A human-like following decision-making control method for intelligent connected vehicles
By extracting characteristics from the following data of human drivers, calibrating the following model and using model prediction control methods, the problem that intelligent connected vehicle follow-up control does not conform to human driving characteristics is solved, and smoother acceleration and deceleration and improvement are achieved. Comfortable ride.
Patent Information
- Application Number
- CN202210727720.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-06-22
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2042-06-22
AI Technical Summary
The existing intelligent connected vehicles follow-up control method cannot effectively meet the characteristics of human driving, and the acceleration and deceleration are not smooth enough, resulting in poor riding comfort.
By calibrating the following model parameters of intelligent connected vehicles from the following data of human drivers, the optimal following acceleration of the vehicle is solved by using the model prediction and control method, and anthropomorphic following decision control is achieved.
The vehicle following process has been optimized and the comfort of following the vehicle has been improved.
Smart Images

Figure CN114987473B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of automatic driving, and in particular relates to a humanoid following decision-making control method for an intelligent network-connected vehicle. Background Art
[0002] With the rapid development of computer technology and communication technology, vehicles are becoming more and more intelligent and networked. Intelligent networked vehicles can achieve collaborative perception, decision-making and control between vehicles and between vehicles and infrastructure through wireless communication with surrounding vehicles and road infrastructure to improve road traffic efficiency and vehicle driving safety. Following the vehicle is one of the most important conditions in traffic. How to enable intelligent networked vehicles to make following decisions and control as reasonably and efficiently as humans is still an urgent problem to be solved.
[0003] Existing research has proposed some effective intelligent vehicle following decision control methods. Chinese invention patent application number CN201910805281.0, entitled "A method, device and vehicle for following an autonomous vehicle", proposes a method for following an autonomous vehicle, which plans an isosceles trapezoidal acceleration / deceleration curve according to the speed relationship between the front vehicle and the vehicle, and realizes the following control of the vehicle. Chinese invention patent application number CN201910325347.6, entitled "Control system and method for following an intelligent driving vehicle", provides a method for following an intelligent driving vehicle, which obtains the corresponding following time distance and safe parking distance according to the real-time vehicle speed and the real-time traffic status of the road, and controls the acceleration and deceleration of the vehicle according to the following time distance. Although the existing method realizes a simple following control function, its following control and decision cannot meet the characteristics of human driving well, and its acceleration and deceleration are not smooth enough, and the ride comfort is poor. Summary of the invention
[0004] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide an anthropomorphic following decision-making control method for an intelligent connected vehicle, so as to solve the problem that the following control in the prior art does not conform to the driving characteristics of human beings; the method of the present invention calibrates the parameters of the following model of the intelligent connected vehicle from the following data of the existing human driver, and then solves the optimal reference trajectory of the intelligent connected vehicle when following the vehicle, and finally solves the optimal following acceleration of the vehicle based on the model predictive control method, thereby realizing anthropomorphic following decision-making control.
[0005] To achieve the above object, the technical solution adopted by the present invention is as follows:
[0006] The method for controlling a human-like vehicle following decision of an intelligent network-connected vehicle of the present invention comprises the following steps:
[0007] 1) Based on the data set of human-driven vehicle following conditions, calculate the average acceleration of all vehicles in the data set The average acceleration of all vehicles in the dataset at the start time and the average acceleration of all vehicles in the data set when the speed is above 70km / h
[0008] 2) The vehicle obtains the acceleration a of the vehicle in front through wireless communication with the vehicle in front p and the length of the vehicle in front L p ;
[0009] 3) Solve the solution obtained in step 1) and the front vehicle acceleration a obtained in step 2) p , the length of the front vehicle L p Input into the following vehicle model to solve the expected longitudinal relative position sequence [Δx d (k+1),Δx d (k+2),...,Δx d (k+N p -1)] and the expected longitudinal relative velocity sequence [Δv d (k+1),Δv d (k+2),...,Δv d (k+N p -1)],N p is the prediction time domain; k+1, k+2, k+N p -1 represents different moments;
[0010] 4) Based on the expected longitudinal relative position sequence and expected longitudinal relative speed sequence of the leading vehicle and the ego vehicle in the predicted time domain obtained in step 3), the ego vehicle's following prediction control method is used to solve the optimal acceleration and optimal front wheel steering angle of the ego vehicle at the current moment; the ego vehicle performs corresponding acceleration, deceleration and steering operations according to the optimal acceleration and optimal front wheel steering angle.
[0011] Furthermore, the data set of the human-driven vehicle following condition in step 1) adopts a public NGSIM data set, which contains actual driving position information, speed information and acceleration information of a large number of vehicles.
[0012] Furthermore, the step 1) specifically includes: averaging the accelerations of all vehicles in the data set to obtain an average acceleration The average acceleration of all vehicles in the data set at the start time is calculated to obtain the average acceleration The average acceleration is obtained by averaging the acceleration of all vehicles in the data set when the speed is above 70km / h.
[0013] Furthermore, the following vehicle model in step 3) is specifically:
[0014] The front vehicle and the vehicle at time t p The expected longitudinal relative velocity Δv d (t p ) and the desired longitudinal relative position Δx d (t p ) are:
[0015]
[0016]
[0017]
[0018] Where t0 and t p are the current time and the end time respectively; a d is the expected acceleration of the vehicle; s is the minimum safe distance between the vehicle in front and the vehicle; v and v p are the longitudinal velocities of the vehicle and the preceding vehicle respectively; x and x p are the longitudinal positions of the ego vehicle and the front vehicle respectively; VD is the following threshold, which determines whether the ego vehicle follows the front vehicle or drives freely; Δx0 is the relative distance between the front vehicle and the ego vehicle at time t0; by p Set to different moments, the expected longitudinal relative position sequence and expected longitudinal relative velocity sequence in the prediction time domain can be obtained as [Δv d (k+1),Δv d (k+2),...,Δv d (k+N p -1)]、[Δx d (k+1),Δx d (k+2),...,Δx d (k+N p -1)],N p is the prediction time domain; k+1, k+2, k+N p -1 represents a different moment.
[0019] Furthermore, the following vehicle predictive control method of the self-vehicle in step 4) is specifically as follows: establishing a vehicle kinematic model as Which expands to:
[0020]
[0021] In the formula, x s is the state quantity, u c is the control quantity, u c =[δ f ,at ] T ; x and y are the longitudinal position and lateral position of the vehicle respectively; is the yaw angle of the vehicle; a t is the acceleration of the vehicle; δ f is the front wheel turning angle of the vehicle; m is the mass of the vehicle; F xf 、F yf 、F xr and F yr are the front wheel longitudinal tire force, the front wheel lateral tire force, the rear wheel longitudinal tire force and the rear wheel lateral tire force respectively; I z is the yaw inertia of the vehicle; l f and l r are the distances between the front and rear axles of the vehicle and the center of gravity; e y is the lateral position error between the vehicle and the lane centerline; is the angle between the front of the vehicle and the centerline of the lane; Δv is the longitudinal relative speed between the front vehicle and the vehicle, Δv = v p -v,v p is the longitudinal velocity of the front vehicle; Δx is the longitudinal relative position of the front vehicle and the vehicle itself, Δx = x p -x,x p is the longitudinal position of the front vehicle; R is the radius of curvature of the driving curve of the vehicle; the above vehicle kinematic model is written as follows:
[0022]
[0023] In the formula, The vehicle kinematic model is discretized to obtain the discrete state equation:
[0024] x s (k+1)=A d x s (k)+B d u c (k)
[0025] In the formula, A d =I+A s T; B d =B s T; T is the sampling time; the output equation is:
[0026] y d (k) = C d x s (k)
[0027] In the formula,
[0028] The model predictive control problem is constructed as:
[0029]
[0030] stx s (k+i+1)=A d x s (k+i)+B d u c (k+i),i=0,1,...,N p -1
[0031] u c,min ≤u k+i ≤u c,max ,i=0,1,...,N c -1
[0032] Where N p and N c They are prediction time domain and control time domain respectively; is the control sequence; Q and R are weight matrices; ρ is the weight coefficient; ε is the relaxation factor; y ref (k+i) is the reference output at time k+i, which is obtained from the expected longitudinal relative velocity sequence and expected longitudinal relative position sequence of the preceding vehicle and the self-vehicle in the prediction time domain in step 3), y ref (k+i)=[Δv d (k+i)Δx d (k+i)] T ;u c,min and u c,max are the minimum control amount and the maximum control amount respectively; by solving the model predictive control problem, the optimal control sequence is obtained The optimal control vector at the current time k in the optimal control sequence is in, is the optimal front wheel steering angle of the vehicle at the current time k, is the optimal acceleration of the ego vehicle at the current moment k; the ego vehicle is required to perform corresponding steering and acceleration / deceleration operations according to the optimal front wheel steering angle and optimal acceleration at the current moment to realize the following vehicle decision control.
[0033] Beneficial effects of the present invention:
[0034] The present invention extracts the following characteristics of human drivers from the following behavior dataset of human drivers, calibrates the following model, and constructs a following decision control model for the vehicle based on the model predictive control method, thereby solving the optimal acceleration and front wheel turning angle of the vehicle under the following condition, optimizing the vehicle following process and improving the comfort of the vehicle following driving. BRIEF DESCRIPTION OF THE DRAWINGS
[0035] Figure 1 The figure is a flow chart of the method of the present invention.
[0036] Figure 2 Schematic diagram of the vehicle's trajectory. DETAILED DESCRIPTION
[0037] In order to facilitate the understanding of those skilled in the art, the present invention is further described below in conjunction with embodiments and drawings. The contents mentioned in the implementation modes are not intended to limit the present invention.
[0038] Reference Figure 1 , Figure 2 As shown, a humanoid following decision control method of an intelligent networked vehicle of the present invention comprises the following steps:
[0039] 1) Based on the data set of human-driven vehicle following conditions, calculate the average acceleration of all vehicles in the data set The average acceleration of all vehicles in the dataset at the start time and the average acceleration of all vehicles in the data set when the speed is above 70km / h
[0040] The data set of the human-driven vehicle following condition adopts the public NGSIM data set, which contains a large amount of actual vehicle driving position information, speed information and acceleration information.
[0041] 2) The vehicle obtains the acceleration a of the vehicle in front through wireless communication with the vehicle in front p and the length of the vehicle in front L p ;
[0042] 3) Solve the solution obtained in step 1) and the front vehicle acceleration a obtained in step 2) p , the length of the front vehicle L p Input into the following vehicle model to solve the expected longitudinal relative position sequence [Δx d (k+1),Δx d (k+2),...,Δx d (k+N p -1)] and the expected longitudinal relative velocity sequence [Δv d (k+1),Δv d (k+2),...,Δv d (k+N p -1)],N p is the prediction time domain; k+1, k+2, k+N p -1 represents different moments;
[0043] The following car model is specifically:
[0044] The front vehicle and the vehicle at time t pThe expected longitudinal relative velocity Δv d (t p ) and the desired longitudinal relative position Δx d (t p ) are:
[0045]
[0046]
[0047]
[0048] Where t0 and t p are the current time and the end time respectively; a d is the expected acceleration of the vehicle; s is the minimum safe distance between the vehicle in front and the vehicle; v and v p are the longitudinal velocities of the vehicle and the preceding vehicle respectively; x and x p are the longitudinal positions of the ego vehicle and the front vehicle respectively; VD is the following threshold, which determines whether the ego vehicle follows the front vehicle or drives freely; Δx0 is the relative distance between the front vehicle and the ego vehicle at time t0; by p Set to different moments, the expected longitudinal relative position sequence and expected longitudinal relative velocity sequence in the prediction time domain can be obtained as [Δv d (k+1),Δv d (k+2),...,Δv d (k+N p -1)]、[Δx d (k+1),Δx d (k+2),...,Δx d (k+N p -1)],N p is the prediction time domain; k+1, k+2, k+N p -1 represents a different moment.
[0049] 4) According to the expected longitudinal relative position sequence and expected longitudinal relative speed sequence of the leading vehicle and the ego vehicle in the prediction time domain obtained in step 3), the ego vehicle's following vehicle prediction control method is used to solve the optimal acceleration and optimal front wheel steering angle of the ego vehicle at the current moment; the ego vehicle performs corresponding acceleration, deceleration and steering operations according to the optimal acceleration and optimal front wheel steering angle; specifically:
[0050] The following vehicle predictive control method of the self-vehicle is specifically as follows: a vehicle kinematic model is established as Which expands to:
[0051]
[0052] In the formula, x s is the state quantity, u c is the control quantity, u c =[δ f ,a t ] T ; x and y are the longitudinal position and lateral position of the vehicle respectively; is the yaw angle of the vehicle; a t is the acceleration of the vehicle; δ f is the front wheel turning angle of the vehicle; m is the mass of the vehicle; F xf 、F yf 、F xr and F yr are the front wheel longitudinal tire force, the front wheel lateral tire force, the rear wheel longitudinal tire force and the rear wheel lateral tire force; I z is the yaw inertia of the vehicle; l f and l r are the distances between the front and rear axles of the vehicle and the center of gravity; e y is the lateral position error between the vehicle and the lane centerline; is the angle between the front of the vehicle and the centerline of the lane; Δv is the longitudinal relative speed between the front vehicle and the vehicle, Δv = v p -v,v p is the longitudinal velocity of the front vehicle; Δx is the longitudinal relative position of the front vehicle and the vehicle itself, Δx = x p -x,x p is the longitudinal position of the front vehicle; R is the radius of curvature of the driving curve of the vehicle; the above vehicle kinematic model is written as follows:
[0053]
[0054] In the formula, The vehicle kinematic model is discretized to obtain the discrete state equation:
[0055] x s (k+1)=A d x s (k)+B d u c (k)
[0056] In the formula, A d =I+A s T; B d =B s T; T is the sampling time; the output equation is:
[0057] y d (k) = C d x s (k)
[0058] In the formula,
[0059] The model predictive control problem is constructed as:
[0060]
[0061] stx s (k+i+1)=A d x s (k+i)+B d u c (k+i),i=0,1,...,N p -1
[0062] u c,min ≤u k+i ≤u c,max ,i=0,1,...,N c -1
[0063] Where N p and N c They are prediction time domain and control time domain respectively; is the control sequence; Q and R are weight matrices; ρ is the weight coefficient; ε is the relaxation factor; y ref (k+i) is the reference output at time k+i, which is obtained from the expected longitudinal relative velocity sequence and expected longitudinal relative position sequence of the preceding vehicle and the self-vehicle in the prediction time domain in step 3), y ref (k+i)=[Δv d (k+i)Δx d (k+i)] T ;u c,min and u c,max are the minimum control amount and the maximum control amount respectively; by solving the model predictive control problem, the optimal control sequence is obtained The optimal control vector at the current time k in the optimal control sequence is in, is the optimal front wheel steering angle of the vehicle at the current time k, is the optimal acceleration of the ego vehicle at the current moment k; the ego vehicle is required to perform corresponding steering and acceleration / deceleration operations according to the optimal front wheel steering angle and optimal acceleration at the current moment to realize the following vehicle decision control.
[0064] The present invention has many specific application paths. The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements can be made without departing from the principle of the present invention. These improvements should also be regarded as the protection scope of the present invention.
Claims
1. A humanoid following decision control method for an intelligent connected vehicle, characterized in that: Here are the steps: 1) Based on the data set of human-driven vehicle following conditions, calculate the average acceleration of all vehicles in the data set The average acceleration of all vehicles in the dataset at the start time and the average acceleration of all vehicles in the data set when the speed is above 70km / h 2) Obtain the acceleration and body length of the preceding vehicle; 3) Solve the solution obtained in step 1) The acceleration of the leading vehicle and the length of the leading vehicle obtained in step 2) are input into the following vehicle model to obtain the expected longitudinal relative position sequence and the expected longitudinal relative velocity sequence of the leading vehicle and the ego vehicle in the prediction time domain; 4) Adopting the vehicle following predictive control method of the ego vehicle, the optimal acceleration and optimal front wheel steering angle of the ego vehicle at the current moment are solved; the ego vehicle performs corresponding acceleration, deceleration and steering operations according to the optimal acceleration and optimal front wheel steering angle; The following car model in step 3) is specifically: The front vehicle and the vehicle at time t p The expected longitudinal relative velocity Δvd(t p ) and the desired longitudinal relative position Δxd(t p ) are: Where t0 and t p are the current time and the end time respectively; a d is the expected acceleration of the vehicle; s is the minimum safe distance between the vehicle in front and the vehicle; v and v p are the longitudinal velocities of the vehicle and the preceding vehicle respectively; x and x p are the longitudinal positions of the ego vehicle and the front vehicle respectively; VD is the following threshold, which determines whether the ego vehicle follows the front vehicle or drives freely; Δx0 is the relative distance between the front vehicle and the ego vehicle at time t0; by p Set to different moments, the expected longitudinal relative position sequence and expected longitudinal relative velocity sequence in the prediction time domain can be obtained as [Δv d (k+1),Δv d (k+2),...,Δv d (k+N p -1)]、[Δx d (k+1),Δx d (k+2),...,Δx d (k+N p -1)],N p is the prediction time domain; k+1, k+2, k+N p -1 represents different moments, a p is the acceleration of the front vehicle, L p is the length of the front vehicle; The following vehicle prediction control method of the self-vehicle in step 4) is specifically as follows: establish a vehicle kinematic model as Which expands to: In the formula, x s is the state quantity, u c is the control quantity, u c =[δ f ,a t ] T ; x and y are the longitudinal position and lateral position of the vehicle respectively; is the yaw angle of the vehicle; a t is the acceleration of the vehicle; δ f is the front wheel turning angle of the vehicle; m is the mass of the vehicle; F xf 、F yf 、F xr and F yr are the front wheel longitudinal tire force, the front wheel lateral tire force, the rear wheel longitudinal tire force and the rear wheel lateral tire force respectively; I z is the yaw inertia of the vehicle; l f and l r are the distances between the front and rear axles of the vehicle and the center of gravity; e y is the lateral position error between the vehicle and the lane centerline; is the angle between the front of the vehicle and the centerline of the lane; Δv is the longitudinal relative speed between the front vehicle and the vehicle, Δv = v p -v,v p is the longitudinal velocity of the front vehicle; Δx is the longitudinal relative position of the front vehicle and the vehicle itself, Δx = x p -x,x p is the longitudinal position of the front vehicle; R is the radius of curvature of the driving curve of the vehicle; the above vehicle kinematic model is written as follows: In the formula, The vehicle kinematic model is discretized to obtain the discrete state equation: x s (k+1)=A d x s (k)+B d u c (k) In the formula, A d =I+A s T; B d =B s T; T is the sampling time; the output equation is: y d (k)=C d x s (k) In the formula, The model predictive control problem is constructed as: s.t.x s (k+i+1)=A d x s (k+i)+B d u c (k+i),i=0,1,...,N p -1 u c,min ≤u k+i ≤u c,max ,i=0,1,...,N c -1 Where Np and Nc are the prediction time domain and the control time domain respectively; uc(t)=[uk,uk +1 ,...,uk +Nc -1] is the control sequence; Q and R are weight matrices; ρ is the weight coefficient; ε is the relaxation factor; y ref (k+i) is the reference output at time k+i, which is obtained from the expected longitudinal relative velocity sequence and expected longitudinal relative position sequence of the preceding vehicle and the self-vehicle in the prediction time domain in step 3), yref(k+i)=[Δv d (k+i)Δx d (k+i)] T uc,min and uc,max are the minimum and maximum control quantities respectively; by solving the model predictive control problem, the optimal control sequence is obtained The optimal control vector at the current time k in the optimal control sequence is in, is the optimal front wheel steering angle of the vehicle at the current time k, is the optimal acceleration of the ego vehicle at the current moment k; the ego vehicle is required to perform corresponding steering and acceleration / deceleration operations according to the optimal front wheel steering angle and optimal acceleration at the current moment to realize the following vehicle decision control.
2. The humanoid following decision control method for an intelligent connected vehicle according to claim 1, characterized in that: The data set of the human-driven vehicle following condition in step 1) adopts the public NGSIM data set.
