Method for cooperative control of vehicle queues in mixed traffic flow

Abstract

The invention relates to the technical field of intelligent traffic and vehicle control, in particular to a method for cooperative control of a vehicle queue in a mixed traffic flow, and aims at improving the prediction accuracy of the driving intention of a man-driven vehicle by identifying and analyzing the following behavior of the man-driven vehicle through deep learning. Through identification of parameters of a car-following controller of an automatic driving vehicle, the intelligent network connection vehicle can adjust a car-following strategy of the intelligent network connection vehicle, effective cooperation with other vehicles is achieved, the overall traffic system performance is improved, and traffic congestion is relieved. Finally, a mixed queue collaborative control theory is constructed, verification is carried out through a demonstration system, and a feasible application demonstration is provided for an intelligent network connection road to be constructed. The method has remarkable beneficial effects in the aspects of improving the car-following behavior prediction accuracy, enhancing the collaboration, adaptability and flexibility of the hybrid queue, reducing the calculation complexity, improving the traffic flow efficiency, promoting the development of an intelligent traffic system and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent transportation and vehicle control, and particularly to a method for collaborative control of vehicle queues in a mixed traffic flow. Background Art

[0002] Due to the addition of autonomous driving technology and intelligent connected vehicles, a mixed traffic flow scenario has emerged on the road, where traditional human-driven vehicles, autonomous driving vehicles, and intelligent connected vehicles coexist. In such a mixed traffic flow, there are complex coupling characteristics between different types of vehicles, and their behavior patterns and interaction methods have become extremely complex, posing a major challenge to existing traffic management and control strategies.

[0003] Traditional human-driven vehicles rely on the operation of human drivers, and the skill levels, experiences, and driving habits of each driver are different, which makes the behavior of the vehicles highly unpredictable. Human drivers may make irrational driving decisions due to distraction, mood swings, or other personal factors, and these behaviors often lead to traffic accidents or traffic jams, seriously affecting the safety and smoothness of road traffic.

[0004] Autonomous driving vehicles, with the help of technologies such as lidar, cameras, and various advanced sensors, have the ability to autonomously perceive the surrounding environment and make decisions, which improves driving safety to a certain extent. However, currently, most autonomous driving vehicles mainly rely on a single-vehicle control mode, and there is a lack of an effective collaborative mechanism between vehicles. This means that during driving, they often can only make decisions based on the information they obtain themselves, and cannot make full use of the real-time state information of other vehicles, making it difficult to achieve efficient collaboration of the entire traffic flow, and their collaborative advantages have not been fully exploited.

[0005] Intelligent connected vehicles, relying on advanced communication technologies, have the ability to exchange real-time information with other vehicles and traffic infrastructure. Through vehicle-to-vehicle (V2V) and vehicle-to-infrastructure (V2I) communications, intelligent connected vehicles can obtain more comprehensive traffic information and theoretically achieve more efficient traffic flow and a safer driving environment. However, due to the huge number of traditional human-driven vehicles, it is not realistic to upgrade all traditional vehicles to intelligent connected vehicles in the short term, and the high retrofit cost makes the large-scale application of intelligent connected vehicles face numerous challenges in the short term.

[0006] Currently, the control research on the three types of vehicles, namely traditional human-driven vehicles, autonomous vehicles, and connected and autonomous vehicles, is relatively independent. Current research often focuses on the characteristics of a certain type of vehicle and fails to fully consider the coordinated control among different types of vehicles. Most traditional traffic management methods are designed based on the behavior patterns and traffic rules of human-driven vehicles, and they seem inadequate when facing the emerging scenario of mixed traffic flow. Due to the failure to effectively integrate the advantages of different types of vehicles, it is difficult to improve the overall efficiency of the traffic system, and the potential of different types of vehicles cannot be fully explored. Summary of the Invention

[0007] Aiming at the problem of low coordination efficiency of the mixed traffic flow formed by traditional human-driven vehicles, autonomous vehicles, and connected and autonomous vehicles in the prior art, the present invention provides a method for collaborative control of vehicle queues in mixed traffic flow.

[0008] The present invention is realized through the following technical solutions: A method for collaborative control of vehicle queues in mixed traffic flow includes the following steps: S1. Identify the following behavior of traditional human-driven vehicles through deep learning and predict the short-term following behavior of human-driven vehicles; Identify the following controller parameters of autonomous vehicles based on pattern recognition; S2. Construct a mixed queue following model based on model predictive control, initialize the terminal state cost, recursively calculate the cost function at each moment from the terminal time, and obtain the optimal control input expression through the Bellman equation.

[0009] Preferably, in S1, the identification process of the following characteristics of traditional human-driven vehicles is as follows: S111. Screen out the data of the second lane (Lane_ID = 2) from the NGSIM dataset, filter out the data without a preceding vehicle (Preceding is 0), and select trucks with relatively more data as the research object; S112. Fit the parameters of the IDM, GM, OVM, LCM, and FVD following models; S113. Use the PCA technique to calculate the covariance matrix, eigenvalues, and eigenvectors after standardizing the original data, and select 2 as the number of principal components for data dimensionality reduction; S114. Use the K-Means clustering algorithm to cluster the data after dimensionality reduction; S115. Construct an LSTM model, split the dataset into a training set and a test set according to the clustering results, input and train the LSTM model, and output the predicted value of V Acc value.

[0010] Preferably, in S111, the Savitzky-Golay filtering algorithm is used to smooth the original data; traverse the data set, determine the front and rear vehicle pair relationships by comparing the Frame_ID and Preceding columns, and extract the front and rear vehicle pair data.

[0011] Preferably, in S112, when performing parameter fitting, the mean square error between the model prediction and the actual data is calculated by the optimization objective function, and the minimize function and the Nelder-Mead algorithm are used to optimize the parameters for each pair of vehicles to obtain the fitting parameters.

[0012] Preferably, in S114, the optimal number of clusters for K-Means clustering is determined by the elbow method.

[0013] Preferably, in S115, the number of hidden layer units in the LSTM model is set to 50, the activation function is tanh, the optimizer is Adam, the loss function is the mean square error MSE, the trained model is used to predict the data and calculate the MSE value, and the model for each cluster is saved.

[0014] Preferably, in S1, the specific process of identifying the parameters of the autonomous vehicle following controller based on pattern recognition is as follows: S121, read the smoothed acceleration data of the leading vehicle trajectory from the NGSIM data set and store it in a DataFrame; create a PIDController class to represent the PID controller, including attributes such as kp, ki, kd, and setpoint; create a Car class to represent the vehicle, including attributes such as v_vel, v_acc, and local_y. S122, define an error_function function to calculate the error between the output generated by the PID parameters and the actual relative acceleration; calculate the relative acceleration of adjacent vehicles based on the leading vehicle acceleration data and store it in the rel_acc_data list; set the value ranges of kp, ki, and kd, call the differential_evolution function for differential evolution optimization, pass in the error function, parameter boundaries, and relative acceleration data, and optimize to obtain the optimal PID parameter values and assign them to the PID controller object. S123, use the updated PID controller parameters to simulate the vehicle, traverse all vehicles, update the speed and position of the vehicle according to the relative acceleration and the PID controller output, print the status information such as the speed, acceleration, and position of each vehicle, and output the optimal PID parameter values and the simulated vehicle driving results.

[0015] Preferably, in S2, when performing optimization to solve, by calculating the Pt matrix, the optimization process is decomposed into multiple sub-problems, and updated forward from the terminal time using the recurrence formula.

[0016] An electronic device includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the method for collaborative control of vehicle queues in the mixed traffic flow are implemented.

[0017] A storage medium stores a computer program thereon. When the computer program is executed by a processor, the steps of the method for collaborative control of vehicle queues in the mixed traffic flow are implemented.

[0018] Compared with the prior art, the present invention has the following beneficial effects: The method for collaborative control of vehicle queues in the mixed traffic flow of the present invention identifies and analyzes the following behavior of human-driven vehicles through deep learning, improves the prediction accuracy of the driving intention of human-driven vehicles, helps intelligent connected vehicles to make responses in advance, and reduces the risk of traffic accidents. By identifying the parameters of the following controller of autonomous vehicles, intelligent connected vehicles can adjust their following strategies, achieve effective collaboration with other vehicles, improve the performance of the overall traffic system, and alleviate traffic congestion. Finally, a theory of collaborative control of mixed queues is constructed and verified through a demonstration system, providing a practical application demonstration for the upcoming intelligent connected highway construction, and having significant beneficial effects in improving the prediction accuracy of following behavior, enhancing the collaboration, adaptability and flexibility of mixed queues, reducing the computational complexity, improving the traffic flow efficiency, and promoting the development of intelligent transportation systems.

[0019] Furthermore, through deep learning technology, considering the complex interactions between vehicles, especially the LSTM model, can accurately predict the short-term following behavior of traditional human-driven vehicles, thereby improving the accuracy and reliability of the prediction.

[0020] Furthermore, a mixed queue following model based on MPC is constructed, which can consider the dynamic behavior of the entire queue and minimize the cost function by optimizing the control input, helping to enhance the collaboration between vehicles, reduce conflicts and unnecessary accelerations and decelerations between vehicles, and thus improve the efficiency and safety of the traffic flow.

[0021] Furthermore, by identifying and fitting various following models (such as IDM, GM, OVM, etc.), it can adapt to different types of vehicles and driving behaviors, and can more comprehensively understand the following characteristics of human-driven vehicles.

[0022] Furthermore, by optimizing the following behavior between vehicles, the spacing fluctuations and speed changes between vehicles can be reduced, thereby reducing traffic congestion and energy consumption. Description of the Drawings

[0023] Figure 1 is the technology roadmap; Figure 2 is the speed smoothing effect diagram of Vehicle No. 1; Figure 3 is the acceleration smoothing effect diagram of Vehicle No. 1; Figure 4 is the parameter fitting result diagram of the traditional human-driven vehicle following model; Figure 5 is the diagram for determining the number of clusters by the elbow method; Figure 6 is the clustering result diagram of driving behaviors; Figure 7 is the comparison diagram of the actual acceleration - predicted acceleration for the first category; Figure 8 is the comparison diagram of the actual acceleration - predicted acceleration for the second category; Figure 9 is the comparison diagram of the actual acceleration - predicted acceleration for the third category; Figure 10 is the training process diagram of the first category of LSTM; Figure 11 is the training process diagram of the second category of LSTM; Figure 12 is the training process diagram of the third category of LSTM; Figure 13 is the parameter fitting flow chart of the autonomous vehicle; Figure 14 is the comparison diagram of the acceleration of the leading vehicle and the following vehicle under the PID controller; Figure 15 is the hybrid queue structure diagram; Figure 16 is the operation result diagram with a prediction time step of 30 steps; Figure 17 is the operation result diagram with a position error weight of 20; Figure 18 is the operation result diagram with a speed error weight of 1; Figure 19 is the hybrid queue cooperative control demonstration system diagram. Specific implementation manners

[0024] The present invention will be further described in detail below in conjunction with specific embodiments, which are explanations rather than limitations of the present invention.

[0025] The present invention discloses a method for cooperative control of vehicle queues in a mixed traffic flow. Referring to Figure 1 , the method includes the following steps: A method for cooperative control of vehicle queues in a mixed traffic flow includes the following steps: S1. Identify the following - vehicle behavior of traditional human - driven vehicles through deep learning and predict the short - term following - vehicle behavior of human - driven vehicles. Specifically, S111. Select the data of the second lane (Lane_ID = 2) from the NGSIM dataset, filter out the data without a preceding vehicle (Preceding is 0), and select trucks with relatively more data as the research object. Use the Savitzky - Golay filtering algorithm with a window length of 21 and a polynomial order of 3 to smooth the original data to obtain accurate vehicle driving data. The smoothing effect is as Figure 2 、 3 shown. Traverse the dataset, determine the following - vehicle pair relationship by comparing the Frame_ID and Preceding columns, and extract the following - vehicle pair data.

[0026] S112. Perform parameter fitting on the following - vehicle models of IDM (Intelligent Driver Model), GM (General Motors), OVM (Optimal Velocity Model), LCM (Longitudinal Control Model), and FVD (Full Velocity Difference). Specifically, define the optimization objective function to calculate the mean square error between the model prediction and the actual data, use the minimize function and the Nelder - Mead algorithm to optimize the parameters for each pair of vehicles, set the maximum number of iterations to 9, and obtain the fitting parameters. The parameter fitting results are as Figure 4 shown.

[0027] S113. After standardizing the original data using PCA technology, calculate the covariance matrix, eigenvalues, and eigenvectors, and select 2 principal components for data dimensionality reduction.

[0028] S114. Use the K - Means clustering algorithm to cluster the dimensionality - reduced data. Specifically, use the elbow method to determine that the optimal number of clusters for K - Means clustering is 3. The elbow method result graph is as Figure 5 shown. Next, use the K - Means clustering algorithm to cluster the data after PCA dimensionality reduction. According to the elbow method, the number of clusters is specified as 3. Execute the K - Means algorithm 10 times, set the maximum number of iterations in a single run to 300, and set a random seed to ensure the reproducibility of the results. The clustering results are as Figure 6 shown. The "1", "2", and "3" marked in red in the figure are the cluster centers of each category, and different colors represent different categories.

[0029] S115. Build an LSTM model. Split the dataset into a training set and a test set according to the clustering results. Use a time step of 10 as the input to train the LSTM model and output the predicted value of v_Acc. Specifically: For the three clustered categories, split the dataset respectively, input the feature dataset with a time step of 10, and output the predicted value of v_Acc. Standardize the feature data, split it into a training set and a test set, build and train an LSTM model. Set the number of hidden layer units to 50, the activation function to tanh, the optimizer to Adam, and the loss function to mean squared error (MSE). Use the trained model to predict the data and calculate the MSE value, and save the model for each cluster. After each category is trained, the actual values and predicted values are as Figure 7 、 8 、shown in Figure 9.

[0030] S116. Take out a vehicle of the first model and make predictions using the three models respectively. Plot the change trends of the training loss and validation loss of each cluster with the number of iterations as shown in Figure 10 、 11 、Figure 12. Verify the prediction effect of the model by comparing the MSE values and the training process. It can be seen from both the MSE values and the training process that the prediction effect of the first model is the best, indicating that there are differences between the three models.

[0031] Identify the parameters of the autonomous vehicle following controller based on the pattern type. Specifically: S121. Read the smooth acceleration data of the leading vehicle trajectory from the NGSIM dataset and store it in a DataFrame; create a PIDController class to represent the PID controller, including attributes such as kp, ki, kd, and setpoint; create a Car class to represent the vehicle, including attributes such as v_vel, v_acc, and local_y; S122. Define an error_function function to calculate the error between the output generated by the PID parameters and the actual relative acceleration; calculate the relative acceleration of the adjacent vehicle according to the leading vehicle acceleration data and store it in the rel_acc_data list; set the value ranges of kp, ki, and kd, and call the differential_evolution function for differential evolution optimization. Pass in the error function, parameter boundaries, and relative acceleration data to optimize and obtain the optimal PID parameter values, and assign them to the PID controller object; S123. Use the updated PID controller parameters to simulate the vehicle. Traverse all vehicles, update the speed and position of the vehicle according to the relative acceleration and the output of the PID controller, print the status information such as the speed, acceleration, and position of each vehicle, and output the optimal PID parameter values and the simulated vehicle driving results. The comparison chart of the acceleration of the leading vehicle and the following vehicle is as Figure 14 shown

[0032] S2. Construct a hybrid queue car-following model based on model predictive control, initialize the terminal state cost, recursively calculate the cost function at each time from the terminal time forward, and obtain the optimal control input expression

[0033] Specifically S21. System equation construction: First, the hybrid queue structure is as Figure 13 shown. Define the vehicle desired spacing expression as shown in Equation (3-1), and construct the system state expression, control input expression, system equation expression, and cost function. The system state expression reflects the difference between the actual distance between the front and rear vehicles and the desired safety distance and the speed difference; the control input expression ensures that the acceleration difference between the vehicles in the platoon remains within a reasonable range; the system equation expression describes the change of the system state over time; the cost function realizes the cooperative control of the hybrid platoon by minimizing the total cost of the deviation of the system state and the deviation of the control input from the expected value

[0034] Specifically, the system state expression is

[0035] In the formula —— Safe headway —— Vehicle driving speed —— Time headway —— Minimum static spacing when the two vehicles decelerate to a stop

[0036] That is, the desired spacing of the vehicle is equal to the speed multiplied by the time headway plus the minimum static spacing when the two vehicles decelerate to a stop

[0037] The control input expression is

[0038] In the formula represents at time in the platoon composed of represents at time the The actual distance between the th vehicle and the th vehicle (usually i.e., the adjacent vehicle); represents the speed difference between the th vehicle and the th vehicle at time; represents the desired distance between the th vehicle and the th vehicle at time; represents the difference between the actual distance and the desired distance between the

[0039] th vehicle and the

[0040]

[0041] In the formula: ; ; .

[0042]

[0043] In the formula: ; ; describes the error caused by prediction, control, and perception uncertainties: . The cost function is:

[0044] S22, Solving the optimal control input: Initialize the terminal state cost, and recursively calculate the cost function at each time step from the terminal time. Take the derivative of the cost function with respect to the control input and set it to zero to obtain the expression of the optimal control input. By calculating the Pt matrix, decompose the optimization process into multiple sub-problems and update from the terminal time forward using the recurrence formula to improve the calculation efficiency.

[0045] S23, Parameter verification: Set the prediction time steps to 1 step, 15 steps, and 30 steps, run the code respectively, analyze the changes in vehicle speed, acceleration, and position, and determine the appropriate prediction time step. Among them, the prediction time step of 30 steps is as Figure 16 shown. Next, set the position error weight to 20 and the speed error weight to 1. The operation results are as Figure 17 , 18 shown. Analyze the influence on vehicle spacing and speed consistency through the position difference and speed difference diagrams, and determine reasonable weight parameters.

[0046] The present invention also discloses a vehicle queue cooperative control system in a mixed traffic flow. Specifically: (1) System development: Use PyQt5 to develop a mixed queue cooperative control demonstration system, design the interface layout, and implement function modules such as vehicle type selection, car quantity selection, control buttons, and vehicle status display. The interface is as Figure 13 shown. The function explanations of each part are as follows: ① Vehicle type selection: There are four selectable types: HDV (red), AV (blue), CAV (green), and random. Among them, random means generating a mixed queue of random types according to the number of cars selected by the user.

[0047] ② Car quantity selection: Radio button box: The user can select the number of cars to be inserted. The options include 3, 4, 5, 6, 7, and 8 cars.

[0048] Input quantity box: The user can directly enter a specific quantity in the input box. Among them, when the user selects the first 3 types in the vehicle type selection, the quantity cannot be selected because by default, only one vehicle of a known type can be inserted each time.

[0049] ③ Control buttons and vehicle status display part: Insert car button: After clicking this button, according to the above selection, insert the corresponding number and type of cars into the simulation environment.

[0050] Start movement button: After clicking this button, the vehicles start to simulate according to the predetermined movement trajectory.

[0051] Stop movement button: After clicking this button, the vehicles pause.

[0052] Clear data button: After clicking this button, clear all current vehicle data and charts, and reset the simulation environment.

[0053] Vehicle status display box: Real-time display the position, speed, and acceleration of each vehicle.

[0054] ④ Visualization of vehicle type and quantity: Colors and squares are used to represent the currently selected vehicle type and quantity. Different colors represent different vehicle types. After clicking "Start Movement", each vehicle starts to move as in the real environment. After movement, the box may exhibit front-back jitter behavior, possibly due to the inconsistent frequencies of vehicle data updates and the graphical interface refresh, resulting in jitter in the displayed vehicle positions on the interface.

[0055] ⑤ Display of speed and acceleration charts: Real-time charts of speed and acceleration changes over time are plotted to help users intuitively understand the dynamic behavior of the vehicles.

[0056] (2) System testing: Test each function of the demonstration system, including inserting small vehicles by selecting different vehicle types, inserting small vehicles by selecting random vehicle types and specifying quantities, and clicking control buttons (insert small vehicle, start movement, stop movement, clear data), and observe the vehicle status display and chart plotting to ensure the correctness and usability of the system functions.

[0057] (3) Analysis of demonstration results: In the demonstration system, the HDV uses the vehicle prediction trajectory trained by LSTM (the first HDV reads the NGSIM dataset), the AV uses a PID controller with fitted parameters, and the CAV uses MPC for cooperative control. Analyze the demonstration results to verify the ability of different types of vehicles to perform intelligent following and maintain reasonable vehicle distances in the system, demonstrating the effectiveness of the hybrid queue following method.

[0058] The present invention also discloses an electronic device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the method for cooperative control of vehicle queues in a mixed traffic flow are implemented.

[0059] The present invention also discloses a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the method for cooperative control of vehicle queues in a mixed traffic flow are implemented.

[0060] The method for cooperative control of vehicle queues in a mixed traffic flow proposed by the present invention targets non-controllable vehicles in the system (traditional human-driven vehicles and autonomous vehicles), explores their following characteristics in the new traffic flow environment, provides an important basis for the cooperative management and control of non-controllable vehicles in a future vehicle networking environment for mixed traffic flows, and is a method that takes into account the coupling characteristics of traditional human-driven vehicles and autonomous vehicles, can solve the problems of insufficient cooperative management and control of different types of vehicles, low traffic efficiency, and high safety risks in the prior art, realizes the safe and efficient driving of vehicles in a mixed traffic flow environment, and improves the performance of the overall traffic system.

[0061] The above are only the preferred embodiments of the present invention and are not intended to limit the technical solutions of the present invention. Those skilled in the art should understand that without departing from the spirit and principles of the present invention, several simple modifications and substitutions can be made to the technical solutions, and these modifications and substitutions also fall within the protection scope covered by the claims.

Claims

1. A method for collaborative control of vehicle queues in mixed traffic flow, characterized in that, It includes the following steps: S1. Identify the following - driving behavior of traditional human - driven vehicles through deep learning and predict the short - term following - driving behavior of human - driven vehicles; Identify the following - driving controller parameters of autonomous driving vehicles based on pattern recognition; S2. Construct a hybrid - queue following - driving model based on model predictive control, initialize the terminal - state cost, recursively calculate the cost function at each moment from the terminal time forward, and obtain the optimal control input expression through the Bellman equation.

2. The method for collaborative control of vehicle queues in mixed traffic flows according to claim 1, characterized in that, In S1, the identification process of the following - driving characteristics of traditional human - driven vehicles is as follows: S111. Screen out the data of the second lane (Lane_ID = 2) from the NGSIM dataset, filter out the data without a preceding vehicle (Preceding is 0), and select trucks with relatively more data as the research object; S112. Fit the parameters of the IDM, GM, OVM, LCM, and FVD following - driving models; S113. After standardizing the original data using PCA technology, calculate the covariance matrix, eigenvalues, and eigenvectors, and select 2 principal components for data dimensionality reduction; S114. Use the K - Means clustering algorithm to cluster the dimensionality - reduced data; S115. Build an LSTM model, split the data set into a training set and a test set according to the clustering results, train the LSTM model, and output the predicted value of V Acc ​ 3. The method for collaborative control of vehicle queues in mixed traffic flows according to claim 1, characterized in that, In S111, use the Savitzky - Golay filtering algorithm to smooth the original data; traverse the dataset, determine the vehicle - pair relationship between the front and rear vehicles by comparing the Frame_ID and Preceding columns, and extract the vehicle - pair data.

4. The method for collaborative control of vehicle queues in mixed traffic flow according to claim 1, characterized in that In S112, during parameter fitting, define the optimization objective function to calculate the mean - square error between the model prediction and the actual data, and use the minimize function and the Nelder - Mead algorithm to optimize the parameters for each pair of vehicles to obtain the fitting parameters.

5. The method for collaborative control of vehicle queues in mixed traffic flows according to claim 1, wherein In S114, the optimal number of clusters for K - Means clustering is determined by the elbow method.

6. The method for collaborative control of vehicle queues in mixed traffic flow according to claim 1, wherein In S115, set the number of hidden - layer units in the LSTM model to 50, the activation function to tanh, the optimizer to Adam, the loss function to mean - square error MSE, use the trained model to predict the data and calculate the MSE value, and save the model for each cluster.

7. The method for collaborative control of vehicle queues in mixed traffic flow according to claim 1, characterized in that, In S1, the specific process of identifying the following - driving controller parameters of autonomous driving vehicles based on pattern recognition is as follows: S121. Read the smoothed acceleration data of the leading - vehicle trajectory from the NGSIM dataset and store it in a DataFrame; create a PIDController class to represent the PID controller, including attributes such as kp, ki, kd, and setpoint; Create a Car class to represent the vehicle, including attributes such as v_vel, v_acc, and local_y; S122. Define the error_function function to calculate the error between the output generated by the PID parameters and the actual relative acceleration; calculate the relative acceleration of adjacent vehicles based on the leading vehicle acceleration data and store it in the rel_acc_data list; set the value ranges of kp, ki, and kd, call the differential_evolution function for differential evolution optimization, pass in the error function, parameter boundaries, and relative acceleration data, and optimize to obtain the optimal PID parameter values, which are assigned to the PID controller object. S123. Use the updated PID controller parameters to simulate the vehicle. Traverse all vehicles, update the speed and position of the vehicle according to the relative acceleration and the PID controller output, print the status information such as the speed, acceleration, and position of each vehicle, and output the optimal PID parameter values and the simulation results of the vehicle driving.

8. The method for collaborative control of vehicle queues in mixed traffic flow according to claim 1, wherein In S2, during the optimization solution, by calculating the Pt matrix, the optimization process is decomposed into multiple sub-problems and updated forward from the terminal time using the recurrence formula.

9. An electronic device, comprising a memory and a processor, the memory storing a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method for cooperative control of vehicle queues in mixed traffic flows according to any one of claims 1 to 8.

10. A storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the method for cooperative control of vehicle queues in mixed traffic flows according to any one of claims 1 to 8.

Images