acquisition), a training part of the neural network model and a model predictive control algorithm design; by combining the established neural network vehicle dynamics prediction model with the model prediction control algorithm, compared with an end-to-end control algorithm, the control algorithm provided by the invention has higher interpretability. And the method can realize the tracking control of the expected track under different road conditions and running conditions, ensure the tracking precision of the path and simultaneously give consideration to the transverse and longitudinal stability, and lay a foundation for developing a high-performance motion controller of an automatic driving vehicle.
Description
Technical Field
The invention relates to the technical field of intelligent vehicle automatic driving, in particular to an automatic driving vehicle track tracking system and method based on a neural network dynamics model.
Background
With the continuous upgrade of the automobile 'new and quadruple' and the rapid development of artificial intelligence technology, the automatic driving automobile has become a trend of the revolution of the traditional automobile industry and a research hotspot of the world vehicle engineering. Autonomous vehicles are expected to free people from tedious long distance driving and have great potential in reducing traffic congestion and reducing traffic accidents. Autonomous vehicles are typically composed of context aware, path planning and control execution systems, where the construction of vehicle models is critical to trajectory planning and control, which is the basis for high safety and high reliability trajectory tracking control.
Currently, automatic driving vehicle trajectory tracking control is mainly classified into a control method based on a vehicle kinematic model and a control method based on a vehicle kinematic model. The controller designed based on the vehicle kinematics model can ensure certain control performance under the working condition of low speed and small curvature. However, when the vehicle body speed is high and the road curvature is high, the dynamics of the vehicle itself is not considered, resulting in degradation of the control performance and degradation of the running quality. The controller designed based on the vehicle dynamics model cannot sufficiently consider the vertical motion of the vehicle, the suspension motion characteristics and the longitudinal and transverse coupling relation of the force of the tire when the vehicle runs at a high speed due to the simplification of the vehicle model. Such models are mostly built based on differential equations, which are typically reduced to two-degree-of-freedom or three-degree-of-freedom vehicle dynamics models due to the high degrees of freedom of the real world vehicle itself. While this approach reduces computational complexity, the vehicle dynamics model employed typically does not adequately account for the vertical motion of the vehicle when traveling at high speeds, suspension motion characteristics, and the cross-coupling of the forces of the tires. Therefore, when the vehicle runs at a high speed, a large track tracking error can be generated, and the actual requirement of high-level automatic driving of the intelligent vehicle cannot be met.
Disclosure of Invention
In order to solve the technical problems, the invention provides an automatic driving vehicle track tracking system based on a neural network vehicle dynamics model, which mainly comprises a neural network vehicle dynamics model part, a vehicle dynamics data acquisition part (comprising a driving simulator and virtual simulation platform CarSim simulation data acquisition process and real-world automatic driving vehicle data acquisition), a training part of the neural network model and a model prediction control algorithm design.
The neural network vehicle dynamics model part designs a neural network model with delay input by using a feedforward neural network, wherein the hidden layer of the model is two layers, each layer has 100 neurons, the activation layer selects Softplus activation functions, the input of the model adopts vehicle control and state information at two moments, and further, the first derivatives of the yaw rate and the lateral speed of the vehicle are predicted.
And a vehicle dynamics data acquisition part for establishing a real-time virtual simulation platform with the CarSim through a driving simulator, selecting an automatic driving test map Mcity and acquiring data based on normal driving behaviors of human beings. Because the road curvature of a vehicle has a great influence on the drivability of the vehicle, in order to collect complete data, the vehicle is driven on different roads, including straight roads, curved roads, and the like, and single-lane and double-lane changes are performed.
In the real-world vehicle data acquisition process, an automatic driving vehicle is controlled by a human driver to perform linear motion, curve motion, single-lane change, double-lane change and the like.
Based on a feedforward neural network vehicle dynamics prediction model training part, the obtained virtual experiment platform simulation data set and real-world real vehicle data are combined and then divided into 80% of training set, 10% of verification set and 10% of test set. The loss function was chosen as the MSE loss function, the optimizer was chosen as Adam, the batch size was set to 1000, and the learning rate was set to 0.0003. The network model is trained based on Pytorch deep learning frameworks.
And designing a model predictive control algorithm part based on the trained neural network vehicle dynamics model, and obtaining the optimal front wheel corner through online solution of rolling optimization so as to track a reference track.
Based on the tracking system, the invention also provides an automatic driving vehicle track tracking method based on the neural network vehicle dynamics model, which comprises the following steps:
s1: establishing a neural network vehicle dynamics model; comprising the following steps:
s1.1, firstly, building a nonlinear monorail model of a vehicle; specific:
The vehicle is front wheel steering, the vehicle body coordinate system is positioned in a left-right symmetry plane of the vehicle, the origin of the mass center of the vehicle is o, the x axis is the longitudinal axis of the vehicle, the y axis is the lateral direction of the vehicle, the z axis meets the right hand rule, is vertical to the oxy upward, and according to Newton's law, a stress balance equation of the vehicle on the y axis and around the z axis is obtained, and the nonlinear vehicle dynamics model can be represented as follows by differential equations:
Where m is the vehicle mass, v x and v y are the longitudinal and lateral accelerations, respectively, of the center of mass in the vehicle body coordinate system, I z is the moment of inertia of the vehicle about the z-axis, l f and l r are the distances of the center of mass of the vehicle from the front and rear axes, F xf and F xr are the resultant of the tire longitudinal forces acting on the front and rear axes, F yf and F yr are the resultant of the tire lateral forces acting on the front and rear axes, respectively, r is the yaw rate of the vehicle, Is the first derivative of the yaw rate of the vehicle,/>Delta f is the front wheel corner, which is the first derivative of the lateral speed of the vehicle;
The nonlinear characteristics generated in the running process of the vehicle under different road conditions are caused by the fact that the tire turns, so that a Fiala model of the tire is introduced, and the calculation formula of the tire lateral force F y is as follows:
Wherein alpha is the cornering angle of the tire, C α is the cornering stiffness of the tire, u is the friction coefficient between the tire and the ground, and F z is the resultant force of the vertical forces of the tire;
F zf、Fzr is the vertical load of the front wheel and the vertical load of the rear wheel under the condition that the vehicle ignores the transverse load displacement and the longitudinal load displacement;
S1.2: determining that the input of the feedforward neural network model is the yaw rate r, the lateral speed v y, the longitudinal speed v x and the front wheel rotation angle delta f of the vehicle and the output of the model is the first derivative of the yaw rate based on the nonlinear monorail model of the vehicle First derivative of longitudinal speed/>
The neural network vehicle dynamics model specifically adopts the following structure: the first layer is an input layer, and the input layer has 8 characteristic inputs, namely a yaw rate r t, a lateral rate v y,t, a longitudinal rate v x,t, a front wheel rotation angle delta f,t, a yaw rate r t-1 at the last moment, a lateral rate v y,t-1, a longitudinal rate v x,t-1 and a front wheel rotation angle delta f,t-1 at the current moment; the second layer is an FC1 full-connection network layer, and the hidden layer is designed to have 100 hidden units; the third layer is an activation layer, and the activation function is selected as Softplus functions; the fourth layer is an FC2 full-connection network layer, and the hidden layer is designed to have 100 hidden units; the fifth layer is an activation layer, and the activation function is selected as Softplus functions; the sixth layer is the output layer, designed with 2 neurons, output is the first derivative of yaw rate at the current momentFirst derivative of lateral speed of vehicle/>
The forward calculation method of the designed neural network vehicle dynamics model is as follows:
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
θ＝(w1,b1,w2,b2,w3,b3)
Where x t is the vehicle state information of a single time step, h t contains the vehicle state information of the current time and the last time, a 1,a2 is the active layer Softplus function expression, θ is the parameter learned by the network, w 1,b1,w2,b2,w3,b3 is the weight and bias of the middle layer of the network, z 1 is the output of the hidden layer of the first layer of the network, and z 2 is the output of the hidden layer of the second layer of the network. And/>The definition is as follows: /(I)And/>Δt=0.03s is the sampling frequency of the data;
S2, acquiring vehicle dynamics data, including simulation data acquisition based on a driving simulator and a virtual simulation platform CarSim and real-world automatic driving vehicle data acquisition;
Establishing a real-time virtual simulation platform with a CarSim through a driving simulator, selecting an automatic driving test map Mcity, collecting data based on normal driving behaviors of human beings, and performing single-lane and double-lane conversion on different roads including straight roads, curved roads and the like for collecting complete data; in real world vehicle data acquisition, the autonomous vehicle is controlled by a human driver to perform linear motion, curvilinear motion, single lane change and double lane change;
S3, training a neural network vehicle dynamics model;
The obtained virtual simulation platform data set and real-world real vehicle data are combined and then divided into 80% of training set, 10% of verification set and 10% of test set, the loss function is set as MSE loss function, the optimizer is set as Adam, the batch size is set as 1000, the learning rate is set as 0.0003, and the network model is trained based on Pytorch deep learning frames;
The MSE loss function is specifically designed as follows:
Where r, v y are the measured yaw rate and lateral rate of the vehicle respectively, The first derivative of the yaw rate and the lateral speed of the vehicle predicted by the neural network vehicle dynamics model, delta t is the sampling time, N is the number of samples, and the first derivative/>, of the yaw rate and the lateral speed of the vehicle predicted by the neural network vehicle dynamics model is utilizedEuler integration is carried out on the yaw rate and transverse speed measured values r, v y obtained by the CarSim software, and the predicted value/>, of the yaw rate and transverse speed at the next moment is obtainedH t contains vehicle state information of the current moment and the last moment, and θ is a parameter learned by the network;
s4, designing a model predictive control algorithm; and obtaining the optimal front wheel steering angle through online solution of rolling optimization, and realizing tracking of the reference track.
Further, the model predictive control algorithm is specifically designed as follows:
Based on a neural network vehicle dynamics model, establishing an automatic driving vehicle path tracking system model to be expressed as
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
Where x t is the vehicle state information for a single time step, h t contains the vehicle state information for the current and last time, f NN is the established neural network vehicle dynamics model,V x and v y are the longitudinal acceleration and lateral acceleration of the centroid in the body coordinate system, r is the yaw rate of the vehicle, delta f is the front wheel steering angle,/>, respectively, for the heading angle of the vehicleIs the first derivative of the yaw rate of the vehicle,/>Is the first derivative of the lateral speed of the vehicle,/>Is the first derivative of the heading angle of the vehicle,/>AndThe first derivatives of the longitudinal displacement and lateral displacement of the vehicle, respectively;
Yaw rate r, lateral rate v y, longitudinal rate v x, longitudinal displacement X and lateral displacement Y, heading angle As state variables of the system, i.e./>Front wheel angle δ f is used as a control variable of the system, namely u= [ delta f ], and the input/>, of the system
Discretizing the described automatic driving vehicle track model by using an Euler formula to obtain a discrete dynamics model of the system as follows
y(k)＝C·S(k)
In the matrixFor the sampling time, T S is the sampling time, T S is the same as the virtual data sampling time, T S =Δt=0.03s, S (k-1) is the state of the last time of the system, S (k) is the state of the current time of the system, F is the established track tracking system model, F NN is the established neural network vehicle dynamics model, r (k), v y(k),vx(k),δf (k) are respectively the yaw rate, lateral speed, longitudinal speed, front wheel rotation angle, r (k-1), v y(k-1),vx(k-1),δf (k-1) are respectively the yaw rate, lateral speed, longitudinal speed, front wheel rotation angle,/>, of the vehicle at the time k-1 before the current sampling time kThe first derivative of the longitudinal position, the first derivative of the transverse position and the first derivative of the course angle of the vehicle at the moment k are respectively;
Defining the prediction time domain of the automatic driving vehicle track model as p, controlling the time domain as c, wherein p is more than or equal to c, and the dynamic state of the vehicle in the [ k+1, k+p ] prediction time domain can be obtained based on the current state of the vehicle, the state of the last moment and the prediction model, namely, at the moment k+p, the state of the vehicle is
At the kth sampling time, the optimal input sequence of the system is obtained as follows
U(K)＝[u(k|k),u(k+1|k),…,u(k+p-1|k)]T
At the kth sampling instant, the predicted output of the system is
Y(K)＝[y(k|k),y(k+1|k),…,y(k+p|k)]T
At the kth sampling time, the reference input sequence of the system is that
R(K)＝[rref(k|k),rref(k+1|k),…,rref(k+p|k)]T
At the kth sampling moment, y (k) is taken as an initial value of the prediction of the control system, namely y (k|k) =y (k), the controller predicts the output of the system in a period of time in the future through a prediction model, the controller is designed to hopefully reach an optimal performance index, the control output is obtained by solving an optimal control problem with constraint, and the prediction output is corrected according to the system output in the next period to complete the control period;
In the design process of the model predictive control algorithm, in order to keep good track tracking of the autonomous vehicle, the expected outputs on the input tracking of the system, namely the longitudinal displacement X, the lateral displacement Y and the course angle of the system, are required to be made in consideration of the tracking performance and the comfort of the vehicle Tracking the desired lateral displacement X ref, longitudinal displacement Y ref and heading angle/>The control targets are:
Wherein Q 1,Q2,Q3 is the weight in the optimization target, and increasing Q 1,Q2 can improve the path tracking performance;
in order to reduce the rate of change of the control actions to ensure the comfort of the passengers, the control targets are:
wherein M is the weight of the optimization target, and the weight coefficient can be adjusted according to the requirement;
Obtaining a total optimized objective function:
further constraints on control quantities should be considered in the MPC solution process
umin≤δf≤umax
Δumin≤Δδf≤Δumax
Wherein u min,umax is the minimum value and the maximum value of the front wheel rotation angle obtained in the MPC solving process, and Deltau min,Δumax is the minimum change rate and the maximum change rate of the front wheel rotation angle obtained in the MPC solving process;
for control stability of the vehicle body, a constraint is imposed on the first derivative of yaw rate
In the middle ofMu is the friction coefficient of the road, g is the gravitational acceleration, v x is the longitudinal speed, which is the first derivative of the yaw rate.
The beneficial effects of the invention are as follows:
1. the invention provides a data acquisition method based on a driving simulator and a CarSim real-time virtual simulation platform, which lays a data foundation for the establishment of a vehicle dynamics model. The data of the real-time response of the vehicle dynamics is obtained by the driver using a driving simulator to maneuver the vehicle in the high-fidelity vehicle dynamics software CarSim. And the freedom degree selection range of the vehicle dynamics simulation data is wide, and the data acquisition cost is reduced.
2. The invention provides a neural network vehicle dynamics prediction model designed by a base and feedforward neural network, which consists of a simple four-layer feedforward neural network, compared with a deep neural network, the calculation cost is reduced, various complex dynamics behaviors in the running process of the vehicle can be accurately identified, and the dynamic response of the unmodeled vehicle, such as the vertical motion of the vehicle during high-speed running, the suspension motion characteristic and the longitudinal and transverse coupling relation of the force of a tire, can be learned.
3. By combining the established neural network vehicle dynamics prediction model with the model prediction control algorithm, compared with an end-to-end control algorithm, the control algorithm provided by the invention has higher interpretability. And the method can realize the tracking control of the expected track under different road conditions and running conditions, ensure the tracking precision of the path and simultaneously give consideration to the transverse and longitudinal stability, and lay a foundation for developing a high-performance motion controller of an automatic driving vehicle.
Drawings
FIG. 1 is a flow chart for autonomous vehicle trajectory tracking based on a neural network vehicle dynamics model;
FIG. 2 is a non-linear monorail model of an autonomous vehicle;
FIG. 3 is a model of vehicle dynamics prediction based on a feedforward neural network;
FIG. 4 is a vehicle dynamics data acquisition module;
FIG. 5 is a training structure diagram of a vehicle dynamics prediction model based on a feedforward feedback neural network;
FIG. 6 is a flowchart of an autonomous vehicle trajectory tracking control algorithm based on neural network vehicle dynamics model predictive control.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
FIG. 1 is a flow chart of an intelligent vehicle track following algorithm based on neural network vehicle dynamics model predictive control, including model training and track following based on model predictive control, specifically as follows:
model training: and (3) acquiring data and a real vehicle data through a driving simulator and a CarSim simulation platform. A vehicle dynamics prediction model is designed based on the feedforward neural network and the model is trained by using the acquired data.
Trajectory tracking based on model predictive control: and designing a model predictive control algorithm by utilizing the trained neural network vehicle dynamics model, and obtaining the optimal front wheel turning angle through online solving through rolling optimization so as to realize tracking control of the reference track.
FIG. 2 is a non-linear monorail model of a vehicle. The nonlinear monorail model of the vehicle makes the following idealized assumptions:
(1) Assuming that the vehicle is traveling on a flat road, only the lateral and longitudinal movements of the vehicle are considered, ignoring the vertical movement of the vehicle.
(2) Assuming that the suspension system of the vehicle is a rigid body, the motion of the suspension and its effect on the coupling relationship are ignored.
(3) The coupling relationship of the lateral and longitudinal tire forces of the vehicle is ignored.
(4) The lateral load displacement and the longitudinal load displacement of the vehicle are ignored.
(5) Ignoring the effect of the tread on the turning radius, a bicycle model is used to describe the movement of the vehicle.
(6) The influence of the air resistance on the yaw characteristics of the vehicle is not considered.
Based on the above assumption, the vehicle has motion only in the x-o-y plane. The vehicle is front wheel steering, the vehicle body coordinate system is positioned in a left-right symmetry plane of the vehicle, the origin of the mass center of the vehicle is o, the x axis is the longitudinal axis of the vehicle, the y axis is the lateral direction of the vehicle, the z axis meets the right rule, and the z axis is vertical to the oxy direction. And obtaining a stress balance equation of the vehicle on the y axis and around the z axis according to Newton's law. The nonlinear vehicle dynamics model can be expressed as follows differential equations:
Where m is the vehicle mass, v x and v y are the longitudinal and lateral accelerations, respectively, of the center of mass in the vehicle body coordinate system, I z is the moment of inertia of the vehicle about the z-axis, l f and l r are the distances of the center of mass of the vehicle from the front and rear axes, F xf and F xr are the resultant of the tire longitudinal forces acting on the front and rear axes, F yf and F yr are the resultant of the tire lateral forces acting on the front and rear axes, respectively, r is the yaw rate of the vehicle, Is the first derivative of the yaw rate of the vehicle,/>Delta f is the front wheel steering angle, which is the first derivative of the vehicle lateral speed.
The nonlinear characteristics generated in the running process of the vehicle under different road conditions are caused by the fact that the tire turns, so that a Fiala model of the tire is introduced, and the calculation formula of the tire lateral force F y is as follows:
Where α is the tire cornering angle, C α is the tire cornering stiffness, u is the coefficient of friction between the tire and the ground, and F z is the tire vertical force resultant force.
F zf、Fzr is the vertical load of the front wheels and the vertical load of the rear wheels in the case where the vehicle ignores the lateral load displacement and the longitudinal load displacement, respectively.
Determining that the input of the feedforward neural network model is the yaw rate r, the lateral speed v y, the longitudinal speed v x and the front wheel rotation angle delta f of the vehicle and the output of the model is the first derivative of the yaw rate based on the nonlinear monorail model of the vehicleFirst derivative of longitudinal speed/>
FIG. 3 is a model of a prediction of vehicle dynamics for a base feedforward neural network. Compared with the traditional nonlinear monorail model, the model can learn the unmodeled vehicle dynamics changes, such as the vertical motion, suspension motion characteristics and longitudinal and transverse coupling relation of tire force of the vehicle when the vehicle runs at a high speed.
The neural network vehicle dynamics model specifically adopts the following structure: the first layer is an input layer, and the input layer has 8 characteristic inputs, namely a yaw rate r t at the current moment, a lateral rate v y,t, a longitudinal rate v x,t, a front wheel rotation angle delta f,t, a yaw rate r t-1 at the last moment, a lateral rate v y,t-1, a longitudinal rate v x,t-1 and a front wheel rotation angle delta f,t-1. The second layer is an FC1 fully connected network layer, and the hidden layer design has 100 hidden units. The third layer is an activation layer, and the activation function is selected as Softplus functions. The fourth layer is an FC2 fully connected network layer, and the hidden layer design has 100 hidden units. The fifth layer is the active layer, and the activation function is chosen to be Softplus functions. The sixth layer is the output layer, designed with 2 neurons, output is the first derivative of yaw rate at the current momentFirst derivative of lateral speed of vehicle/>
The forward calculation method of the designed neural network vehicle dynamics model is as follows:
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
θ＝(w1,b1,w2,b2,w3,b3)
Where x t is the vehicle state information for a single time step, h t contains the vehicle state information for the current time and the last time. a 1,α2 is an activation layer Softplus function expression, θ is a parameter learned by the network. w 1,b1,w2,b2,w3,b3 is the weight and bias of the middle layer of the network, z 1 is the output of the first hidden layer of the network, and z 2 is the output of the second hidden layer of the network. And/>The definition is as follows: /(I)And/>Where Δt=0.03 s is the sampling frequency of the data.
FIG. 4 is a vehicle dynamics data acquisition module. The vehicle parameters in the vehicle dynamics simulation software CarSim are modified for the vehicle parameters of the autonomous vehicle in the real world. And establishing a real-time simulation platform with the CarSim through a driving simulator, selecting an automatic driving test map Mcity, and collecting data based on normal driving habits of human beings. Because the road curvature of a vehicle has a great influence on the drivability of the vehicle, in order to collect complete data, the vehicle is driven on different roads, including straight roads, curved roads, and the like, and single-lane and double-lane changes are performed.
In the real-world vehicle data acquisition process, an autonomous vehicle is controlled to perform linear motion, curved motion, single-lane change, double-lane change, and the like by using a human driver.
FIG. 5 is a training structure diagram of a vehicle dynamics prediction model based on a feedforward neural network. The vehicle dynamics simulation data and the real vehicle data based on the normal driving behaviors of the human are collected by using the CarSim software for training, and the obtained data set is divided into 80% of training set, 10% of verification set and 10% of test set. The loss function was chosen as the MSE loss function, the optimizer was chosen as Adam, the batch size was set to 1000, the learning rate was set to 0.0003, the network model was trained based on Pytorch deep learning framework, and the loss function of the model was as follows:
Where r, v y are the measured yaw rate and lateral rate of the vehicle respectively, The first derivatives of the yaw rate and lateral speed of the vehicle predicted for the neural network vehicle dynamics model, Δt is the sampling time, and N is the number of samples. First derivative/>, of yaw rate and lateral rate of vehicle predicted using neural network vehicle dynamics modelEuler integration is carried out on the yaw rate and transverse speed measured values r, v y obtained by the CarSim software, and the predicted value/>, of the yaw rate and transverse speed at the next moment is obtainedH t contains vehicle state information of the current time and the last time, and θ is a parameter learned by the network.
FIG. 6 is a flowchart of an autonomous vehicle trajectory tracking control algorithm based on neural network vehicle dynamics model predictive control. Based on the established neural network vehicle dynamics model, the autonomous vehicle path tracking system model may be represented as
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
Where x t is the vehicle state information for a single time step, and h t contains the vehicle state information for the current time and the last time. And f NN is an established neural network vehicle dynamics model.V x and v y are the longitudinal acceleration and lateral acceleration of the centroid in the body coordinate system, r is the yaw rate of the vehicle, and δ f is the front wheel steering angle, respectively, for the heading angle of the vehicle. /(I)Is the first derivative of the yaw rate of the vehicle,/>Is the first derivative of the lateral speed of the vehicle,/>Is the first derivative of the heading angle of the vehicle,/>AndThe first derivatives of the longitudinal displacement and lateral displacement of the vehicle, respectively.
Yaw rate r, lateral rate v y, longitudinal rate v x, longitudinal displacement X and lateral displacement Y, heading angleAs state variables of the system, i.e./>Front wheel angle δ f is used as a control variable of the system, namely u= [ delta f ], and the input/>, of the system
Discretizing the described automatic driving vehicle track model by using an Euler formula to obtain a discrete dynamics model of the system as follows
y(k)＝C·S(k)
In the matrixThe sampling time is T S, and T S is the same as the virtual data sampling time, and T S =Δt=0.03 s. S (k-1) is the state of the last time of the system, and S (k) is the state of the current time of the system. F is the established trajectory tracking system model, and F NN is the established neural network vehicle dynamics model. r (k), v y(k),vx(k),δf (k) are respectively the yaw rate, lateral rate, longitudinal rate, front wheel rotation angle of the vehicle at the current sampling instant k. r (k-1), v y(k-1),vx(k-1),δf (k-1) are the yaw rate, lateral speed, longitudinal speed, front wheel steering angle of the vehicle at time k-1 before the current sampling time k, respectively. /(I)The first derivative of the longitudinal position, the first derivative of the transverse position and the first derivative of the heading angle of the vehicle at the moment k are respectively obtained.
According to the invention, the prediction time domain of the automatic driving vehicle track model is defined as p, the control time domain is defined as c, and p is more than or equal to c. The dynamics of the vehicle in the [ k+1, k+p ] prediction domain can be obtained based on the current state of the vehicle, the last time state and the prediction model. Namely at time k+p, the state of the vehicle is
Thus, at the kth sampling instant, the optimal input sequence for the system is available as
U(K)＝[u(k|k),u(k+1|k),…,u(k+p-1|k)]T
At the kth sampling instant, the predicted output of the system is
Y(K)＝[y(k|k),y(k+1|k),…,y(k+p|k)]T
At the kth sampling time, the reference input sequence of the system is that
R(K)＝[rref(k|k),rref(k+1|k),…,rref(k+p|k)]T
At the kth sampling instant, y (k) is taken as an initial value for control system prediction, i.e., y (k|k) =y (k). The controller predicts the output of the system in a period of time in the future through a prediction model, designs the optimal performance index which the controller hopes to reach, obtains the control output by solving the optimal control problem with constraint, corrects the prediction output according to the system output in the next period, and completes the control period.
In the process of designing a model predictive control algorithm, the performance of tracking, comfort and the like of the vehicle are considered. To maintain good trajectory tracking of an autonomous vehicle, it is necessary to have the system inputs track the desired outputs, namely the system output longitudinal displacement X, lateral displacement Y and heading angleTracking desired lateral displacement X ref, longitudinal displacement Y ref and heading angleThe control targets are as follows:
where Q 1,Q2,Q3 is a weight in the optimization objective, increasing σ 1,Q2 can significantly improve the path tracking performance.
In order to reduce the rate of change of the control actions so as to ensure the comfort of passengers, the control targets are as follows:
Wherein M is the weight of the optimization target, and the weight coefficient can be adjusted according to the requirement.
To sum up, a total optimized objective function is obtained, i.e
Further constraints on control quantities should be considered in the MPC solution process
umin≤δf≤umax
Δumin≤Δδf≤Δumax
Wherein u min,umax is the minimum value and the maximum value of the front wheel rotation angle obtained in the MPC solving process respectively. Δu min,Δumax is the minimum change rate and the maximum change rate of the front wheel corner obtained in the MPC solving process respectively. For control stability of the vehicle body, a constraint is imposed on the first derivative of yaw rate
In the middle ofMu is the friction coefficient of the road, g is the gravitational acceleration, v x is the longitudinal speed, which is the first derivative of the yaw rate.
Therefore, the required optimization problem with constraint is established, a joint simulation model is established through CarSim and Matlab/Simulink, and a nonlinear optimization solver fmincon is applied to solve an optimization equation on line, so that the control quantity is obtained.
The above list of detailed descriptions is only specific to practical embodiments of the present invention, and they are not intended to limit the scope of the present invention, and all equivalent manners or modifications that do not depart from the technical scope of the present invention should be included in the scope of the present invention.
Claims (8)
1. An automatic driving vehicle track tracking system based on a neural network dynamics model is characterized by comprising a neural network vehicle dynamics model part, a vehicle dynamics data acquisition part, a training part of the neural network model and a model predictive control algorithm part;
The neural network vehicle dynamics model portion: designing a neural network model with delay input by using a feedforward neural network, wherein the hidden layer of the model is two layers, each layer has 100 neurons, an activation layer selects Softplus activation functions, the input of the model adopts vehicle control and state information at two moments, and then the first derivatives of the yaw rate and the lateral speed of the vehicle are predicted;
The vehicle dynamics data acquisition section: the method comprises the steps of obtaining simulation data based on a driving simulator and a virtual simulation platform CarSim and obtaining real-world automatic driving vehicle data, establishing a real-time virtual simulation platform through the driving simulator and the CarSim, selecting an automatic driving test map Mcity, collecting data based on normal driving behaviors of human beings, and driving the vehicle to run on different roads for collecting complete data, including straight-line roads and curved roads, and performing single-lane change and double-lane change; in real world vehicle data acquisition, the autonomous vehicle is controlled by a human driver to perform linear motion, curvilinear motion, single lane change and double lane change;
Training part of the neural network model: the obtained virtual simulation platform data set and real-world real vehicle data are combined and then divided into 80% of training set, 10% of verification set and 10% of test set, the loss function is set as MSE loss function, the optimizer is set as Adam, the batch size is set as 1000, the learning rate is set as 0.0003, and the network model is trained based on Pytorch deep learning frames;
The model predictive control algorithm part: obtaining an optimal front wheel steering angle through online solution of rolling optimization, and tracking a reference track;
The neural network vehicle dynamics model part is used for determining that the input of the feedforward neural network model is the yaw rate r, the lateral speed v y, the longitudinal speed v x and the front wheel rotation angle delta f of the vehicle based on the vehicle nonlinear monorail model, and the output of the model is the first derivative of the yaw rate First derivative of lateral velocity/>
The nonlinear monorail model of the vehicle comprises the following steps: the vehicle is set as front wheel steering, the vehicle body coordinate system is positioned in a left-right symmetry plane of the vehicle, the origin of the mass center of the vehicle is o, the x axis is the longitudinal axis of the vehicle, the y axis is the lateral direction of the vehicle, the z axis meets the right rule, and the z axis is vertical to the oxy upward; according to Newton's law, a stress balance equation of the vehicle on the y axis and around the z axis is obtained, and a nonlinear vehicle dynamics model is expressed as the following differential equation:
Where m is the vehicle mass, I z is the moment of inertia of the vehicle about the z-axis, l f and l r are the distances from the vehicle center of mass to the front and rear axles, respectively, F xf and F xr are the resultant of the tire longitudinal forces acting on the front and rear axles, respectively, F yf and F yr are the resultant of the tire lateral forces acting on the front and rear axles, respectively, r is the yaw rate of the vehicle, Is the first derivative of the yaw rate of the vehicle,/>Delta f is the front wheel corner, which is the first derivative of the lateral speed of the vehicle;
Introducing Fiala model of the tire, the calculation formula of the tire side force F y is as follows:
Wherein alpha is the cornering angle of the tire, C α is the cornering stiffness of the tire, u is the friction coefficient between the tire and the ground, and F z is the resultant force of the vertical forces of the tire;
F zf、Fzr is the vertical load of the front wheels and the vertical load of the rear wheels in the case where the vehicle ignores the lateral load displacement and the longitudinal load displacement, respectively.
2. The automatic driving vehicle track tracking system based on the neural network dynamics model according to claim 1, wherein the specific structure of the neural network vehicle dynamics model is as follows:
The first layer is an input layer, and the input layer has 8 characteristic inputs, namely a yaw rate r t, a lateral rate v y,t, a longitudinal rate v x,t, a front wheel rotation angle delta f,t, a yaw rate r t-1 at the last moment, a lateral rate v y,t-1, a longitudinal rate v x,t-1 and a front wheel rotation angle delta f,t-1 at the current moment;
The second layer is an FC1 full-connection network layer, and the hidden layer is designed to have 100 hidden units;
The third layer is an activation layer, and the activation function is selected as Softplus functions;
the fourth layer is an FC2 full-connection network layer, and the hidden layer is designed to have 100 hidden units;
The fifth layer is an activation layer, and the activation function is selected as Softplus functions;
the sixth layer is the output layer, designed with 2 neurons, output is the first derivative of yaw rate at the current moment First derivative of lateral speed of vehicle/>
3. An automatic driving vehicle track tracking system based on a neural network dynamics model according to claim 2, wherein the forward calculation method of the neural network vehicle dynamics model is as follows:
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
θ＝(w1,b1,w2,b2,w3,b3)
Wherein x t is vehicle state information of a single time step, h t contains vehicle state information of the current time and the last time, a 1,a2 is an activation layer Softplus function expression, θ is a parameter learned by a network, w 1,b1,w2,b2,w3,b3 is weight and bias of a middle layer of the network, z 1 is output of a hidden layer of a first layer of the network, and z 2 is output of a hidden layer of a second layer of the network; And The definition is as follows: /(I)And/>Δt=0.03 s is the sampling frequency of the data.
4. An autonomous vehicle trajectory tracking system based on a neural network dynamics model according to claim 1, characterized in that the training part of the neural network model designs the MSE loss function as follows:
Where r, v y are the measured yaw rate and lateral rate of the vehicle respectively, The first derivative of the yaw rate and the lateral speed of the vehicle predicted by the neural network vehicle dynamics model, delta t is the sampling time, N is the number of samples, and the first derivative/>, of the yaw rate and the lateral speed of the vehicle predicted by the neural network vehicle dynamics model is utilizedEuler integration is carried out on the yaw rate and transverse speed measured values r, v y obtained by the CarSim software, and the predicted value/>, of the yaw rate and transverse speed at the next moment is obtainedH t contains vehicle state information of the current time and the last time, and θ is a parameter learned by the network.
5. An automated driving vehicle trajectory tracking system based on a neural network dynamics model according to claim 1, wherein the model predictive control algorithm section designs an automated driving vehicle trajectory tracking system model based on the established neural network vehicle dynamics model, expressed as
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
Where x t is the vehicle state information for a single time step, h t contains the vehicle state information for the current and last time, f NN is the established neural network vehicle dynamics model,For the heading angle of the vehicle, r is the yaw rate of the vehicle, delta f is the front wheel angle,/>Is the first derivative of the yaw rate of the vehicle,/>Is the first derivative of the lateral speed of the vehicle,/>Is the first derivative of the heading angle of the vehicle,/>And/>The first derivatives of the longitudinal displacement and lateral displacement of the vehicle, respectively;
Yaw rate r, lateral rate v y, longitudinal rate v x, longitudinal displacement X and lateral displacement Y, heading angle As state variables of the system, i.e./>Front wheel angle δ f is used as a control variable of the system, namely u= [ delta f ], and the input/>, of the system
Discretizing the described automatic driving vehicle track model by using an Euler formula to obtain a discrete dynamics model of the system as follows
y(k)＝C·S(k)
In the matrixK is sampling time, T S is sampling time, T S is the same as the virtual data sampling time, T S =Δt=0.03 s; s (k-1) is the state of the system at the last moment, S (k) is the state of the system at the current moment, F is the established track tracking system model, F NN is the established neural network vehicle dynamics model, r (k), v y(k),vx(k),δf (k) are the yaw rate, lateral velocity, longitudinal velocity, front wheel rotation angle, r (k-1), v y(k-1),vx(k-1),δf (k-1) are the yaw rate, lateral velocity, longitudinal velocity, front wheel rotation angle, v/O > of the vehicle at the moment k-1 before the current sampling moment k, respectivelyThe first derivative of the longitudinal position, the first derivative of the transverse position and the first derivative of the course angle of the vehicle at the moment k are respectively;
Defining the prediction time domain of the automatic driving vehicle track model as p, controlling the time domain as c, wherein p is more than or equal to c, and the dynamic state of the vehicle in the [ k+1, k+p ] prediction time domain is obtained based on the current state of the vehicle, the state at the last moment and the prediction model, namely, at the moment k+p, the state of the vehicle is
At the kth sampling time, the optimal input sequence of the system is obtained as follows
U(K)＝[u(k|k),u(k+1|k),…,u(k+p-1|k)]T
At the kth sampling instant, the predicted output of the system is
Y(K)＝[y(k|k),y(k+1|k),…,y(k+p|k)]T
At the kth sampling time, the reference input sequence of the system is that
R(K)＝[rref(k|k),rref(k+1|k),…,rref(k+p|k)]T
At the kth sampling time, y (k) is taken as an initial value predicted by the control system, namely y (k|k) =y (k); the control system predicts the output of the system in a period of time in the future through a prediction model, designs the optimal performance index which the control system hopes to reach, obtains the control output by solving the optimal control problem with constraint, corrects the prediction output according to the system output in the next period, and completes the control period.
6. An autonomous vehicle trajectory tracking system based on neural network dynamics model as claimed in claim 5, characterized in that in said model predictive control algorithm design, in order to keep the autonomous vehicle well-tracked, it is necessary to make the system inputs track the desired outputs, namely the system output longitudinal displacement X, lateral displacement Y and heading angleTracking the desired lateral displacement X ref, longitudinal displacement Y ref and heading angle/>The control targets are as follows:
wherein Q 1,Q2,Q3 is the weight in the optimization target, and increasing Q 1,Q2 can improve the path tracking performance;
In order to reduce the change rate of the control action and ensure the comfort of passengers, the control targets are as follows:
wherein M is the weight of the optimization target, and the weight coefficient is adjusted according to the requirement;
to sum up, a total optimized objective function is obtained, i.e
Further constraints on control quantities should be considered in the MPC solution process
umin≤δf≤umax
Δumin≤Δδf≤Δumax
Wherein u min,umax is the minimum value and the maximum value of the front wheel rotation angle obtained in the MPC solving process, and Deltau min,Δumax is the minimum change rate and the maximum change rate of the front wheel rotation angle obtained in the MPC solving process; for control stability of the vehicle body, a constraint is imposed on the first derivative of yaw rate
In the middle ofMu is the friction coefficient of the road, g is the gravitational acceleration, v x is the longitudinal speed, which is the first derivative of the yaw rate.
7. An automatic driving vehicle track tracking method based on a neural network dynamics model is characterized by comprising the following steps:
s1: establishing a neural network vehicle dynamics model; comprising the following steps:
s1.1, firstly, building a nonlinear monorail model of a vehicle; specific:
The vehicle is front wheel steering, the vehicle body coordinate system is positioned in a left-right symmetry plane of the vehicle, the origin of the mass center of the vehicle is o, the x axis is the longitudinal axis of the vehicle, the y axis is the lateral direction of the vehicle, the z axis meets the right rule, is vertical to the oxy upward, a stress balance equation of the vehicle around the z axis is obtained according to Newton's law, and a nonlinear vehicle dynamics model is represented by the following differential equation:
Where m is the vehicle mass, v x and v y are the longitudinal and lateral velocities of the centroid in the vehicle body coordinate system, respectively, I z is the moment of inertia of the vehicle about the z-axis, l f and l r are the distances of the vehicle centroid to the front and rear axes, respectively, F xf and F xr are the resultant of the tire longitudinal forces acting on the front and rear axes, F yf and F yr are the resultant of the tire lateral forces acting on the front and rear axes, respectively, r is the yaw rate of the vehicle, Is the first derivative of the yaw rate of the vehicle,/>Delta f is the front wheel corner, which is the first derivative of the lateral speed of the vehicle;
The nonlinear characteristics generated in the running process of the vehicle under different road conditions are caused by the fact that the tire turns, so that a Fiala model of the tire is introduced, and the calculation formula of the tire lateral force F y is as follows:
Wherein alpha is the cornering angle of the tire, C α is the cornering stiffness of the tire, u is the friction coefficient between the tire and the ground, and F z is the resultant force of the vertical forces of the tire;
F zf、Fzr is the vertical load of the front wheel and the vertical load of the rear wheel under the condition that the vehicle ignores the transverse load displacement and the longitudinal load displacement;
S1.2: determining that the input of the feedforward neural network model is the yaw rate r, the lateral speed v y, the longitudinal speed v x and the front wheel rotation angle delta f of the vehicle and the output of the model is the first derivative of the yaw rate based on the nonlinear monorail model of the vehicle First derivative of lateral velocity/>
The neural network vehicle dynamics model specifically adopts the following structure: the first layer is an input layer, and the input layer has 8 characteristic inputs, namely a yaw rate r t, a lateral rate v y,t, a longitudinal rate v x,t, a front wheel rotation angle delta f,t, a yaw rate r t-1 at the last moment, a lateral rate v y,t-1, a longitudinal rate v x,t-1 and a front wheel rotation angle delta f,t-1 at the current moment; the second layer is an FC1 full-connection network layer, and the hidden layer is designed to have 100 hidden units; the third layer is an activation layer, and the activation function is selected as Softplus functions; the fourth layer is an FC2 full-connection network layer, and the hidden layer is designed to have 100 hidden units; the fifth layer is an activation layer, and the activation function is selected as Softplus functions; the sixth layer is the output layer, designed with 2 neurons, output is the first derivative of yaw rate at the current momentFirst derivative of lateral speed of vehicle/>
The forward calculation method of the designed neural network vehicle dynamics model is as follows:
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
θ＝(w1,b1,w2,b2,w3,b3)
Wherein x t is vehicle state information of a single time step, h t contains vehicle state information of the current time and the last time, a 1,a2 is an activation layer Softplus function expression, θ is a parameter learned by a network, w 1,b1,w2,b2,w3,b3 is weight and bias of a middle layer of the network, z 1 is output of a hidden layer of a first layer of the network, and z 2 is output of a hidden layer of a second layer of the network; And The definition is as follows: /(I)And/>Δt=0.03 s is the sampling frequency of the data;
S2, acquiring vehicle dynamics data, including simulation data acquisition based on a driving simulator and a virtual simulation platform CarSim and real-world automatic driving vehicle data acquisition;
Establishing a real-time virtual simulation platform with a CarSim through a driving simulator, selecting an automatic driving test map Mcity, collecting data based on normal driving behaviors of human beings, and driving a vehicle to run on different roads, including a straight road and a curved road, for collecting complete data, and performing single-lane change and double-lane change; in real world vehicle data acquisition, the autonomous vehicle is controlled by a human driver to perform linear motion, curvilinear motion, single lane change and double lane change;
S3, training a neural network vehicle dynamics model;
The obtained virtual simulation platform data set and real-world real vehicle data are combined and then divided into 80% of training set, 10% of verification set and 10% of test set, the loss function is set as MSE loss function, the optimizer is set as Adam, the batch size is set as 1000, the learning rate is set as 0.0003, and the network model is trained based on Pytorch deep learning frames;
The MSE loss function is specifically designed as follows:
Where r, v y are the measured yaw rate and lateral rate of the vehicle respectively, The first derivative of the yaw rate and the lateral speed of the vehicle predicted by the neural network vehicle dynamics model, delta t is the sampling time, N is the number of samples, and the first derivative/>, of the yaw rate and the lateral speed of the vehicle predicted by the neural network vehicle dynamics model is utilizedEuler integration is carried out on the yaw rate and transverse speed measured values r, v y obtained by the CarSim software, and the predicted value/>, of the yaw rate and transverse speed at the next moment is obtainedH t contains vehicle state information of the current moment and the last moment, and θ is a parameter learned by the network;
s4, designing a model predictive control algorithm; and obtaining the optimal front wheel steering angle through online solution of rolling optimization, and realizing tracking of the reference track.
8. The method for tracking the track of the automatic driving vehicle based on the neural network dynamics model according to claim 7, wherein the model predictive control algorithm is specifically designed as follows:
Based on a neural network vehicle dynamics model, establishing an automatic driving vehicle path tracking system model to be expressed as
xt＝(r,vy,vx,δf)
ht＝[xt,xt-1]
Where x t is the vehicle state information for a single time step, h t contains the vehicle state information for the current and last time, f NN is the established neural network vehicle dynamics model,V x and v y are the longitudinal acceleration and lateral acceleration of the centroid in the body coordinate system, r is the yaw rate of the vehicle, delta f is the front wheel steering angle,/>, respectively, for the heading angle of the vehicleIs the first derivative of the yaw rate of the vehicle,/>Is the first derivative/>, of the lateral speed of the vehicleIs the first derivative of the heading angle of the vehicle,/>And/>The first derivatives of the longitudinal displacement and lateral displacement of the vehicle, respectively;
Yaw rate r, lateral rate v y, longitudinal rate v x, longitudinal displacement X and lateral displacement Y, heading angle As state variables of the system, i.e./>Front wheel angle δ f is used as a control variable of the system, namely u= [ delta f ], and the input/>, of the system
Discretizing the described automatic driving vehicle track model by using an Euler formula to obtain a discrete dynamics model of the system as follows
y(k)＝C·S(k)
In the matrixK is sampling time, T S is sampling time, T S is the same as virtual data sampling time, T S =Δt=0.03s, S (k-1) is the state of the last time of the system, S (k) is the state of the current time of the system, F is the established track tracking system model, F NN is the established neural network vehicle dynamics model, r (k), v y(k),vx(k),δf (k) are respectively the yaw rate, lateral speed, longitudinal speed, front wheel rotation angle, r (k-1), v y(k-1),vx(k-1),δf (k-1) are respectively the yaw rate, lateral speed, longitudinal speed, front wheel rotation angle,/>, of the vehicle at time k-1 before the current sampling time kThe first derivative of the longitudinal position, the first derivative of the transverse position and the first derivative of the course angle of the vehicle at the moment k are respectively;
Defining the prediction time domain of the automatic driving vehicle track model as p, controlling the time domain as c, wherein p is more than or equal to c, and the dynamic state of the vehicle in the [ k+1, k+p ] prediction time domain is obtained based on the current state of the vehicle, the state at the last moment and the prediction model, namely, at the moment k+p, the state of the vehicle is
At the kth sampling time, the optimal input sequence of the system is obtained as follows
U(K)＝[u(k|k),u(k+1|k),…,u(k+p-1|k)]T
At the kth sampling instant, the predicted output of the system is
Y(K)＝[y(k|k),y(k+1|k),…,y(k+p|k)]T
At the kth sampling time, the reference input sequence of the system is that
R(K)＝[rref(k|k),rref(k+1|k),…,rref(k+p|k)]T
At the kth sampling moment, y (k) is taken as an initial value of the prediction of the control system, namely y (k|k) =y (k), the control system predicts the output of the system in a period of time in the future through a prediction model, the control system is designed to achieve an optimal performance index, the control output is obtained by solving an optimal control problem with constraint, and the prediction output is corrected according to the system output in the next period to complete the control period;
In the design process of the model predictive control algorithm, in order to keep good track tracking of the autonomous vehicle, the expected outputs on the input tracking of the system, namely the longitudinal displacement X, the lateral displacement Y and the course angle of the system, are required to be made in consideration of the tracking performance and the comfort of the vehicle Tracking the desired lateral displacement X ref, longitudinal displacement Y ref and heading angle/>The control targets are:
Wherein Q 1,Q2,Q3 is the weight in the optimization target, and increasing Q 1,Q2 improves the path tracking performance;
in order to reduce the rate of change of the control actions to ensure the comfort of the passengers, the control targets are:
wherein M is the weight of the optimization target, and the weight coefficient is adjusted according to the requirement;
Obtaining a total optimized objective function:
further constraints on control quantities should be considered in the MPC solution process
umin≤δf≤umax
Δumin≤Δδf≤Δumax
Wherein u min,umax is the minimum value and the maximum value of the front wheel rotation angle obtained in the MPC solving process, and Deltau min,Δumax is the minimum change rate and the maximum change rate of the front wheel rotation angle obtained in the MPC solving process;
for control stability of the vehicle body, a constraint is imposed on the first derivative of yaw rate
In the middle ofMu is the friction coefficient of the road, g is the gravitational acceleration, v x is the longitudinal speed, which is the first derivative of the yaw rate.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111508163.7A CN114379583B (en) | 2021-12-10 | 2021-12-10 | Automatic driving vehicle track tracking system and method based on neural network dynamics model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111508163.7A CN114379583B (en) | 2021-12-10 | 2021-12-10 | Automatic driving vehicle track tracking system and method based on neural network dynamics model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114379583A CN114379583A (en) | 2022-04-22 |
| CN114379583B true CN114379583B (en) | 2024-05-14 |
Family
ID=81196295
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111508163.7A Active CN114379583B (en) | 2021-12-10 | 2021-12-10 | Automatic driving vehicle track tracking system and method based on neural network dynamics model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114379583B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115525054B (en) * | 2022-09-20 | 2023-07-11 | 武汉理工大学 | Method and system for controlling tracking of edge path of unmanned sweeper in large industrial park |
| CN115432009B (en) * | 2022-10-09 | 2023-09-05 | 海南大学 | Automatic driving vehicle track tracking control system |
| CN116560241B (en) * | 2023-07-10 | 2023-09-15 | 北京科技大学 | Explicit circulation model predictive control track tracking method and device for articulated vehicle |
| CN117389276B (en) * | 2023-11-05 | 2024-05-28 | 理工雷科智途(北京)科技有限公司 | Unmanned vehicle driving path tracking control method based on driving risk prediction |
| CN117601857B (en) * | 2023-12-18 | 2024-05-10 | 广东工业大学 | Man-machine co-driving switching control method based on track prediction |
| CN117911414A (en) * | 2024-03-20 | 2024-04-19 | 安徽大学 | Automatic driving automobile motion control method based on reinforcement learning |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107561942A (en) * | 2017-09-12 | 2018-01-09 | 重庆邮电大学 | Intelligent vehicle track following model predictive control method based on model compensation |
| CN109795502A (en) * | 2018-09-27 | 2019-05-24 | 吉林大学 | Intelligent electric automobile path trace model predictive control method |
| CN110780674A (en) * | 2019-12-04 | 2020-02-11 | 哈尔滨理工大学 | Method for improving automatic driving track tracking control |
| CN111258323A (en) * | 2020-03-30 | 2020-06-09 | 华南理工大学 | Intelligent vehicle trajectory planning and tracking combined control method |
| CN111890951A (en) * | 2020-08-07 | 2020-11-06 | 吉林大学 | Intelligent electric automobile trajectory tracking and motion control method |
| CN112092815A (en) * | 2020-09-02 | 2020-12-18 | 北京航空航天大学 | Vehicle track changing tracking control method based on model prediction |
| CN112622903A (en) * | 2020-10-29 | 2021-04-09 | 东北大学秦皇岛分校 | Longitudinal and transverse control method for autonomous vehicle in vehicle following driving environment |
| CN113320542A (en) * | 2021-06-24 | 2021-08-31 | 厦门大学 | Tracking control method for automatic driving vehicle |
| CN113386781A (en) * | 2021-05-24 | 2021-09-14 | 江苏大学 | Intelligent vehicle trajectory tracking control method based on data-driven vehicle dynamics model |
| CN113408047A (en) * | 2021-08-05 | 2021-09-17 | 江苏大学 | Vehicle dynamics prediction model based on time-lag feedback neural network, training data acquisition method and training method |
-
2021
- 2021-12-10 CN CN202111508163.7A patent/CN114379583B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107561942A (en) * | 2017-09-12 | 2018-01-09 | 重庆邮电大学 | Intelligent vehicle track following model predictive control method based on model compensation |
| CN109795502A (en) * | 2018-09-27 | 2019-05-24 | 吉林大学 | Intelligent electric automobile path trace model predictive control method |
| CN110780674A (en) * | 2019-12-04 | 2020-02-11 | 哈尔滨理工大学 | Method for improving automatic driving track tracking control |
| CN111258323A (en) * | 2020-03-30 | 2020-06-09 | 华南理工大学 | Intelligent vehicle trajectory planning and tracking combined control method |
| CN111890951A (en) * | 2020-08-07 | 2020-11-06 | 吉林大学 | Intelligent electric automobile trajectory tracking and motion control method |
| CN112092815A (en) * | 2020-09-02 | 2020-12-18 | 北京航空航天大学 | Vehicle track changing tracking control method based on model prediction |
| CN112622903A (en) * | 2020-10-29 | 2021-04-09 | 东北大学秦皇岛分校 | Longitudinal and transverse control method for autonomous vehicle in vehicle following driving environment |
| CN113386781A (en) * | 2021-05-24 | 2021-09-14 | 江苏大学 | Intelligent vehicle trajectory tracking control method based on data-driven vehicle dynamics model |
| CN113320542A (en) * | 2021-06-24 | 2021-08-31 | 厦门大学 | Tracking control method for automatic driving vehicle |
| CN113408047A (en) * | 2021-08-05 | 2021-09-17 | 江苏大学 | Vehicle dynamics prediction model based on time-lag feedback neural network, training data acquisition method and training method |
Non-Patent Citations (3)
| Title |
|---|
| 四驱混合动力轿车转弯工况路径跟踪控制;赵治国;周良杰;王凯;;同济大学学报(自然科学版);20190524(05);107-115 * |
| 智能车辆轨迹跟踪控制器设计;柴瑞强;孙涛;田敏杰;;软件导刊;20200415(04);175-179 * |
| 自动驾驶4WS车辆路径跟踪最优控制算法仿真;郭旭东;杨世春;;计算机仿真;20200415(04);132-138 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114379583A (en) | 2022-04-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN114379583B (en) | Automatic driving vehicle track tracking system and method based on neural network dynamics model | |
| Li et al. | Shared control driver assistance system based on driving intention and situation assessment | |
| CN113386781B (en) | Intelligent vehicle track tracking control method based on data-driven vehicle dynamics model | |
| CN107015477B (en) | Vehicle route based on state feedback tracks H ∞ control method | |
| CN113320542B (en) | Tracking control method for automatic driving vehicle | |
| CN111845774A (en) | Automatic driving automobile dynamic trajectory planning and tracking method based on transverse and longitudinal coordination | |
| CN104859650B (en) | A kind of vehicle yaw stability rolling optimization control method of Multiple Time Scales | |
| CN107561942A (en) | Intelligent vehicle track following model predictive control method based on model compensation | |
| CN108227491A (en) | A kind of intelligent vehicle Trajectory Tracking Control method based on sliding formwork neural network | |
| Awad et al. | Model predictive control with fuzzy logic switching for path tracking of autonomous vehicles | |
| Cai et al. | Implementation and development of a trajectory tracking control system for intelligent vehicle | |
| CN110716562A (en) | Decision-making method for multi-lane driving of unmanned vehicle based on reinforcement learning | |
| CN108791491A (en) | Vehicle lateral tracking control method based on self-evaluation learning | |
| Mohajer et al. | Enhancing passenger comfort in autonomous vehicles through vehicle handling analysis and optimization | |
| CN104881030A (en) | Unmanned vehicle-side longitudinal coupling tracking control method based on rapid terminal sliding mode principle | |
| CN110162046A (en) | Unmanned vehicle path following method based on event trigger type model predictive control | |
| Pauca et al. | Predictive control for the lateral and longitudinal dynamics in automated vehicles | |
| Hu et al. | Probabilistic lane-change decision-making and planning for autonomous heavy vehicles | |
| CN113408047B (en) | Vehicle dynamics prediction model based on time-lag feedback neural network, training data acquisition method and training method | |
| CN115303289A (en) | Vehicle dynamics model based on depth Gaussian, training method, intelligent vehicle trajectory tracking control method and terminal equipment | |
| CN112578672A (en) | Unmanned vehicle trajectory control system based on chassis nonlinearity and trajectory control method thereof | |
| Németh et al. | Nonlinear analysis and control of a variable-geometry suspension system | |
| CN113581201A (en) | Potential field model-based collision avoidance control method and system for unmanned automobile | |
| CN114684199A (en) | Vehicle dynamics series hybrid model driven by mechanism analysis and data, intelligent automobile trajectory tracking control method and controller | |
| Bruschetta et al. | A nonlinear model predictive control based virtual driver for high performance 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |