An ecological driving control method for intelligent connected vehicles at intersections

Abstract

The present invention relates to an ecological driving control method for an intelligent networked vehicle intersection, comprising: collecting traffic information in real time based on GPS technology and V2I communication; globally linearizing a vehicle dynamics model based on simulated sampled vehicle trajectory data; combining intersection signal timing constraints and a globally linearized vehicle dynamics model as constraints to construct an ecological driving optimization control model based on a spatial domain; based on model predictive control technology, inputting the real-time collected traffic information into the ecological driving optimization control model, and generating an optimal motion trajectory for ecological driving through iterative dynamic solution, and updating and optimizing automatic driving control instructions in real time, so that the vehicle drives according to the optimal motion trajectory for ecological driving. Compared with the prior art, the present invention can accurately and quickly make decisions and plans on the driving trajectory of intelligent networked vehicles, thereby realizing efficient and green passage of intelligent networked vehicles at urban intersections.

Description

Technical Field

[0001] The present invention relates to the technical field of traffic control for networked autonomous driving vehicles, and in particular to an ecological driving control method for intersections of intelligent networked vehicles. Background Art

[0002] The transportation sector is one of the main sources of energy consumption and pollution emissions in the world. In 2014, carbon dioxide emissions caused by the transportation industry accounted for 20.5% of the global total, second only to the electricity and heat industries. In 2016, the energy consumed by the transportation industry accounted for 26% of the total global energy demand, and this proportion is still rising. The energy consumption and carbon dioxide emissions of road traffic account for 72% and 80% of the entire transportation industry. Therefore, it is urgent to solve the problems of transportation energy consumption and pollution emissions.

[0003] The development of eco-driving assistance systems is the main response to current traffic energy consumption and pollution issues. The eco-driving assistance system is based on networked autonomous driving technology, and saves fuel consumption by optimizing driving distances, adjusting driving speeds, and reducing the number of stops. Eco-driving is mainly divided into eco-cruise driving and intersection green speed guidance. The former is suitable for scenarios without signal light control, such as highways and expressways; the latter takes into account the influence of intersection signals and coordinates the acceleration, speed and intersection signal timing of networked autonomous driving vehicles to achieve the purpose of saving fuel. With the development of vehicle-road collaborative technology, networked autonomous driving vehicles can obtain signal timing information in real time, which has also played a huge role in promoting the development of intersection green speed guidance technology.

[0004] In urban road environments, intersection signal control often interrupts the normal operation of vehicles, thus having a significant impact on their driving efficiency and energy consumption. How to plan vehicle trajectories in real time so that they can pass through intersections with low energy consumption and high efficiency is a major research focus of intelligent connected vehicle decision-making and control issues. The existing ecological driving assistance system for achieving green speed guidance at intersections mainly adopts the optimization control method, but it has the following obvious defects:

[0005] (1) Existing eco-driving technologies often only consider the longitudinal kinematic characteristics of the vehicle, mostly ignoring the lateral dynamics and kinematics, and even less considering the longitudinal and lateral coupling characteristics of vehicle operation, which leads to a mismatch between theoretical modeling and the actual operation of the vehicle, and a large model error.

[0006] (2) In the existing ecological driving technology methods that take into account the longitudinal and lateral coupling characteristics of the vehicle, the vehicle dynamics system has nonlinear and highly coupled characteristics, which have a certain impact on the real-time and accuracy of vehicle control. The current methods are difficult to achieve efficient and accurate closed-loop feedback control of the vehicle dynamics system.

[0007] (3) Existing eco-driving technologies based on optimization control lack optimization methods for the control time domain. The control time domain is often given in advance through simulation experiments and other means. This limits the flexibility of vehicle driving strategy selection, making it impossible to achieve optimal energy savings for the vehicle, and also leads to a significant loss of vehicle maneuverability. Summary of the invention

[0008] The purpose of the present invention is to overcome the defects of the above-mentioned prior art and to provide an ecological driving control method for intelligent connected vehicles at intersections. The vehicle dynamics model is globally linearized based on massive simulated sampling data, and a model predictive control algorithm based on the spatial domain is constructed to realize real-time decision-making and planning of the driving trajectory of intelligent connected vehicles in urban road environments, and to realize efficient and green passage of intelligent connected vehicles at urban intersections.

[0009] The purpose of the present invention can be achieved by the following technical solution: A method for controlling ecological driving at an intersection of an intelligent networked vehicle comprises the following steps:

[0010] S1. Based on GPS technology and V2I (Vehicle to Infrastructure) communication, real-time traffic information is collected;

[0011] S2, globally linearize the vehicle dynamics model based on the simulated sampled vehicle trajectory data;

[0012] S3. Combine intersection signal timing constraints and the global linearized vehicle dynamics model as constraints to construct an eco-driving optimization control model based on the spatial domain;

[0013] S4. Based on model predictive control technology, the real-time collected traffic information is input into the eco-driving optimization control model. Through iterative dynamic solution, the optimal motion trajectory of eco-driving is generated in real time through real-time planning, and the automatic driving control instructions are updated and optimized in real time, so that the vehicle can travel according to the optimal motion trajectory of eco-driving.

[0014] Furthermore, in step S1, the intelligent connected vehicle specifically obtains and collects traffic information from the on-board GPS device, the on-board sensing communication device and the roadside sensing communication module.

[0015] Furthermore, the traffic information collected in step S1 includes the main vehicle position coordinates, vehicle speed, yaw angle, the environment vehicle position coordinates, vehicle speed, and intersection signal timing information.

[0016] Furthermore, the step S2 specifically adopts a nonlinear dynamic model with high coupling in longitudinal and lateral directions to simulate the sampled trajectory data offline, and based on the Koopman operator, uses the simulated sampled trajectory data as input to globally linearize the vehicle dynamics model.

[0017] Furthermore, the step S2 comprises the following steps:

[0018] S21, basis function vector selection and state vector amplification;

[0019] The nonlinear vehicle dynamics model is x k+1 =f(x k ,u k ),in, It is represented as the vehicle state vector at the kth moment, including the vehicle's longitudinal and lateral position coordinates, longitudinal and lateral velocities, yaw angle, and yaw angular velocity, is the vehicle control input vector at the kth moment, including the vehicle acceleration and front wheel steering angle;

[0020] According to Koopman linear operator theory, by expanding the state vector x k The dimension makes the original nonlinear dynamic model tend to be linear, that is, there is a linear With the basis function φ, the following equation holds:

[0021]

[0022] Select radial basis function ψ and linear function As the basis function φ,

[0023]

[0024] Among them, ψ i (x k )=||x k -x i || 2 log(||x k -x i ||), x i is a randomly selected parameter vector;

[0025] Then the original state vector Expanded to z k =(x k ,ψ1(x k ),…,ψ N (x k ));

[0026] Just remember Linear Koopman Then the nonlinear dynamic system x k+1 =f(x k ,u k ) is transformed into a linear dynamic system z k+1 =Az k +Buk , A, B are the parameter matrices to be determined;

[0027] S22, offline simulation sampling of trajectory data;

[0028] Based on the virtual simulation platform, K (K>>n, m, N) control inputs u are randomly generated within the feasible range to form a control input data set U = [u1, ..., u k , ..., u K ], the control input in U acts on the nonlinear dynamic system x in sequence k+1 =f(x k ,u k ), then we get the state sequence X = [x1, ..., x k , ..., x K+1 ], substituting the state sequence into the radial basis function in turn, we can get the expanded state sequence Z = [z1, ..., z k , ..., z K ] and Y = [z2, ..., z k , ..., z K+1 ];

[0029] S23, global linearization of vehicle dynamics model;

[0030] The undetermined parameter matrices A and B of the linear dynamics system are described using the least squares problem:

[0031]

[0032] The solution to this least squares problem is Then, the simulated sampled trajectory data is substituted into the formula to obtain the parameter matrices A and B, thus completing the global linearization of the nonlinear vehicle dynamics model.

[0033] Furthermore, the step S3 specifically includes the following steps:

[0034] S31. Considering the energy consumption and mobility of intelligent connected vehicles, a model cost function is constructed;

[0035] S32, according to the current state of the traffic light, according to the two types of states of red light and green light, establish corresponding signal timing constraints;

[0036] S33. Considering the spatial fixity of intelligent connected vehicles passing through intersections, the time domain model is converted into a space domain model. The longitudinal position of the vehicle is used as the underlying variable and time is used as the system state variable to construct an ecological driving optimization control model based on the space domain.

[0037] Furthermore, the model cost function in step S31 is specifically:

[0038]

[0039] The corresponding conversion into vector matrix form is:

[0040]

[0041] Among them, β0, β1, and β2 are weight coefficients; v represents the longitudinal speed of the intelligent connected vehicle, which is the variable to be determined; v des is the expected speed of the intelligent networked vehicle; M is the mass of the intelligent networked vehicle; a is the longitudinal acceleration, which is the variable to be determined; f is the resistance of the vehicle; s tf is the final longitudinal position of the vehicle; L is the longitudinal distance between the vehicle and the stop line; t0 is the initial time when the vehicle arrives at the intersection, which is a known quantity; t f is the time when the vehicle passes through the intersection, is the quantity to be determined, the outer term of the integral of the cost function and the first term inside the integral are used to describe the mobility demand of the vehicle, and the second term inside the integral is used to describe the energy consumption demand of the vehicle; Q tf , Q, S are weight coefficient matrices.

[0042] Furthermore, the signal timing constraints in step S32 are specifically:

[0043] If the current light is red, then t f The following constraints must be met:

[0044]

[0045] Among them, r rest is the remaining red light time, g is the green light time;

[0046] If the light is green at the current moment, then t f The following constraints must be met:

[0047] When rest ≥τ, t f ≤g rest +t0;

[0048] When rest When t f ≥g rest +r+t0;

[0049] Among them, g rest is the remaining green light time; r is the red light time; τ is the critical time, which is determined by the maximum acceleration limit and maximum speed limit of the vehicle.

[0050] Furthermore, the step S33 specifically converts the time domain model into a space domain model through the following conversion formula:

[0051]

[0052] Based on this transformation method, t0→t f The uncertain time interval is transformed into the fixed space interval s0→L, and the original free time domain optimization control problem is transformed into a constrained fixed space domain optimization control problem.

[0053] Furthermore, the specific process of iterative dynamic solution in step S4 is:

[0054] S41. Discretize the model cost function to obtain the weight coefficient matrix Q for each spatial step k , S k , k = (0, 1, ..., M);

[0055] S42, set the number of iterations i=1, the convergence threshold ε, and set the initial value of the control input sequence U to U0;

[0056] S43, reverse calculation of control law gain:

[0057] When k = M, set the gain matrix P M =Q M ;

[0058] From k=M-1 to k=1, calculate the following equation in reverse recursion:

[0059]

[0060] S44, forward calculation control input u k :

[0061]

[0062] z k+1 =Az k +Bu k

[0063] Get the control input sequence U and state sequence Z;

[0064] S45, judging whether the current time state meets the signal timing constraint condition, if so, executing step S46; if not, correcting the control input U according to the boundary value of the constraint condition, and then executing step S46;

[0065] S46. Convergence judgment: if ||U-U0||≤ε is satisfied, the optimal solution is obtained and the solution is completed; if not satisfied, set U0=U, i=i+1 and return to execute step S43.

[0066] Compared with the prior art, the present invention has the following advantages:

[0067] 1. The present invention proposes a data-driven ecological driving control method for intelligent connected vehicles at intersections. First, the nonlinearity and longitudinal and lateral coupling effects of the vehicle dynamics model are considered, and massive simulated sampling data is used to globally linearize the vehicle dynamics model based on the Koopman operator to improve the execution efficiency of the decision-making control algorithm; then, the spatiotemporal coupling characteristics of urban roads are considered, and an ecological driving optimization control model based on the spatial domain is constructed. Then, the model predictive control algorithm is combined, and GPS and V2I communication data are used as model inputs. Accurate and rapid decision-making planning of the driving trajectory of the intelligent connected vehicle is achieved, thereby realizing efficient and green passage of the intelligent connected vehicle at urban intersections.

[0068] 2. The present invention fully considers the longitudinal and lateral coupling effects of the nonlinear vehicle dynamics model, and adopts the Koopman operator theory to globally linearize the vehicle dynamics model, which can effectively improve the efficiency of subsequent calculations and solutions, thereby realizing real-time rolling optimization control of intelligent connected vehicles.

[0069] 3. The present invention takes into account the impact of signal timing constraints on vehicle operation in urban intersection environments, while taking into account the space-time coupling effects of urban roads, and constructs a model prediction control method based on the spatial domain, thereby converting the time uncertainty problem into a spatial deterministic problem, which can eliminate the impact of the free control time domain on the model solution and ensure the reliability of the model solution results.

[0070] Fourth, the present invention takes into account both the energy consumption and driving maneuverability of the vehicle, implements dual-objective optimization of energy consumption and driving maneuverability, and constructs a model cost function. While saving driving energy consumption and achieving energy conservation and emission reduction, it can minimize the loss of vehicle maneuverability and ensure that the vehicle's driving is both environmentally friendly and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] Figure 1 It is a schematic diagram of the method flow of the present invention. DETAILED DESCRIPTION

[0072] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0073] Example

[0074] like Figure 1 As shown, an intelligent networked vehicle intersection ecological driving control method includes the following steps:

[0075] S1. Collect traffic information in real time based on GPS technology and V2I communication;

[0076] S2, globally linearize the vehicle dynamics model based on the simulated sampled vehicle trajectory data;

[0077] S3. Combine intersection signal timing constraints and the global linearized vehicle dynamics model as constraints to construct an eco-driving optimization control model based on the spatial domain;

[0078] S4. Based on model predictive control technology, the real-time collected traffic information is input into the eco-driving optimization control model. Through iterative dynamic solution, the optimal motion trajectory of eco-driving is generated in real time through real-time planning, and the automatic driving control instructions are updated and optimized in real time, so that the vehicle can travel according to the optimal motion trajectory of eco-driving.

[0079] This embodiment applies the above technical solution, and its main contents include:

[0080] Step 1: Based on GPS technology and V2I communication, real-time traffic information (main vehicle information, surrounding vehicle information, signal timing information, etc.) is collected.

[0081] In step 1, the intelligent connected vehicle can obtain information based on the on-board GPS device, on-board sensing and communication device, and roadside sensing and communication module. The information includes: the main vehicle position coordinates, vehicle speed, yaw angle, the environment vehicle position coordinates, vehicle speed, and intersection signal timing information.

[0082] In order to collect the above information, in this embodiment, the intelligent connected vehicle used should have a vehicle information collection device, a communication device, a vehicle database, and an on-board control unit; the intelligent traffic light should have an information sensing device, a communication device, and a roadside data processing device.

[0083] Step 2: Based on the simulated sampled vehicle trajectory data, the vehicle dynamics model is globally linearized to improve the computational efficiency of the decision-making control algorithm.

[0084] In step 2, a nonlinear dynamic model with high coupling in longitudinal and lateral directions is used to simulate the sampled trajectory data offline, and then based on the Koopman operator, the vehicle dynamics model is globally linearized with the simulated sampled trajectory data as input. The following steps are included:

[0085] Step 2.1: Basis function vector selection and state vector expansion. Let the exact nonlinear vehicle dynamics model be x k+1 =f(x k ,u k ),in, It is represented as the vehicle state vector at the kth moment, which includes the vehicle's longitudinal and lateral position coordinates, longitudinal and lateral velocities, yaw angle, and yaw angular velocity. is the vehicle control input vector at the kth moment, including the vehicle acceleration and front wheel angle. According to Koopman linear operator theory, the state vector x can be expanded k The dimension makes the original nonlinear dynamic model tend to be linear, that is, there is a linear With the basis function φ, the following equation holds:

[0086]

[0087] Select radial basis function ψ and linear function As the basis function φ,

[0088]

[0089] Among them, ψ i (x k )=||x k -x i || 2 log(||x k -x i ||), x i is a randomly selected parameter vector.

[0090] Then the original state vector Can be expanded to z k =(x k ,ψ1(x k ),…,ψ N (x k )).

[0091] Remember the linear Koopman operator Then the nonlinear dynamic system x k+1 =f(x k ,u k ) can be transformed into a linear dynamic system z k+1 =Az k +Bu k , A, B are the parameter matrices to be determined.

[0092] Step 2.2: Offline simulation sampling of trajectory data. Based on the virtual simulation platform, randomly generate K (K>>n, m, N) control inputs u within the feasible range to form a control input data set U = [u1, ..., u k , ..., u K ], the control input in U acts on the nonlinear dynamic system x in sequence k+1 =f(x k ,u k ), then we can get the state sequence X = [x1, ..., x k , ..., x K+1 ], substituting the state sequence into the radial basis function in turn, we can get the expanded state sequence Z = [z1, ..., z k , ..., z K ] and Y = [z2, ..., zk , ..., z K+1 ].

[0093] Step 2.3: Global linearization of the vehicle dynamics model. The unknown parameter matrices A and B of the linear dynamics system can be described by the following least squares problem:

[0094]

[0095] The solution to this least squares problem is Then, by substituting the simulated sampled data into the formula, we can obtain the parameter matrices A and B, thus completing the global linearization of the nonlinear vehicle dynamics model.

[0096] Step 3: Based on the intersection signal timing constraints, an eco-driving optimization control model based on the spatial domain is constructed.

[0097] In step 3, considering the time-space coupling characteristics of the road and the intersection signal timing constraints, an eco-driving optimization control model based on the spatial domain is constructed. The following steps are included:

[0098] Step 3.1: Model cost function construction. The model cost function mainly considers the energy consumption and mobility of intelligent connected vehicles. The mathematical description of the cost function J can be given by the following formula:

[0099]

[0100] Among them, β0, β1, and β2 are weight coefficients; v represents the longitudinal speed of the intelligent connected vehicle, which is the variable to be determined; v des is the expected speed of the intelligent networked vehicle; M is the mass of the intelligent networked vehicle; a is the longitudinal acceleration, which is the variable to be determined; f is the resistance of the vehicle; s tf is the final longitudinal position of the vehicle; L is the longitudinal distance between the vehicle and the stop line; t0 is the initial time when the vehicle arrives at the intersection, which is a known quantity; t f is the time when the vehicle passes through the intersection, and is the quantity to be determined. The term outside the integral sign of the cost function and the first term inside the integral sign are used to describe the mobility requirements of the vehicle, and the second term inside the integral sign is used to describe the energy consumption requirements of the vehicle.

[0101] The cost function can be further transformed into a vector matrix form:

[0102]

[0103] Among them, Q tf , Q, and S are weight coefficient matrices, all of which are composed of constants such as β0, β1, β2, M, and f.

[0104] Step 3.2: Signal timing constraints are established. Signal timing is classified and discussed according to the current state of the traffic light (red light or green light). The mathematical description of the constraint is as follows:

[0105] If the current light is red, then t f The following constraints must be met:

[0106]

[0107] Among them, r rest is the remaining red light time, and g is the green light time.

[0108] If the light is green at the current moment, then t f The following constraints must be met:

[0109] When rest ≥τ, t f ≤g rest +t0

[0110] When rest When t f ≥g rest +r+t0

[0111] Among them, g rest is the remaining green light time; r is the red light time; τ is the critical time, which is determined by the maximum acceleration limit and maximum speed limit of the vehicle.

[0112] Step 3.3: Transformation of time domain model to space domain model. In the time domain optimization control model, time t f -t0 is a variable and is constrained by signal timing, so it is difficult to solve the time domain model. Considering the spatial fixity of intelligent connected vehicles passing through intersections, the time domain model can be converted into a space domain model for solution, with the longitudinal position of the vehicle as the underlying variable and time as the system state variable. The model is converted from a final state free problem to a final state fixed problem. The conversion method is as follows:

[0113]

[0114] Based on this transformation method, t0→t f The uncertain time interval is transformed into the fixed space interval s0→L. The original free time domain optimization control problem is transformed into a constrained fixed space domain optimization control problem.

[0115] Step 4: Based on the model characteristics and the idea of ​​model predictive control, an iterative dynamic programming algorithm is designed to plan the optimal motion trajectory of ecological driving in real time and optimize the autonomous driving control instructions online.

[0116] In step 4, the iterative dynamic programming algorithm includes the following steps:

[0117] Step 4.1: Discretize the cost function and obtain the weight coefficient matrix Q for each spatial step k , S k , (k=0,1,....,M).

[0118] Step 4.2: Set the number of iterations i = 1, the convergence threshold ε, and set the initial value of the control input sequence U to U0.

[0119] Step 4.3: Inversely calculate the control law gains:

[0120] When k = M, let the gain matrix P M =Q M ;

[0121] From k=M-1 to k=1, calculate the following equation in reverse recursion:

[0122]

[0123] Step 4.4: Forward calculation of control input u k :

[0124]

[0125] z k+1 =Az k +Bu k

[0126] Get the control input sequence U and state sequence Z

[0127] Step 4.5: Determine whether the desired time state satisfies the signal timing constraint. If so, execute step 4.6. If not, correct the control input U according to the boundary value of the constraint, and then execute step 4.6.

[0128] Step 4.6: Convergence judgment. If ||U-U0||≤ε is satisfied, the optimal solution is obtained and the algorithm ends; if it does not satisfy this equation, let U0=U, i=i+1 and return to step 4.3.

[0129] In summary, this technical solution is based on the vehicle longitudinal and lateral coupling dynamics model, Koopman operator theory, optimization control theory, and model predictive control method. It realizes a vehicle dynamics system linearization scheme based on simulated trajectory data and a spatial domain model predictive control scheme for urban intersections. It can make accurate and rapid decisions and plans on the driving trajectory of intelligent connected vehicles, thereby ensuring the maneuverability and ecology of intelligent connected vehicles in urban intersection environments.

Claims

1. An ecological driving control method for an intelligent networked vehicle at an intersection, characterized in that: The following steps are involved: S1. Collect traffic information in real time based on GPS technology and V2I communication; S2, globally linearize the vehicle dynamics model based on the simulated sampled vehicle trajectory data; S3. Combine intersection signal timing constraints and the global linearized vehicle dynamics model as constraints to construct an eco-driving optimization control model based on the spatial domain; S4. Based on the model predictive control technology, the real-time collected traffic information is input into the eco-driving optimization control model, and the optimal motion trajectory of eco-driving is generated through iterative dynamic solution in real time. The automatic driving control instructions are updated and optimized in real time, so that the vehicle can drive according to the optimal motion trajectory of eco-driving; The step S2 specifically adopts a nonlinear dynamic model with high coupling in longitudinal and lateral directions to simulate the sampled trajectory data offline, and based on the Koopman operator, uses the simulated sampled trajectory data as input to globally linearize the vehicle dynamics model; The step S2 comprises the following steps: S21, basis function vector selection and state vector amplification; The nonlinear vehicle dynamics model is x k+1 =f(x k ,u k ),in, It is represented as the vehicle state vector at the kth moment, including the vehicle's longitudinal and lateral position coordinates, longitudinal and lateral velocities, yaw angle, and yaw angular velocity, is the vehicle control input vector at the kth moment, including the vehicle acceleration and front wheel steering angle; According to Koopman linear operator theory, by expanding the state vector x k The dimension of makes the original nonlinear dynamic model tend to linearize, that is, there is a linear operator With the basis function φ, the following equation holds: Select radial basis function ψ and linear function As the basis function φ, Among them, ψ i (x k )=||x k -x i || 2 log(||x k -x i ||), x i is a randomly selected parameter vector; Then the original state vector Expanded to z k =(x k ,ψ1(x k ),…,ψ N (x k )); Remember the linear Koopman operator Then the nonlinear dynamic system x k+1 =f(x k ,u k ) is transformed into a linear dynamic system z k+1 =Az k +Bu k , A, B are the parameter matrices to be determined; S22, offline simulation sampling of trajectory data; Based on the virtual simulation platform, K control inputs u are randomly generated within the feasible range, K>>n, m, N, forming a control input data set U=[u1,...,u k , ..., u K ], the control input in U acts on the nonlinear dynamic system x in sequence k+1 =f(x k ,u k ), then we get the state sequence X = [x1, ..., x k , ..., x K+1 ], substituting the state sequence into the radial basis function in turn, we can get the expanded state sequence Z = [z1, ..., z k , ..., z K ] and Y = [z2, ..., z k , ..., z K+1 ]; S23, global linearization of vehicle dynamics model; The undetermined parameter matrices A and B of the linear dynamics system are described using the least squares problem: The solution to this least squares problem is Then, the simulated sampled trajectory data is substituted into the formula to obtain the parameter matrices A and B, thus completing the global linearization of the nonlinear vehicle dynamics model. The step S3 specifically comprises the following steps: S31. Considering the energy consumption and mobility of intelligent connected vehicles, a model cost function is constructed; S32, according to the current state of the traffic light, according to the two types of states of red light and green light, establish corresponding signal timing constraints; S33. Considering the spatial fixity of intelligent connected vehicles passing through intersections, the time domain model is converted into a space domain model. The longitudinal position of the vehicle is used as the underlying variable and time is used as the system state variable to construct an ecological driving optimization control model based on the space domain.

2. The method for controlling the ecological driving of an intelligent networked vehicle at an intersection according to claim 1, characterized in that: In the step S1, the intelligent connected vehicle specifically obtains and collects traffic information from the vehicle-mounted GPS device, the vehicle-mounted sensing and communication device, and the roadside sensing and communication module.

3. The method for controlling the ecological driving of an intelligent networked vehicle at an intersection according to claim 1, characterized in that: The traffic information collected in step S1 includes the main vehicle position coordinates, vehicle speed, yaw angle, ambient vehicle position coordinates, and intersection signal timing information.

4. The method for controlling the ecological driving of an intelligent networked vehicle at an intersection according to claim 1, characterized in that: The model cost function in step S31 is specifically: The corresponding conversion into vector matrix form is: Among them, β0, β1, and β2 are weight coefficients; v represents the longitudinal speed of the intelligent connected vehicle, which is the variable to be determined; v des is the expected speed of the intelligent networked vehicle; M is the mass of the intelligent networked vehicle; a is the longitudinal acceleration, which is the variable to be determined; f is the resistance of the vehicle; s tf is the final longitudinal position of the vehicle; L is the longitudinal distance between the vehicle and the stop line; t0 is the initial time when the vehicle arrives at the intersection, which is a known quantity; t f is the time when the vehicle passes through the intersection, is the quantity to be determined, the outer term of the integral of the cost function and the first term inside the integral are used to describe the mobility demand of the vehicle, and the second term inside the integral is used to describe the energy consumption demand of the vehicle; Q tf , Q, S are weight coefficient matrices.

5. The method for controlling the ecological driving of an intelligent networked vehicle at an intersection according to claim 1, characterized in that: The signal timing constraints in step S32 are specifically: If the current light is red, then t f The following constraints must be met: Among them, r rest is the remaining red light time, g is the green light time; If the light is green at the current moment, then t f The following constraints must be met: When rest ≥τ, t f ≤g rest +t0; When rest When t f ≥g rest +r+t0; Among them, g rest is the remaining green light time; r is the red light time; τ is the critical time, which is determined by the maximum acceleration limit and maximum speed limit of the vehicle.

6. The method for controlling the ecological driving of an intelligent networked vehicle at an intersection according to claim 5, characterized in that: The step S33 specifically converts the time domain model into a space domain model through the following conversion formula: Based on this conversion formula, t0→t f The uncertain time interval is transformed into the fixed space interval s0→L, and the original free time domain optimization control problem is transformed into a constrained fixed space domain optimization control problem.

7. The method for controlling the ecological driving of an intelligent networked vehicle at an intersection according to claim 4, characterized in that: The specific process of iterative dynamic solution in step S4 is: S41. Discretize the model cost function to obtain the weight coefficient matrix Q for each spatial step k , S k , k = (0, 1, ..., M); S42, set the number of iterations i=1, the convergence threshold ε, and set the initial value of the control input sequence U to U0; S43, reverse calculation of control law gain: When k = M, set the gain matrix P M =Q M ; From k=M-1 to k=1, calculate the following equation in reverse recursion: S44, forward calculation control input u k : z k+1 =The k +Bu k Get the control input sequence U and state sequence Z; S45, judging whether the current time state meets the signal timing constraint condition, if so, executing step S46; If not, the control input U is corrected according to the boundary value of the constraint condition, and then step S46 is executed; S46. Convergence judgment: if ||U-U0||≤ε is satisfied, the optimal solution is obtained and the solution is completed; if not satisfied, set U0=U, i=i+1 and return to execute step S43.

Images