Multi-agent cooperative traffic signal control method, system, device and medium

Abstract

This disclosure relates to a multi-agent collaborative traffic signal control method, system, device, and medium. The method includes: determining candidate green wave paths based on road traffic flow data, and determining green wave strategies for the candidate green wave paths; in response to each agent acquiring a green wave strategy, inputting the green wave strategy and real-time traffic data into a pre-trained decision model for analysis and processing to obtain a signal timing control strategy corresponding to each agent; for each agent, controlling the traffic lights corresponding to the agent during the transition period between the current time and the green wave initiation time according to the signal timing control strategy, and ensuring that the traffic lights turn on the green light for the specified direction at the green wave initiation time. According to the technical solution of this disclosure, global green wave optimization is achieved, improving road traffic efficiency, alleviating urban congestion, and enhancing the intelligence level of traffic signal control.

Description

Technical Field

[0001] This disclosure relates to the field of data processing technology, and in particular to a multi-agent collaborative traffic signal control method, system, device and medium. Background Technology

[0002] As urban traffic volume continues to rise, congestion problems are becoming increasingly prominent. Traditional traffic lights mostly use fixed durations or simple sensor control, which is difficult to cope with complex and ever-changing traffic conditions. For example, there are serious queues during peak hours, while there may be empty queues during off-peak hours, resulting in significant waste of resources. In addition, the complex structure of modern urban road networks makes it difficult to coordinate overall traffic with optimization of a single intersection, and local control is insufficient to achieve smooth traffic flow in a given area.

[0003] Currently, the development of artificial intelligence has driven the rise of intelligent transportation systems, creating an urgent need for a traffic signal control method to improve the intelligence level of traffic lights, optimize traffic flow distribution, and alleviate urban congestion. Summary of the Invention

[0004] To address the aforementioned technical issues, this disclosure provides a multi-agent collaborative traffic signal control method, system, device, and medium.

[0005] In a first aspect, embodiments of this disclosure provide a multi-agent cooperative traffic signal control method, including:

[0006] Candidate green wave paths are determined based on road traffic flow data, and green wave strategies for the candidate green wave paths are determined; the green wave strategy includes the green wave start time and duration.

[0007] In response to each agent acquiring the green wave strategy, the green wave strategy and real-time traffic data are input into a pre-trained decision model for analysis and processing to obtain the signal timing control strategy corresponding to each agent.

[0008] For each agent, the traffic light corresponding to the agent is controlled during the transition period between the current time and the green wave start time according to the signal timing control strategy, and the traffic light is turned on with a green light in the specified direction at the green wave start time.

[0009] Secondly, embodiments of this disclosure provide a multi-agent collaborative traffic signal control system, including: an edge computing center and multiple agents;

[0010] The edge computing center is used to determine candidate green wave paths based on road traffic data, and to determine the green wave strategy for the candidate green wave paths; the green wave strategy includes the green wave start time and duration.

[0011] The edge computing center is also used to send the green wave strategy to the multiple intelligent agents;

[0012] The intelligent agent is used to input the green wave strategy and real-time traffic data into a pre-trained decision model for analysis and processing to obtain a corresponding signal timing control strategy; according to the signal timing control strategy, the traffic lights are controlled during the transition period between the current time and the green wave start time, and the traffic lights are made to turn on the green light for the specified direction at the green wave start time.

[0013] Thirdly, embodiments of this disclosure provide an electronic device, including: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the multi-agent cooperative traffic signal control method described in the first aspect above.

[0014] Fourthly, embodiments of this disclosure provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the multi-agent cooperative traffic signal control method described in the first aspect.

[0015] Compared with the prior art, the technical solution provided in this disclosure has the following advantages: By adopting a multi-agent collaborative architecture, distributed intelligent control of traffic lights is realized. The edge computing center aggregates global traffic data and determines candidate green wave paths and green wave strategies, thereby improving decision-making efficiency and response capabilities. Each agent makes intelligent decisions through a decision model and controls the traffic lights corresponding to the agent according to the decision control strategy, thereby achieving global green wave optimization, improving road traffic efficiency, alleviating urban congestion, and enhancing the level of intelligence in traffic light control. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0017] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a multi-agent cooperative traffic signal control method provided in an embodiment of this disclosure;

[0019] Figure 2 This is a schematic diagram of a green wave strategy provided in an embodiment of the present disclosure;

[0020] Figure 3 A flowchart illustrating another multi-agent cooperative traffic signal control method provided in this embodiment of the present disclosure;

[0021] Figure 4 This is a schematic diagram of the structure of a multi-agent cooperative traffic signal control system provided in an embodiment of the present disclosure;

[0022] Figure 5 This is a schematic diagram of a single agent design provided in an embodiment of this disclosure. Detailed Implementation

[0023] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0024] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0025] Figure 1 This is a flowchart illustrating a multi-agent collaborative traffic signal control method provided in an embodiment of this disclosure. The method provided in this embodiment can be executed by a multi-agent collaborative traffic signal control device, which can be implemented in software and / or hardware and can be integrated on any electronic device with computing capabilities.

[0026] like Figure 1 As shown, the multi-agent cooperative traffic signal control method provided in this disclosure embodiment may include:

[0027] Step 101: Determine candidate green wave paths and green wave strategies for candidate green wave paths based on road traffic data.

[0028] The method of this disclosure is applied to a traffic signal control scenario, in which an edge computing center and multiple agents are set up to determine control strategies and execute control actions.

[0029] In this embodiment, the green wave strategy includes the green wave start time and duration. The edge computing center acquires road traffic flow data within the area. Optionally, based on real-time road traffic flow data, it identifies main roads with traffic flow exceeding a set threshold as candidate green wave paths. Then, it dynamically estimates the green wave start time and duration by combining the distance between stations on the candidate green wave paths and historical traffic speeds, thus achieving preliminary signal timing planning. After determining the green wave strategy, the edge computing center sends the green wave strategy to multiple agents, enabling the multiple agents to coordinate and control the traffic lights to achieve a continuous green wave effect.

[0030] As an example, multiple agents correspond to multiple traffic lights, with each agent controlling one traffic light. After determining candidate green wave paths, spatiotemporal collaborative estimation is performed. This estimation includes distance calculation, temporal planning, speed adaptation, and optimization objectives. The distance calculation step calculates the duration based on the distance between stations and historical traffic speeds. The temporal planning step calculates the green light activation timing for each traffic light on the candidate green wave path to ensure continuous traffic flow. The speed adaptation step adjusts the green wave width and tilt angle based on real-time traffic conditions. The optimization objective maximizes the green wave coverage to reduce the number of stops and waiting times. Figure 2 As shown.

[0031] Step 102: In response to each agent acquiring the green wave strategy, the green wave strategy and real-time traffic data are input into the pre-trained decision model for analysis and processing to obtain the signal timing control strategy corresponding to each agent.

[0032] In this embodiment, for each agent, a decision model is pre-trained. The input to the decision model includes collected traffic data, and the output is a traffic light timing control strategy. This strategy dynamically adjusts control parameters such as the phase duration and switching sequence of traffic lights, enabling refined regulation and dynamic optimization of traffic flow. The training process of the decision model includes: acquiring training sample data, which includes traffic data and corresponding labels for the traffic light phases; and training the decision model based on the training sample data.

[0033] As an example, the input to the decision model is traffic data and green wave constraints. When the agent receives a green wave policy, it determines the input green wave constraints according to the green wave policy. When the agent does not receive a green wave policy, the green wave constraints are set to default values ​​so that the decision model can make intelligent decisions based on the collected traffic data.

[0034] Step 103: For each agent, control the traffic light corresponding to the agent during the transition period between the current time and the start time of the green wave according to the signal timing control strategy, and make the traffic light turn on the green light of the specified direction at the start time of the green wave.

[0035] In this embodiment, each intelligent agent receives a coordination signal sent by the edge computing center and makes local dynamic adjustments. During the transition period, it intelligently schedules the phase changes of traffic lights based on real-time traffic data, while ensuring that the traffic lights turn on the green light in the designated direction at the start of the green wave, so as to achieve seamless connection of the green wave band across intersections. Through the coordinated control of multiple intelligent agents, a continuous green wave effect is achieved.

[0036] According to the technical solution of this disclosure embodiment, distributed intelligent control of traffic lights is realized by adopting a multi-agent collaborative architecture. The edge computing center aggregates global traffic data and determines candidate green wave paths and green wave strategies, thereby improving decision-making efficiency and response capabilities. Each agent makes intelligent decisions through a decision model and controls the traffic lights corresponding to the agent according to the decision control strategy, thereby realizing global green wave optimization, improving road traffic efficiency, alleviating urban congestion, and enhancing the intelligence level of traffic light control.

[0037] Based on the above embodiments, the decision model training process will be described below.

[0038] Figure 3 This is a flowchart illustrating another multi-agent cooperative traffic signal control method provided in an embodiment of this disclosure, as shown below. Figure 3 As shown, it includes the following steps:

[0039] Step 301: Obtain sample traffic data, input the sample traffic data into the language model for processing, and generate pseudo-true value labels for the corresponding traffic light phases.

[0040] In this embodiment, the decision model is used to analyze and process the collected real-time traffic data to generate the current optimal signal timing control strategy. Model training can be divided into three stages: data acquisition, model fine-tuning, and reinforcement learning. Training sample data for the decision model is constructed by generating pseudo-true value labels, eliminating the need for manual annotation.

[0041] In one embodiment of this disclosure, after generating pseudo-truth value labels for the corresponding traffic light phases, the shortest waiting time value for the corresponding action in the current state is calculated for each pseudo-truth value label based on the value evaluation mechanism of the Bellman equation. Then, the pseudo-truth value labels are cleaned according to the shortest waiting time value.

[0042] As an example, the traffic scene is analyzed using the GPT-4 model to directly generate pseudo-true value labels for the corresponding traffic light phases. In this example, to improve the effectiveness and decision-making quality of the pseudo-true value labels, a value evaluation mechanism based on the Bellman equation is further introduced. For each pseudo-true value label, the "shortest waiting time" (V value) of its corresponding action in the current state is calculated, and only the action label with the highest V value is retained. This achieves automated data cleaning and optimization, ensuring the policy guidance and quality stability of the training data.

[0043] Step 302: Construct training sample data based on sample traffic data and the pseudo-true value labels of the corresponding traffic light phases.

[0044] In this embodiment, a structured instruction construction strategy is adopted, designing each training sample as a four-segment prompt. The four-segment prompt consists of a role and task description, scene-aware information, decision-making basis, and common-sense supplement. The role and task description clarifies the model's responsibilities and objectives in the current context; the scene-aware information provides contextual information generated by an environment-aware module (such as vision or sensors); the decision-making basis lists key information and causal logic that can be used for reasoning or judgment; and the common-sense supplement completes the background knowledge and reasoning points that the model may lack.

[0045] As an example, consider the following: "You are a traffic management expert. You can use your traffic common sense to solve this traffic light control task.\n\n###Instructions:\nA traffic light controls a four-way intersection, including north, south, east, and west directions, each with two lanes: one for going straight and one for turning left. Each lane is further divided into three segments: segment 1 is closest to the intersection, segment 2 is in the middle, and segment 3 is furthest. Each lane may have two types of vehicles: 'early queue vehicles' that have arrived at the intersection and are waiting for the green light, and 'approaching vehicles' that have not yet arrived but are approaching the intersection.\n\nThis traffic signal has four phases, each phase allowing vehicles in two specific lane directions to pass. Below is the current state of the intersection, including:\n-The lane directions controlled by each phase;\n-The number of early queue vehicles in the allowed lanes;\n-The number of approaching vehicles in different segments of the allowed lanes."\n\nSignal:ETWT\nRelieves:Eastern and western through lanes\n-Early queued:0(East),0(West),0(Total)\n-Segment 1:0(East),0(West),0(Total)\n-Segment 2:0(East),0(West),0(Total)\n-Segment 3:0(East),0(West),0(Total)\n\nSignal:NTST\nRelieves:Northern and southern through lanes\n-Early queued:0(North),0(South),0(Total)\n-Segment 1:0(North),0(South),0(Total)\n-Segment 2:0(North),0(South),0(Total)\n-Segment 3:0(North),0(South),0(Total)\n\nSignal:ELWL\nRelieves:Eastern and western left-turn lanes\n-Early queued:0(East),0(West),0(Total)\n-Segment 1:0(East),0(West),0(Total)\n-Segment 2:0(East),0(West),0(Total)\n-Segment 3:0(East),0(West),0(Total)\n\nSignal:NLSL\nRelieves:Northern and southern left-turn lanes\n-Early queued:0(North),0(South),0(Total)\n-Segment 1:0(North),0(South),0(Total)\n-Segment 2:0(North),0(South),0(Total)\n-Segment 3:0 (North), 0 (South), 0 (Total)\n\nPlease answer:\nIn the next round of signal phases, which signal will be most effective in improving the current traffic situation?\n\nRequirements:\n-Let's think about it step by step.\n-You can only choose one signal from the list above.\n-You must provide your analysis following these steps:\nStep 1: Analysis, explaining how you determined the optimal signal;\nStep 2: Give your chosen signal.You must complete the analysis before making a selection. Your selection must be labeled with the following tags: <signal> Your choice< / signal> \n\n###Answer.

[0046] Step 303: Train the decision model based on the training sample data.

[0047] In this embodiment, LoRA (Low-Rank Adaptation, a model fine-tuning technique) is used for training, where the training objective is to minimize the negative log-likelihood. LoRA enables lightweight and efficient training, aiming to make the model's output distribution as close as possible to the decision distribution of human experts, thereby improving task perception and generalization capabilities.

[0048] Based on the above embodiments, the data acquisition will be described below.

[0049] In one embodiment of this disclosure, real-time traffic data is dynamically acquired by sensing devices deployed at intersections; the real-time traffic data includes traffic flow and vehicle queue length.

[0050] In this embodiment, acquiring real-time traffic data includes: processing traffic images captured by a camera based on target detection and slice-assisted fusion algorithms to obtain vehicle target detection and recognition results; intelligently segmenting lane lines in the traffic images to extract lane boundary ranges, determining the current driving lane and position relationship of the vehicle based on the lane boundary ranges; and completing and enhancing the current visual information by combining historical traffic record data.

[0051] As an example, high-precision perception of the traffic environment is achieved based on visual information fusion and intelligent analysis technology. In this example, vehicle target detection and recognition utilize a fusion algorithm based on DETR (Detection Transformer) and SAHI (Slicing Aided Hyper Inference) to process images captured by cameras, achieving high-precision identification and localization of vehicles on the road, improving detection accuracy and real-time performance in dense or long-distance scenes. Lane segmentation and boundary extraction use a deep learning model to intelligently segment lane lines in the image, extracting lane boundary ranges to help determine the vehicle's current lane and position. Multi-source information fusion and completion combines historical traffic record data to complete and enhance current visual information, improving perception robustness and completeness under conditions of limited camera field of view, occlusion, or low illumination. Through these steps, comprehensive and accurate perception of intersection and road traffic conditions is achieved, providing crucial support for the generation of subsequent control strategies.

[0052] This disclosure also proposes a multi-agent cooperative traffic signal control system. Figure 4 This is a schematic diagram of the structure of a multi-agent cooperative traffic signal control system provided in an embodiment of the present disclosure, as shown below. Figure 4 As shown, the multi-agent collaborative traffic signal control system includes an edge computing center and multiple agents.

[0053] The edge computing center is used to determine candidate green wave paths and green wave strategies based on road traffic data. The green wave strategy includes the green wave initiation time and duration. The edge computing center also sends the green wave strategy to multiple agents.

[0054] The intelligent agent is used to input the green wave strategy and real-time traffic data into a pre-trained decision model for analysis and processing to obtain the corresponding signal timing control strategy; according to the signal timing control strategy, the traffic lights are controlled during the transition period between the current time and the start time of the green wave, and the traffic lights are turned on with green lights for the specified direction at the start time of the green wave.

[0055] As an example, for N traffic lights in an area, N+1 intelligent agents are set up, one of which acts as an edge computing center, and the N agents correspond to the N traffic lights. For any given agent, it includes a real-time status monitoring module, an intelligent decision-making module, and a signal execution control module. The real-time status monitoring module dynamically collects key traffic parameters such as traffic flow and vehicle queue length through sensors deployed at intersections to achieve real-time monitoring of road operation. The intelligent decision-making module analyzes and processes the collected real-time traffic data to generate the optimal signal timing control strategy. The signal execution control module dynamically adjusts control parameters such as the phase duration and switching sequence of the traffic lights based on the intelligent decision-making results, achieving refined control and dynamic optimization of traffic flow. In this example, refer to... Figure 5 The text signals include traffic light phases, vehicle conditions on the road, and global dispatch information. The decision-making model can be implemented based on deep neural networks. The intelligent decision-making results represent phases, phase durations, switching sequences, etc., for example... Figure 5 Phases P1, P2, P3, and P4.

[0056] This disclosure also proposes a multi-agent collaborative traffic signal control device, comprising: a determination module, a decision module, and a control module.

[0057] The determination module is used to determine candidate green wave paths based on road traffic data, and to determine the green wave strategy for the candidate green wave paths; the green wave strategy includes the green wave start time and duration.

[0058] The decision module is used to respond to each agent acquiring the green wave strategy, input the green wave strategy and real-time traffic data into the pre-trained decision model for analysis and processing, and obtain the signal timing control strategy corresponding to each agent.

[0059] The control module is used to control the traffic lights corresponding to each agent during the transition period between the current time and the green wave start time according to the signal timing control strategy, and to make the traffic lights turn on the green light in the specified direction at the green wave start time.

[0060] The multi-agent cooperative traffic signal control device provided in this disclosure can execute any multi-agent cooperative traffic signal control method provided in this disclosure, and has the corresponding functional modules and beneficial effects for executing the method. Content not described in detail in the device embodiments of this disclosure can be referred to the description in any method embodiment of this disclosure.

[0061] This disclosure also provides an electronic device including one or more processors and a memory. The processor may be a central processing unit (CPU) or other processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. The memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and / or cache memory. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor may execute the program instructions to implement the methods of the embodiments of this disclosure above and / or other desired functions. Various contents such as input signals, signal components, and noise components may also be stored in the computer-readable storage medium.

[0062] In one example, the electronic device may also include input and output devices, which are interconnected via a bus system and / or other forms of connection. Furthermore, the input device may include, for example, a keyboard, a mouse, etc. The output device can output various information to the outside, including determined distance information, direction information, etc. The output device may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc. In addition, depending on the specific application, the electronic device may include any other suitable components such as a bus, input / output interfaces, etc.

[0063] In addition to the methods and apparatus described above, embodiments of this disclosure may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform any of the methods provided in the embodiments of this disclosure.

[0064] Computer program products can be written in any combination of one or more programming languages ​​to perform the operations of embodiments of this disclosure. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on a user's computing device, partially on a user's computing device, as a standalone software package, partially on a user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0065] Furthermore, embodiments of this disclosure may also be computer-readable storage media storing computer program instructions that, when executed by a processor, cause the processor to perform any of the methods provided in the embodiments of this disclosure.

[0066] Computer-readable storage media may take the form of any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may, for example, include, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: electrical connections having one or more wires, portable disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0067] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0068] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A multi-agent cooperative traffic signal control method, characterized in that, The method includes: Candidate green wave paths are determined based on road traffic flow data, and green wave strategies for the candidate green wave paths are determined; the green wave strategy includes the green wave start time and duration. In response to each agent acquiring the green wave strategy, the green wave strategy and real-time traffic data are input into a pre-trained decision model for analysis and processing to obtain the signal timing control strategy corresponding to each agent. For each agent, the traffic light corresponding to the agent is controlled during the transition period between the current time and the green wave start time according to the signal timing control strategy, and the traffic light is turned on with a green light in the specified direction at the green wave start time. The method further includes: Acquire sample traffic data, input the sample traffic data into a language model for processing, and generate pseudo-true value labels for the corresponding traffic light phases; Training sample data is constructed based on the sample traffic data and the pseudo-true value labels of the corresponding traffic light phases; The decision model is trained based on the training sample data; Each training sample uses a four-segment prompt, which consists of a role and task description, scene awareness information, decision-making basis, and common sense supplement. Training the decision model based on the training sample data includes: LoRA is used for training; the training objective is to minimize the negative log-likelihood.

2. The method as described in claim 1, characterized in that, After generating the pseudo-truth value label for the corresponding traffic light phase, the method further includes: Based on the value assessment mechanism of Bellman equation, the shortest waiting time value of the corresponding action in the current state is calculated for each pseudo-true value label; Data cleaning is performed on the pseudo-true value labels based on the shortest waiting time value.

3. The method as described in claim 1, characterized in that, The process of determining candidate green wave paths based on road traffic flow data, and determining the green wave strategy for the candidate green wave paths, includes: Based on real-time road traffic data, main roads with traffic flow exceeding a set threshold are identified as candidate green wave paths; By combining the distance between stations on the candidate green wave path and the historical travel speed, the start time and duration of the green wave are dynamically estimated.

4. The method as described in claim 1, characterized in that, Before inputting the green wave strategy and real-time traffic data into the pre-trained decision model for analysis and processing, the method further includes: The real-time traffic data is dynamically acquired by sensors deployed at intersections; the real-time traffic data includes traffic flow and vehicle queue length.

5. The method as described in claim 4, characterized in that, The acquisition of the real-time traffic data includes: Based on the target detection and slice-assisted fusion algorithm, traffic images captured by the camera are processed to obtain vehicle target detection and recognition results; The lane lines in the traffic image are intelligently segmented to extract the lane boundary range, and the current driving lane and position relationship of the vehicle are determined based on the lane boundary range. The current visual information is supplemented and enhanced by combining historical traffic record data.

6. A multi-agent cooperative traffic signal control system, characterized in that, include: Edge computing centers and multiple intelligent agents; The edge computing center is used to determine candidate green wave paths based on road traffic data, and to determine the green wave strategy for the candidate green wave paths; the green wave strategy includes the green wave start time and duration. The edge computing center is also used to send the green wave strategy to the multiple intelligent agents; The intelligent agent is used to input the green wave strategy and real-time traffic data into a pre-trained decision model for analysis and processing to obtain a corresponding signal timing control strategy; and to control the traffic lights according to the signal timing control strategy during the transition period between the current time and the green wave start time, and to make the traffic lights turn on the green light for the specified direction at the green wave start time. The system also includes: Acquire sample traffic data, input the sample traffic data into a language model for processing, and generate pseudo-true value labels for the corresponding traffic light phases; Training sample data is constructed based on the sample traffic data and the pseudo-true value labels of the corresponding traffic light phases; The decision model is trained based on the training sample data; Each training sample uses a four-segment prompt, which consists of a role and task description, scene awareness information, decision-making basis, and common sense supplement. Training the decision model based on the training sample data includes: LoRA is used for training; the training objective is to minimize the negative log-likelihood.

7. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the multi-agent cooperative traffic signal control method according to any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, implements the multi-agent cooperative traffic signal control method as described in any one of claims 1-5.

Images