A Vehicle Queue Optimal Control Method and System Based on Function Deviation Triggering Mechanism

Abstract

This invention relates to the field of vehicle control technology, and particularly to a vehicle platooning optimal control method and system based on a function deviation triggering mechanism. The method includes: 1. Constructing a leading vehicle's dynamics model and a following vehicle's dynamics model; 2. Defining the local tracking error of the i-th following vehicle, and defining the control input of the following dynamics model as a function of the local tracking error according to the optimal control objective, thereby defining an optimal value function and determining the optimal control input based on the optimal value function; 3. Obtaining an estimate of the optimal value function through an adaptive critic neural network; 4. Designing a function deviation triggering mechanism based on the optimal value function estimate and the local tracking error to dynamically adjust the optimal control input of the following dynamics model; 5. Setting the communication relationship and initial state between the leading vehicle and the following vehicles to obtain the platooning control result. This invention enables efficient cooperative driving in complex dynamic environments.

Description

Technical Field

[0001] This invention relates to the field of vehicle control technology, and in particular to a vehicle queuing optimal control method and system based on a function deviation triggering mechanism. Background Technology

[0002] With the increasing prominence of traffic congestion and energy consumption problems, vehicle platooning control, as a key technology in intelligent transportation systems, has received widespread attention. By coordinating the distance, speed, and acceleration between vehicles, vehicle platooning control can achieve safe and efficient cooperative driving, thereby effectively alleviating traffic pressure and reducing fuel consumption. Traditional vehicle platooning control strategies, such as lead-follow control, model predictive control, and sliding mode control, have achieved some success, but these methods typically rely on accurate dynamic models and require complex online optimization, which limits their flexibility and robustness in dynamic and uncertain traffic environments.

[0003] Reinforcement learning, as a method to obtain optimal control strategies without requiring precise system dynamics models, has received increasing attention in solving complex queue control problems. Meanwhile, the development of wireless communication technology enables vehicles to share real-time information through vehicle-to-vehicle communication, but limited bandwidth and the need for high real-time communication pose challenges to the implementation of queue control. Event-triggered mechanisms, by communicating only when preset trigger conditions are met, can reduce redundant data transmission and alleviate communication load. However, the integration of existing event-triggered mechanisms with reinforcement learning frameworks within the vehicle queue control framework still requires further research. Furthermore, traditional static event-triggered mechanisms lack adaptability and struggle to dynamically adjust trigger conditions during the learning process; dynamic event-triggered mechanisms, on the other hand, find it difficult to pre-design trigger conditions based on specified performance, leading to certain reliability issues in practical applications. Summary of the Invention

[0004] This invention aims to address at least one of the technical problems existing in related technologies. To this end, this invention provides a vehicle queuing optimal control method and system based on a function deviation triggering mechanism, solving the problems of reliance on precise models, high communication overhead, and insufficient adaptability of event triggering mechanisms in existing technologies, thereby achieving efficient cooperative driving in complex dynamic environments.

[0005] This invention provides a vehicle queuing optimal control method based on a function deviation triggering mechanism, comprising the following steps:

[0006] Step 1: Construct the navigation dynamics model of the pilot vehicle and A following dynamics model for a vehicle;

[0007] Step 2: Define the local tracking error of the i-th following vehicle. Based on the optimal control objective, define the control input of the following dynamics model as a function of the local tracking error, then define the optimal value function, and finally determine the optimal control input based on the optimal value function.

[0008] Step 3: Obtain the estimated value of the optimal value function through an adaptive critic neural network;

[0009] Step 4: Based on the estimated optimal value function and local tracking error, design a function deviation triggering mechanism to dynamically adjust the optimal control input of the following dynamics model;

[0010] Step 5: Set the communication relationship and initial state between the lead vehicle and the follower vehicle, and run steps 1 to 4 to obtain the queue control results.

[0011] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step one specifically includes:

[0012] The vehicle convoy consisted of a lead vehicle and A number of following vehicles are defined, with the index of the lead vehicle set to 0, and the index of the following vehicles set to... ;

[0013] The navigation dynamics model of the pilot vehicle is ,

[0014] in, for The state vector of the navigator vehicle at any given moment. , for The location of the lead vehicle at all times. for The speed of the lead car at all times for The constant acceleration of the lead car For the system matrix;

[0015] The following dynamics model of the following vehicle is ,

[0016] in, for The state vector of the i-th following vehicle at time i. , for The position of the i-th following vehicle at time i. for The speed of the i-th following vehicle at time i. for The acceleration of the i-th following vehicle at time i. For the system matrix, for The control input for the i-th following vehicle at time i.

[0017] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step two specifically includes:

[0018] Define the local tracking error of the i-th following vehicle as: ,but

[0019]

[0020] in, Let be the local tracking error of the i-th following vehicle. Let i represent the communication relationship between the i-th following vehicle and the j-th following vehicle. If the i-th following vehicle and the j-th following vehicle can communicate directly, then... ,otherwise ; Let represent the communication relationship between the i-th following vehicle and the lead vehicle. If the i-th following vehicle can directly receive information from the lead vehicle, then... ,otherwise ; , Let be the expected relative distance between the i-th following vehicle and the lead vehicle. Let be the expected relative distance between the i-th following vehicle and the j-th following vehicle. To maintain the desired distance between the i-th following vehicle and the lead vehicle. To maintain the desired distance between the i-th following vehicle and the j-th following vehicle; for The state vector of the j-th following vehicle at time j. , for The position of the j-th following vehicle at time j. for The speed of the j-th following vehicle at time j. for The acceleration of the j-th following vehicle at time j.

[0021] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step two further includes the following steps:

[0022] If the control input is defined as a function of the local tracking error, then Control input of the i-th following vehicle at time i for ;

[0023] Introducing time-domain quadratic performance metrics

[0024] in, Let be the performance index of the i-th following vehicle. This is the weighting matrix of local tracking errors. It is a positive definite matrix. To control the weighted scalar input, ;

[0025] Define value function for

[0026]

[0027] in, This refers to the integration variable in integration operations;

[0028] The optimal value function is .

[0029] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step two further includes the following steps:

[0030] Based on the principle of optimality, the Hamiltonian function is defined. for

[0031]

[0032] in, Let i be the set of control inputs from the neighboring vehicle of the i-th following vehicle. Let be the set of local tracking errors of the i-th following vehicle's neighbors. for right gradient, The number of neighboring vehicles associated with the i-th following vehicle. This is the control input for the j-th following vehicle;

[0033] The optimal control input can be obtained by minimizing the Hamiltonian function. .

[0034] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step two further includes the following steps:

[0035] An event-triggered control strategy is adopted, and the event triggering time sequence is defined as follows: In each i-th following vehicle The moment when the control input is updated is triggered. The local tracking error of the i-th following vehicle It is sent to the controller, which then updates the optimal control input;

[0036] During the interval between two consecutive triggering events, the control input remains unchanged based on a zero-order hold mechanism, i.e. ,

[0037] in, For the i-th following vehicle The moment when the control input is updated is triggered. For the i-th following vehicle The moment when the control input is updated is triggered.

[0038] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step three includes:

[0039] Design an adaptive critic neural network, and leverage its approximation capabilities to represent the optimal value function as follows:

[0040] in, For the ideal weight vector, For the activation function of the neural network, This is an approximation error;

[0041] Approximation based on an adaptive reviewer neural network Its form is

[0042]

[0043] in, This is the estimated value of the optimal value function. This represents the estimated weight vector of the critic network;

[0044] Based on the estimated value of the optimal value function, the optimal control input is obtained as follows:

[0045]

[0046] in, This is an estimate of the optimal control input. The gradient of the activation function with respect to the local tracking error;

[0047] Under an event-triggered mechanism, if the control input remains constant between consecutive trigger moments, then we have:

[0048] in, This is an estimate of the optimal control input under the event-triggered mechanism;

[0049] Based on the integral reinforcement learning method, for a given time period, the time difference error is defined as...

[0050]

[0051] in, The time difference vector of the activation function. for The performance cost over a given time period For time difference error, For time period;

[0052] definition Define an instantaneous error function to represent the error in estimating commenter weights. Using gradient descent, the update law for the estimated weight vector of the critic network is:

[0053]

[0054] in, For learning rate, , It is an auxiliary variable.

[0055] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step four specifically includes:

[0056] Design a function deviation trigger mechanism

[0057]

[0058] in, The gradient of the optimal value function estimate with respect to the local tracking error. For the preset decay function, The first parameter of the trigger mechanism, , The second parameter of the trigger mechanism, , The preset threshold parameter related to vehicle i represents the upper bound of the optimal value function relative to the local tracking error. express The smallest eigenvalue.

[0059] A further improvement of the vehicle queuing optimal control method based on the function deviation triggering mechanism of this invention is that step five specifically includes:

[0060] Set the communication relationship, initial state, and time period between the lead vehicle and the follower vehicle. Weighted matrix of local tracking error Weighted scalar of control input Learning rate Triggering mechanism first parameter The second parameter of the triggering mechanism Preset threshold parameters related to vehicle i Run steps one through four to obtain the queue control results.

[0061] This invention also provides a vehicle queuing optimal control system based on a function deviation triggering mechanism. The vehicle queuing optimal control system is used to execute the vehicle queuing optimal control method described above. The vehicle queuing optimal control system includes:

[0062] The model building module is used to build the navigation dynamics model of the pilot vehicle and A following dynamics model for a vehicle;

[0063] The control input determination module is used to define the local tracking error of the following vehicle. According to the optimal control objective, the control input of the following dynamics model is defined as a function of the local tracking error, and then the optimal value function is defined to determine the optimal control input.

[0064] The value function estimation module is used to obtain the optimal value function estimate through an adaptive critic neural network;

[0065] The control input adjustment module is used to dynamically adjust the optimal control input of the following dynamics model based on the optimal value function estimate and the local tracking error design function deviation triggering mechanism.

[0066] The queue control operation module is used to set the communication relationship and initial state between the lead vehicle and the follower vehicles. It sequentially calls the functions of the model building module, control input determination module, value function estimation module, and control input adjustment module to obtain the queue control results.

[0067] This invention dynamically adjusts the triggering conditions based on the deviation between the optimal value function and the preset decay function, and combines integral reinforcement learning and adaptive commentator neural network to ensure system stability and control accuracy while reducing communication overhead.

[0068] This invention employs reinforcement learning, enabling the learning of optimal control strategies without requiring a precise system model, thus adapting to uncertainties in dynamic traffic environments. Utilizing a triggering mechanism based on function deviation, the triggering conditions are dynamically adjusted according to the deviation between the current optimal function and a preset attenuation function, achieving a trade-off between communication frequency and stability. The control input is designed based on optimal control theory, ensuring that it theoretically minimizes preset performance indicators, providing a globally optimal theoretical basis for cooperative driving in vehicle platoons.

[0069] This invention supports multiple communication topologies. For bidirectional lead-follow, bidirectional lead-tail follow, bidirectional lead-follow, and bidirectional dual lead-follow topologies, the designed methods can ensure good performance stability of the queuing system.

[0070] This invention focuses on designing a triggering mechanism based on function deviation, aiming to solve the vehicle queuing control problem based on reinforcement learning. Its core method involves introducing a dynamic deviation term between the optimal value function and a preset decay function, using this as the basis for adjusting the triggering conditions, thereby constructing a performance-aware adaptive triggering mechanism. On the one hand, compared to traditional static event triggering mechanisms, this mechanism exhibits greater flexibility, effectively reducing unnecessary control updates and communication events, thus lowering communication overhead. On the other hand, it overcomes the theoretical bottleneck of traditional dynamic event triggering mechanisms under preset performance constraints, significantly improving the utilization efficiency of communication resources while ensuring the system's preset performance. Ultimately, this mechanism provides a highly efficient cooperative control scheme for vehicle queuing systems that guarantees preset performance while possessing dynamic adaptability.

[0071] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0072] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0073] Figure 1 This is a flowchart of the optimal vehicle queue control method based on the function deviation triggering mechanism provided by the present invention.

[0074] Figure 2 These are schematic diagrams of different communication topologies.

[0075] Figure 3 This is a simulation experiment of the vehicle queuing optimal control method based on the function deviation triggering mechanism provided by this invention. Figure 1 .

[0076] Figure 4 This is a simulation experiment of the vehicle queuing optimal control method based on the function deviation triggering mechanism provided by this invention. Figure 2 .

[0077] Figure 5 This is a simulation experiment of the vehicle queuing optimal control method based on the function deviation triggering mechanism provided by this invention. Figure 3 . Detailed Implementation

[0078] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention. The following embodiments are used to illustrate this invention but should not be used to limit the scope of this invention.

[0079] The following is combined Figure 1 The vehicle queuing optimal control method based on the function deviation triggering mechanism of the present invention includes the following steps:

[0080] Step 1: Construct the navigation dynamics model of the pilot vehicle and A following dynamics model for a vehicle;

[0081] Step 2: Define the local tracking error of the i-th following vehicle. Based on the optimal control objective, define the control input of the following dynamics model as a function of the local tracking error, then define the optimal value function, and finally determine the optimal control input based on the optimal value function.

[0082] Step 3: Obtain the estimated value of the optimal value function through an adaptive critic neural network;

[0083] Step 4: Based on the estimated optimal value function and local tracking error, design a function deviation triggering mechanism to dynamically adjust the optimal control input of the following dynamics model;

[0084] Step 5: Set the communication relationship and initial state between the lead vehicle and the follower vehicle, and run steps 1 to 4 to obtain the queue control results.

[0085] Preferably, by constructing a navigation dynamics model for the pilot vehicle and A following dynamics model of the following vehicles can accurately simulate the motion state of a vehicle platoon, providing a foundation for subsequent control strategy design. This helps improve the overall motion accuracy and coordination of the platoon. Defining the control input of the following dynamics model as a function of local tracking error ensures that the following vehicles closely track the trajectory of the lead vehicle, reducing error accumulation and improving platoon stability. Furthermore, determining the optimal control input based on the optimal value function allows for further optimization of the control strategy, achieving more efficient energy utilization and lower energy consumption. Obtaining the optimal value function estimate through an adaptive critic neural network enables real-time adjustment of the control strategy to adapt to different traffic environments and vehicle states, improving control flexibility and adaptability, and helping the vehicle platoon maintain stable operation in complex and changing traffic environments. A function deviation triggering mechanism designed based on the optimal value function estimate and local tracking error dynamically adjusts the optimal control input of the following dynamics model, correcting deviations promptly and ensuring the vehicle platoon always maintains the desired motion trajectory, thus improving the safety and reliability of the platoon. Setting the communication relationship and initial state between the lead and following vehicles and running the entire control process yields accurate platoon control results.

[0086] In a preferred embodiment of the vehicle queuing optimal control method based on the function deviation triggering mechanism of the present invention, step one specifically includes:

[0087] The vehicle convoy consisted of a lead vehicle and A number of following vehicles are defined, with the index of the lead vehicle set to 0, and the index of the following vehicles set to... ;

[0088] The navigation dynamics model of the pilot vehicle is ,

[0089] in, for The state vector of the navigator vehicle at any given moment. , for The location of the lead vehicle at all times. for The speed of the lead car at all times for The constant acceleration of the lead car For the system matrix;

[0090] The following dynamics model of the following vehicle is ,

[0091] in, for The state vector of the i-th following vehicle at time i. , for The position of the i-th following vehicle at time i. for The speed of the i-th following vehicle at time i. for The acceleration of the i-th following vehicle at time i. For the system matrix, for The control input for the i-th following vehicle at time i.

[0092] Furthermore, step two specifically includes:

[0093] Define the local tracking error of the i-th following vehicle as: ,but

[0094]

[0095] in, Let be the local tracking error of the i-th following vehicle. Let i represent the communication relationship between the i-th following vehicle and the j-th following vehicle. If the i-th following vehicle and the j-th following vehicle can communicate directly, then... ,otherwise ; Let represent the communication relationship between the i-th following vehicle and the lead vehicle. If the i-th following vehicle can directly receive information from the lead vehicle, then... ,otherwise ; , Let be the expected relative distance between the i-th following vehicle and the lead vehicle. Let be the expected relative distance between the i-th following vehicle and the j-th following vehicle. To maintain the desired distance between the i-th following vehicle and the lead vehicle. To maintain the desired distance between the i-th following vehicle and the j-th following vehicle; for The state vector of the j-th following vehicle at time j. , for The position of the j-th following vehicle at time j. for The speed of the j-th following vehicle at time j. for The acceleration of the j-th following vehicle at time j.

[0096] Better place, This indicates that the state deviation of the i-th following vehicle from that of other following vehicles is from... arrive The traversal and accumulation.

[0097] Furthermore, step two also includes the following steps:

[0098] If the control input is defined as a function of the local tracking error, then Control input of the i-th following vehicle at time i for ;

[0099] Introducing time-domain quadratic performance metrics

[0100] in, Let be the performance index of the i-th following vehicle. This is the weighting matrix of local tracking errors. It is a positive definite matrix. The requirement of a positive definite matrix is ​​met, thereby ensuring that the performance index is strictly positive definite in its penalty for error, avoiding the performance index being zero when the error is non-zero, and ensuring that the control strategy effectively guides the system to stability; To control the weighted scalar input, ;

[0101] To calculate the optimal control input for each following vehicle, for any given allowable control input, a value function is defined. for

[0102]

[0103] in, This refers to the integration variable in integration operations;

[0104] The optimal value function is .

[0105] Furthermore, step two also includes the following steps:

[0106] Based on the principle of optimality, the Hamiltonian function is defined. for

[0107]

[0108] in, Let i be the set of control inputs from the neighboring vehicle of the i-th following vehicle. Let be the set of local tracking errors of the i-th following vehicle's neighbors. for right gradient, The number of neighboring vehicles associated with the i-th following vehicle. This is the control input for the j-th following vehicle;

[0109] The optimal control input can be obtained by minimizing the Hamiltonian function. .

[0110] Furthermore, step two also includes the following steps:

[0111] To reduce computational and communication burden, an event-triggered control strategy is adopted, defining the event trigger time sequence as follows: In each i-th following vehicle The moment when the control input is updated is triggered. The local tracking error of the i-th following vehicle It is sent to the controller, which then updates the optimal control input;

[0112] During the interval between two consecutive triggering events, the control input remains unchanged based on a zero-order hold mechanism, i.e. ,

[0113] in, For the i-th following vehicle The moment when the control input is updated is triggered. For the i-th following vehicle The moment when the control input is updated is triggered.

[0114] Therefore, the control objective can be defined as: given a vehicle platoon control system that satisfies the lead dynamics model and follow dynamics model in step one, design a suitable event triggering mechanism to determine the trigger sequence. The corresponding events trigger optimal control inputs to reduce communication frequency while ensuring system performance.

[0115] Preferably, the triggered event can be the norm of the error or the result of its comparison with a preset threshold. When the local tracking error exceeds the threshold, the event is triggered, and this moment is recorded as the event trigger moment. Subsequently, the local tracking error information of the following vehicle is transmitted to the controller to update the control input.

[0116] Ideally, the triggering time can also be a time variable, and the optimal triggering time sequence can be determined through algorithm optimization to achieve a balance between communication frequency and system performance.

[0117] Preferably, the triggering conditions are dynamically adjusted based on the deviation between the current optimal value function and the preset attenuation function to achieve a trade-off between communication frequency and stability. The control input is designed based on optimal control theory to ensure that it theoretically minimizes the preset performance indicators, providing a globally optimal theoretical basis for cooperative driving of vehicle platoons.

[0118] Furthermore, step three includes:

[0119] Design an adaptive critic neural network, and leverage its approximation capabilities to represent the optimal value function as follows:

[0120] in, For the ideal weight vector, For the activation function of the neural network, This is an approximation error;

[0121] Approximation based on an adaptive reviewer neural network Its form is

[0122]

[0123] in, This is the estimated value of the optimal value function. This represents the estimated weight vector of the critic network;

[0124] Based on the estimated value of the optimal value function, the optimal control input is obtained as follows:

[0125]

[0126] in, This is an estimate of the optimal control input. The gradient of the activation function with respect to the local tracking error;

[0127] Under an event-triggered mechanism, if the control input remains constant between consecutive trigger moments, then we have:

[0128] in, This is an estimate of the optimal control input under the event-triggered mechanism; under the event-triggered mechanism, when When the system meets the event-triggered control strategy, it triggers an update of the control input and generates new control input. And upon reaching the Prior to the next trigger moment, the control input remains unchanged, therefore, During the time period, the control input is continuously based on State deviation at time Output constant action;

[0129] Based on the integral reinforcement learning method, for a given time period, the time difference error is defined as...

[0130]

[0131] in, The time difference vector of the activation function. for The performance cost over a given time period For time difference error, For time period;

[0132] definition Define an instantaneous error function to represent the error in estimating commenter weights. Using gradient descent, the update law for the estimated weight vector of the critic network is:

[0133]

[0134] in, For learning rate, , It is an auxiliary variable.

[0135] Preferably, the adaptive critic neural network possesses powerful approximation capabilities, enabling the representation of complex optimal value functions into concise mathematical forms. This simplifies the design process of vehicle platooning control systems, reduces the requirements for system modeling accuracy, and improves the adaptability and robustness of the control system. Through the adaptive critic neural network, accurate estimation of the optimal control input can be achieved. Based on the estimated optimal value function and considering the gradient of the activation function with respect to local tracking errors, the accuracy and effectiveness of the control input are ensured, contributing to improved tracking accuracy and stability of the vehicle platooning and enhancing overall control performance.

[0136] Preferably, based on the integral reinforcement learning method, the estimated weight vector of the adaptive critic neural network can be continuously updated to minimize the instantaneous error function. The gradient descent method ensures the convergence and stability of the learning. Through continuous learning and optimization, the control system can gradually approach the optimal solution, achieving more efficient and intelligent vehicle queuing control.

[0137] Furthermore, step four specifically includes:

[0138] Design a function deviation trigger mechanism

[0139]

[0140] in, The gradient of the optimal value function estimate with respect to the local tracking error. For the preset decay function, The first parameter of the trigger mechanism, , The second parameter of the trigger mechanism, , The preset threshold parameter related to vehicle i represents the upper bound of the optimal value function relative to the local tracking error. express The smallest eigenvalue, This is the function deviation term, i.e., the deviation between the preset decay function and the estimated optimal value function. Under this function deviation event triggering mechanism, the system will only proceed if the triggering condition is met:

[0141] When the function deviation event triggers an update of the control input, compared to the traditional static event triggering mechanism, this function deviation event triggering mechanism can dynamically adjust the triggering conditions according to the function deviation term, showing greater flexibility while ensuring that the system meets the stability requirements.

[0142] Preferably, the function deviation triggering mechanism helps reduce energy consumption. By precisely controlling the relative position and speed between vehicles, unnecessary acceleration and braking operations are reduced, thereby reducing overall energy consumption, improving energy utilization efficiency, and enhancing the robustness of the system. In the face of complex traffic environments and variable vehicle dynamics characteristics, the function deviation triggering mechanism can maintain stable control effects and ensure that the vehicle platoon can maintain its optimal state under various conditions.

[0143] Furthermore, step five specifically includes: setting the communication relationship, initial state, and time period between the lead vehicle and the follower vehicle. Weighted matrix of local tracking error Weighted scalar of control input Learning rate Triggering mechanism first parameter The second parameter of the triggering mechanism Preset threshold parameters related to vehicle i Run steps one through four to obtain the queue control results.

[0144] This invention also provides a vehicle queuing optimal control system based on a function deviation triggering mechanism. The vehicle queuing optimal control system is used to execute the vehicle queuing optimal control method described above. The vehicle queuing optimal control system includes:

[0145] The model building module is used to build the navigation dynamics model of the pilot vehicle and A following dynamics model for a vehicle;

[0146] The control input determination module is used to define the local tracking error of the following vehicle. According to the optimal control objective, the control input of the following dynamics model is defined as a function of the local tracking error, and then the optimal value function is defined to determine the optimal control input.

[0147] The value function estimation module is used to obtain the optimal value function estimate through an adaptive critic neural network;

[0148] The control input adjustment module is used to dynamically adjust the optimal control input of the following dynamics model based on the optimal value function estimate and the local tracking error design function deviation triggering mechanism.

[0149] The queue control operation module is used to set the communication relationship and initial state between the lead vehicle and the follower vehicles. It sequentially calls the functions of the model building module, control input determination module, value function estimation module, and control input adjustment module to obtain the queue control results.

[0150] In a specific implementation example, a simulation experiment was conducted using the aforementioned optimal vehicle platoon control method. A total of 5 vehicles were used: one lead vehicle and four follower vehicles. The communication relationships within the vehicle platoon were established, and the communication topology of the vehicle platoon is as follows: Figure 2 As shown, (a) Two-way Lead-Front-Wheel Follow (BLPF), (b) Two-way Lead-Front Follow (BLTF), (c) Two-way Front-Wheel Follow (BPF), and (d) Two-way Dual Lead-Front-Wheel Follow (BTLPF) are all set to the initial state of the vehicle.

[0151]

[0152] Setting parameters The desired relative distance between adjacent vehicles is 7 meters. The trigger frequency is defined as the number of triggers divided by the number of simulation steps. The simulation results are as follows: Figure 3 , Figure 4 and Figure 5 As shown, Figure 3 , Figure 4 and Figure 5 The values ​​represent the local tracking error, velocity, and acceleration of the following vehicle under the BLPF topology, respectively. The trigger frequencies of each following vehicle are 6.64%, 9.62%, 4.80%, 40.44%, and 6.22%, respectively. Table 1 shows the average absolute distance error under different topologies.

[0153] Table 1

[0154]

[0155] Experiments show that the following vehicle can maintain the desired relative position with the lead vehicle, and its speed and acceleration remain synchronized. Under the designed function deviation triggering mechanism, the communication frequency between vehicles is greatly reduced.

[0156] Preferably, compared with the prior art, the specific beneficial technical effects of the present invention are as follows: It employs a reinforcement learning method, which can learn the optimal control strategy without a precise system model, adapting to the uncertainties in dynamic traffic environments; it utilizes a triggering mechanism based on function deviation, dynamically adjusting the triggering conditions according to the deviation between the current optimal value function and the preset attenuation function, achieving a trade-off between communication frequency and stability; it designs control inputs based on optimal control theory, ensuring that the control inputs theoretically minimize preset performance indicators, providing a globally optimal theoretical basis for cooperative driving of vehicle platoons; and it supports multiple communication topologies, ensuring good performance stability of the platoon system for bidirectional lead-follow, bidirectional lead-tail follow, bidirectional lead-follow, and bidirectional dual lead-follow topologies.

[0157] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A vehicle queuing optimal control method based on a function deviation triggering mechanism, characterized in that, Includes the following steps: Step 1: Construct the navigation dynamics model of the pilot vehicle and A following dynamics model for a vehicle; The vehicle convoy consisted of a lead vehicle and A number of following vehicles are defined, with the index of the lead vehicle set to 0, and the index of the following vehicles set to... ; The navigation dynamics model of the pilot vehicle is , in, for The state vector of the navigator vehicle at any given moment. , for The location of the lead vehicle at all times. for The speed of the lead car at all times for The constant acceleration of the lead car For the system matrix; The following dynamics model of the following vehicle is , in, for The state vector of the i-th following vehicle at time i. , for The position of the i-th following vehicle at time i. for The speed of the i-th following vehicle at time i. for The acceleration of the i-th following vehicle at time i. For the system matrix, for The control input for the i-th following vehicle at time i; Step 2: Define the local tracking error of the i-th following vehicle. Based on the optimal control objective, define the control input of the following dynamics model as a function of the local tracking error, then define the optimal value function, and finally determine the optimal control input based on the optimal value function. An event-triggered control strategy is adopted, and the event triggering time sequence is defined as follows: In each i-th following vehicle The moment when the control input is updated is triggered. The local tracking error of the i-th following vehicle It is sent to the controller, which then updates the optimal control input; During the interval between two adjacent triggering events, the control input remains unchanged based on a zero-order hold mechanism, i.e. , in, For the i-th following vehicle The moment when the control input is updated is triggered. For the i-th following vehicle The moment when the control input is updated is triggered. for The control input for the i-th following vehicle at time i; Step 3: Obtain the estimated value of the optimal value function through an adaptive critic neural network; Step 4: Based on the estimated optimal value function and local tracking error, design a function deviation triggering mechanism to dynamically adjust the optimal control input of the following dynamics model; Design a function deviation trigger mechanism in, Let be the local tracking error of the i-th following vehicle. This is the estimated value of the optimal value function. The gradient of the optimal value function estimate with respect to the local tracking error. For the preset decay function, The first parameter of the trigger mechanism, , The second parameter for triggering the mechanism, , The preset threshold parameter related to vehicle i represents the upper bound of the optimal value function relative to the local tracking error; This is the weighting matrix of local tracking errors. It is a positive definite matrix. for The smallest eigenvalue, The number of neighboring vehicles associated with the i-th following vehicle; Let represent the communication relationship between the i-th following vehicle and the lead vehicle. If the i-th following vehicle can directly receive information from the lead vehicle, then... ,otherwise ; Let i represent the communication relationship between the i-th following vehicle and the j-th following vehicle. If the i-th following vehicle and the j-th following vehicle can communicate directly, then... ,otherwise ; Let be the estimated value of the optimal control input for the i-th following vehicle. This is an estimate of the optimal control input for the j-th following vehicle; Step 5: Set the communication relationship and initial state between the lead vehicle and the follower vehicle, and run steps 1 to 4 to obtain the queue control results.

2. The optimal vehicle queue control method based on function deviation triggering mechanism according to claim 1, characterized in that, Step two specifically includes: Define the local tracking error of the i-th following vehicle as: ,but in, Let be the local tracking error of the i-th following vehicle. Let i represent the communication relationship between the i-th following vehicle and the j-th following vehicle. If the i-th following vehicle and the j-th following vehicle can communicate directly, then... ,otherwise ; Let represent the communication relationship between the i-th following vehicle and the lead vehicle. If the i-th following vehicle can directly receive information from the lead vehicle, then... ,otherwise ; , Let be the expected relative distance between the i-th following vehicle and the lead vehicle. Let be the expected relative distance between the i-th following vehicle and the j-th following vehicle. To maintain the desired distance between the i-th following vehicle and the lead vehicle To maintain the desired distance between the i-th following vehicle and the j-th following vehicle; for The state vector of the j-th following vehicle at time j. , for The position of the j-th following vehicle at time j. for The speed of the j-th following vehicle at time j. for The acceleration of the j-th following vehicle at time j.

3. The optimal vehicle queue control method based on function deviation triggering mechanism according to claim 2, characterized in that, Step two also includes the following steps: If the control input is defined as a function of the local tracking error, then Control input of the i-th following vehicle at time i for ; Introducing time-domain quadratic performance metrics in, Let i be the performance index of the i-th following vehicle. This is the weighting matrix of local tracking errors. It is a positive definite matrix. To control the weighted scalar input, ; Define value function for in, This refers to the integration variable in integration operations; The optimal value function is .

4. The optimal vehicle queue control method based on function deviation triggering mechanism according to claim 3, characterized in that, Step two also includes the following steps: Based on the principle of optimality, the Hamiltonian function is defined. for in, Let i be the set of control inputs from the neighboring vehicle of the i-th following vehicle. Let be the set of local tracking errors of the i-th following vehicle's neighbors. for right gradient, The number of neighboring vehicles associated with the i-th following vehicle. This is the control input for the j-th following vehicle; The optimal control input can be obtained by minimizing the Hamiltonian function. .

5. The optimal vehicle queue control method based on function deviation triggering mechanism according to claim 4, characterized in that, Step three includes: Design an adaptive critic neural network, and leverage its approximation capabilities to represent the optimal value function as follows: in, For the ideal weight vector, For the activation function of the neural network, This is an approximation error; Approximation based on an adaptive reviewer neural network Its form is in, This is the estimated value of the optimal value function. This represents the estimated weight vector of the critic network; Based on the estimated value of the optimal value function, the optimal control input is obtained as follows: in, This is an estimate of the optimal control input. The gradient of the activation function with respect to the local tracking error; Under an event-triggered mechanism, if the control input remains constant between consecutive trigger moments, then we have: in, This is an estimate of the optimal control input under the event-triggered mechanism. For the i-th following vehicle The timing of this event trigger; Based on the integral reinforcement learning method, for a given time period, the time difference error is defined as... in, The time difference vector of the activation function. for The performance cost over a given time period For time difference error, For time period; definition Define an instantaneous error function to represent the error in estimating commenter weights. Using gradient descent, the update law for the estimated weight vector of the critic network is: in, For learning rate, , It is an auxiliary variable.

6. The optimal vehicle queue control method based on function deviation triggering mechanism according to claim 5, characterized in that, Step five specifically includes: Set the communication relationship, initial state, and time period between the lead vehicle and the follower vehicle. Weighted matrix of local tracking error Weighted scalar of control input Learning rate Triggering mechanism first parameter The second parameter of the triggering mechanism Preset threshold parameters related to vehicle i Run steps one through four to obtain the queue control results.

7. A vehicle queuing optimal control system based on a function deviation triggering mechanism, characterized in that, The vehicle queuing optimal control system is used to execute the vehicle queuing optimal control method as described in any one of claims 1 to 6, the vehicle queuing optimal control system comprising: The model building module is used to build the navigation dynamics model of the pilot vehicle and A following dynamics model for a vehicle; The control input determination module is used to define the local tracking error of the following vehicle. According to the optimal control objective, the control input of the following dynamics model is defined as a function of the local tracking error, and then the optimal value function is defined to determine the optimal control input. The value function estimation module is used to obtain the optimal value function estimate through an adaptive critic neural network; The control input adjustment module is used to dynamically adjust the optimal control input of the following dynamics model based on the optimal value function estimate and the local tracking error design function deviation triggering mechanism. The queue control operation module is used to set the communication relationship and initial state between the lead vehicle and the follower vehicles. It sequentially calls the functions of the model building module, control input determination module, value function estimation module, and control input adjustment module to obtain the queue control results.

Images