Unmanned vehicle control system based on multi-agent cooperative control technology

Abstract

The present invention relates to the field of unmanned driving control technology, and discloses a vehicle unmanned driving control system based on multi-agent collaborative control technology. The system regards unmanned vehicles, traffic signals and road facilities as independent agents, and realizes unmanned driving and traffic collaborative control through information sharing and collaborative decision-making. The vehicle agent uses a variety of sensors to collect environmental information and make path planning and speed decisions; the traffic signal agent adjusts the signal light time according to vehicle information and traffic flow; the road facility agent adjusts its own state according to the vehicle situation. The central control module integrates the information of each agent and coordinates and optimizes the decision. The use of multi-agent collaborative control and distributed architecture improves the system reliability, adaptability and flexibility, and adapts to different traffic scenarios and road condition changes.

Description

Technical Field

[0001] The present invention relates to the field of unmanned driving control technology, and in particular to a vehicle unmanned driving control system based on multi-agent collaborative control technology. Background Art

[0002] Autonomous driving refers to the use of on-board sensors, controllers, actuators and other systems to enable vehicles to automatically and safely navigate and drive without direct human driver operation. With the continuous advancement of science and technology, autonomous driving technology has become a research hotspot in the field of transportation. However, existing autonomous driving technology solutions still have some problems in practical applications.

[0003] First, current unmanned driving technologies mostly focus on the autonomous driving capabilities of a single vehicle, lacking coordinated control of the vehicle and other elements in the traffic environment.

[0004] There is a lack of effective information exchange and collaborative decision-making between vehicles and traffic signals. In existing technologies, driverless vehicles can usually only passively receive instructions from traffic signals, and cannot interact with traffic signals to adjust according to real-time traffic conditions. For example, when traffic flow changes, traffic signals cannot adjust the signal time according to the distribution of vehicles in a timely manner, resulting in traffic congestion or long waiting time for vehicles.

[0005] Insufficient coordination between vehicles and road facilities. In the existing technology, road facilities are often static and cannot dynamically adjust their status according to the location and driving direction of the vehicle to provide a better driving environment for the vehicle. For example, street lights cannot automatically adjust their brightness according to the approach of the vehicle, and traffic signs cannot be displayed adaptively according to the needs of the vehicle.

[0006] Secondly, most of the current unmanned driving systems adopt a centralized control architecture, which has the following problems:

[0007] Low reliability. In a centralized control architecture, once the central control module fails, the entire system will be paralyzed. For example, when the central control server fails, all driverless vehicles may lose control, resulting in serious safety accidents.

[0008] Poor scalability. With the increase in the number of driverless vehicles and the continuous expansion of the transportation system, the centralized control architecture is difficult to meet the needs of large-scale systems. The addition of new vehicles and transportation facilities may bring huge computing pressure to the central control module, resulting in a decrease in system performance.

[0009] Insufficient response speed and real-time performance. In a centralized control architecture, all information needs to be transmitted to the central control module for processing, and then the decision results are fed back to each vehicle and facility. This information transmission and processing process will cause a long delay and cannot meet the real-time requirements of the unmanned driving system. For example, in an emergency, the vehicle may not be able to obtain instructions from the central control module in time, thus missing the best time to respond.

[0010] In summary, existing unmanned driving technologies have many problems in collaborative control and control architecture, and cannot meet the needs of increasingly complex traffic scenarios and changing road conditions. Summary of the invention

[0011] The main purpose of the present invention is to provide a vehicle unmanned driving control system based on multi-agent collaborative control technology, which can effectively solve the problems mentioned in the background technology.

[0012] To achieve the above object, the technical solution adopted by the present invention is:

[0013] The vehicle unmanned driving control system based on multi-agent cooperative control technology includes a vehicle agent, a traffic signal agent, a road facility agent and a central control module. The vehicle agent, the traffic signal agent and the road facility agent are respectively connected to the central control module in communication. The vehicle agent includes a sensor module, an information processing module, a decision-making planning module and a communication module. The traffic signal agent includes a data receiving module, a traffic condition evaluation module, a signal light time adjustment module and a communication module. The road facility agent includes a sensor module, a state adjustment module and a communication module. The central control module includes a communication adaptation module, an information analysis module, an information fusion module and a decision coordination module. The system realizes unmanned driving of the vehicle through multi-agent cooperative control technology. The multi-agent cooperative control technology includes three main steps: information interaction, cooperative decision-making and execution feedback, which are specifically:

[0014] ① Information interaction: The vehicle agent will preliminarily process and analyze the collected environmental information, extract key features, and send it to the communication network to share with other agents; the traffic signal agent will comprehensively analyze the current traffic conditions based on the received vehicle information and traffic flow data, determine the optimal switching time and cycle of the traffic lights, and share this information with the vehicle agent; the road facility agent will share its own location and status information with the vehicle agent to provide navigation and safety tips for the vehicle; the central control module receives information from all agents, integrates and comprehensively analyzes it, and provides a global perspective for collaborative control;

[0015] ② Collaborative decision-making: The vehicle agent uses intelligent decision-making algorithms to plan routes and make speed decisions based on shared information and its own goals; the traffic signal agent dynamically adjusts the signal time based on the distribution of vehicles and traffic flow; the road facility agent adjusts its own state based on the location and driving direction of the vehicle; the central control module coordinates and optimizes the decisions of all agents;

[0016] ③ Execution feedback: The vehicle agent executes driving operations according to the decision results to realize unmanned driving, continuously monitors its own status and changes in the surrounding environment, and sends feedback information to the communication network; the traffic signal agent adjusts the status of the traffic lights according to the decision, continuously monitors changes in traffic flow and sends feedback information to the communication network; the road facility agent performs corresponding adjustment operations, monitors vehicle feedback and sends information to the communication network; the central control module receives feedback information for real-time adjustment and optimization, continuously supervises the decision execution of each agent, evaluates system performance, processes feedback information and optimizes decisions.

[0017] Preferably, in the vehicle agent, the sensor module is used to collect information about the vehicle's surrounding environment, the information processing module is used to preprocess and extract features from the sensor data, the decision-making and planning module is used to make path planning and speed decisions based on shared information and its own goals, and the communication module is used to exchange information with other agents; wherein:

[0018] The sensor module includes a laser radar, a camera, and a millimeter-wave radar. The laser radar measures the distance and angle of the target object, the camera obtains image information, and the millimeter-wave radar detects the relative speed of the target. The information processing module uses a median filter to remove noise points in the laser radar data, uses a Gaussian filter to smooth the camera image, uses a convolutional neural network to extract camera image features, uses a clustering algorithm to extract target features from the laser radar point cloud data, extracts target relative speed features from the millimeter-wave radar data, and performs data compression;

[0019] The decision-making planning module adopts a deep reinforcement learning algorithm for path planning and speed decision-making. The state space is defined as a vector composed of the vehicle's current position, speed, direction, and surrounding environment information. The action space includes acceleration, deceleration, and steering operations. The reward function design includes the weighted sum of rewards for approaching the target, rewards for complying with traffic rules, and rewards for avoiding collisions.

[0020] Preferably, the reward function is specifically designed as follows: a positive reward is given when the vehicle approaches the target location, obeys traffic rules, keeps a safe distance from other vehicles, and drives in a well-lit and clear traffic sign environment, and a negative reward is given otherwise; the reward function It is expressed as the weighted sum of multiple reward items, namely:

[0021] = + +

[0022] in , , , is the weight coefficient, is the reward for approaching the target, designed to Rewards for obeying traffic rules; It is a reward for avoiding collisions, based on the distance to surrounding vehicles; It is a driving environment reward.

[0023] Preferably, in the traffic signal agent, the data receiving module is used to receive vehicle information and traffic flow data, the traffic condition evaluation module is used to evaluate the current traffic condition, the signal light time adjustment module is used to determine the optimal switching time and cycle of the signal light, and the communication module is used to exchange information with other agents; wherein:

[0024] The traffic condition assessment module estimates traffic density and flow using a macro traffic flow model and determines traffic conditions based on vehicle information and traffic flow data. The signal light time adjustment module uses an optimization algorithm to determine the optimal switching time and cycle of signal lights with the objective function of minimizing the average vehicle delay time.

[0025] Preferably, the traffic condition assessment module uses the LWR (Lighthill-Whitham-Richards) model to assess the current traffic condition; the basic equation of the LWR model is: ,in represents the traffic density, Indicates the traffic flow, Indicates time, Indicates spatial location;

[0026] The optimal switching time and cycle of the traffic light are determined as follows: Assuming that minimizing the average delay time of vehicles is the objective function, the fitness function is expressed as: ,in Indicates the green light time. Indicates yellow light time. Indicates the signal light cycle, Indicates the number of vehicles, Indicates The delay time of the vehicle.

[0027] Preferably, in the road facility agent, the sensor module is used to collect its own position and state information, the state adjustment module is used to adjust its own state according to the vehicle position and driving direction, and the communication module is used to exchange information with other agents; wherein:

[0028] The state adjustment module adjusts the brightness of the street lamp according to the vehicle position and light intensity, and adjusts the display angle and brightness of the traffic sign according to the vehicle driving direction and distance.

[0029] Preferably, the street light brightness is adjusted by using a photosensitive sensor to detect the ambient light intensity, and the street light brightness is adjusted according to the position of the vehicle and the light intensity; the light intensity is set to , the threshold is , the street light brightness is , then use the following rules to adjust the street light brightness: if < And when the vehicle approaches, the brightness of the street light is increased, which is expressed as +Δ , where Δ Adjust the step size for brightness;

[0030] The traffic sign display adjustment is to ensure that the traffic sign is displayed clearly, and adjust the display angle and brightness of the traffic sign according to the vehicle's driving direction and distance; if the vehicle is driving directly towards the traffic sign, increase the brightness of the sign; if the vehicle approaches the traffic sign from the side, adjust the display angle of the sign.

[0031] Preferably, in the central control module, the communication adaptation module is used to adapt the communication protocols of different intelligent agents, the information analysis module is used to analyze the information of each intelligent agent, the information fusion module is used to fuse the information of different intelligent agents, and the decision coordination module is used to coordinate and optimize the decisions of all intelligent agents; wherein:

[0032] The information fusion module uses Bayesian estimation to perform information fusion, and the fused information is sent to each intelligent agent after being displayed in a map or chart. The decision coordination module coordinates the decisions of different intelligent agents, optimizes the decisions of the entire system based on the comprehensive analysis results, and establishes a feedback mechanism to feed back the coordinated and optimized decisions to each intelligent agent.

[0033] Preferably, Bayesian estimation is used for information fusion. and , and their probability distributions are and , according to Bayes' theorem, the probability distribution after fusion is .

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] The present invention adopts multi-agent collaborative control technology, treating unmanned vehicles, traffic signals and road facilities as independent agents, and realizes the collaborative control of unmanned vehicles and traffic through information sharing and collaborative decision-making. This method can improve the reliability, adaptability and flexibility of the system and adapt to different traffic scenarios and road condition changes.

[0036] The present invention adopts a distributed control architecture to disperse the functions of the central control module to each intelligent agent to achieve distributed control. This method can improve the reliability and scalability of the system and avoid the impact of single point failure. At the same time, the distributed control architecture can also improve the response speed and real-time performance of the system to meet the requirements of the unmanned driving system. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is a schematic diagram of the system composition architecture flow of the present invention;

[0038] Figure 2 It is a schematic diagram of the information interaction process in the multi-agent collaborative control technical solution of the present invention;

[0039] Figure 3 It is a schematic diagram of the collaborative decision-making process in the multi-agent collaborative control technical solution of the present invention;

[0040] Figure 4 A schematic diagram of the feedback process executed in the multi-agent collaborative control technical solution of the present invention. DETAILED DESCRIPTION

[0041] In order to make the technical means, creative features, objectives and effects of the present invention easy to understand, the present invention is further described below in conjunction with specific implementation methods. The vehicle unmanned driving control system of this embodiment is intended to achieve unmanned driving of vehicles and efficient collaborative management of traffic through multi-agent collaborative control technology. The system regards unmanned vehicles, traffic signals and road facilities as independent intelligent entities, and improves the reliability, adaptability and flexibility of the system through information sharing and collaborative decision-making.

[0042] like Figure 1-Figure 4 Shown

[0043] System composition

[0044] 1. Vehicle Agent:

[0045] Sensor module: equipped with multiple sensors such as laser radar, camera, millimeter wave radar, etc. Laser radar can accurately measure the distance and angle of surrounding objects and provide the vehicle with three-dimensional information of the environment. Camera can obtain rich image information for identifying targets such as vehicles, pedestrians, and traffic signs. Millimeter wave radar is good at detecting the relative speed of the target.

[0046] Information processing module: Preprocess the data collected by the sensor. For example, median filtering is used to remove noise points in the lidar data to make the measurement results more accurate. For camera images, Gaussian filtering is used for smoothing to reduce noise interference in the image. Then, convolutional neural networks (CNNs) are used to extract image features, clustering algorithms are used to extract target features from lidar point cloud data, and features such as target relative speed are extracted from millimeter-wave radar data. Data compression is performed to reduce the amount of transmission.

[0047] Decision-making and planning module: Use deep reinforcement learning algorithms for path planning and speed decision-making. The state space is defined as a vector consisting of the vehicle's current position, speed, direction, and surrounding environment information. The action space includes operations such as acceleration, deceleration, and steering. The reward function design takes into account multiple factors, such as rewards for approaching the target, rewards for complying with traffic rules, and rewards for avoiding collisions. For example, when the vehicle approaches the target position, a positive reward is given, if it violates traffic rules (such as running a red light, speeding, etc.), a negative reward is given, and when the distance to the surrounding obstacles is less than the safe distance, a penalty is given, otherwise a reward is given.

[0048] Communication module: Use lightweight communication protocols (such as MQTT, CoAP, etc.) to exchange information with other agents. Send key features such as vehicle location, speed, obstacle information, etc. to other agents, and receive information from traffic signal agents and road facility agents.

[0049] 2. Traffic Signal Agent:

[0050] Data receiving module: Receives information such as the position and speed of the vehicle agent and traffic flow data from traffic flow monitoring equipment. For example, traffic flow information is obtained in real time through devices such as loop coil detectors and video detectors.

[0051] Traffic condition assessment module: Use macro traffic flow models (such as LWR model) to assess current traffic conditions. Estimate traffic density and flow based on vehicle information and traffic flow data. For example, traffic density is calculated by the number of vehicles and road length. The formula is: ,in Indicates the number of vehicles, represents the length of the road, where represents the number of vehicles and represents the length of the road.

[0052] Traffic light time adjustment module: Use optimization algorithms (such as genetic algorithms, simulated annealing algorithms, etc.) to determine the optimal switching time and cycle of traffic lights. The objective function is to minimize the average delay time of vehicles, for example ,in Indicates the green light time. Indicates yellow light time. Indicates the signal light cycle, Indicates the number of vehicles, Indicates The delay time of the vehicle.

[0053] Communication module: Uses communication protocols to exchange information with other intelligent agents, sends information such as traffic light status, optimal switching time and cycle to vehicle intelligent agents, and receives vehicle distribution information and traffic flow data.

[0054] 3. Road facility agent:

[0055] Sensor module: The light intensity is detected by the light sensor, and the position sensor determines its own position. For example, the light sensor can automatically adjust the brightness of the street light according to the changes in the ambient light.

[0056] State adjustment module: adjusts the state of the vehicle according to its position and direction of travel. For example, when the vehicle approaches a street light, the brightness of the street light is adjusted according to the light intensity and the vehicle's speed; when the vehicle approaches a traffic sign, the traffic sign is ensured to be clearly displayed.

[0057] Communication module: uses communication protocols to exchange information with vehicle agents, shares its own location and status information with vehicle agents, and receives feedback information from vehicles.

[0058] 4. Central control module:

[0059] Communication adapter module: adapts the communication protocols used by different agents in order to receive information from each agent. For example, communication middleware (such as MQTT proxy server, CoAP server, etc.) is used to achieve conversion and adaptation between different communication protocols.

[0060] Information parsing module: parses the received information and extracts key information of each agent. For example, the vehicle position, speed, obstacle information, etc. are parsed from the information sent by the vehicle agent; the signal light switching time and cycle are parsed from the information sent by the traffic signal agent; the street light brightness status, traffic sign location information, etc. are parsed from the information sent by the road facility agent.

[0061] Information fusion module: Use data fusion algorithms (such as Bayesian estimation, evidence theory, fuzzy logic, etc.) to fuse information from different agents. For example, when using Bayesian estimation for information fusion, for two information sources and , according to Bayes' theorem, the probability distribution after fusion is The fused information is presented in a suitable form.

[0062] Decision coordination module: Coordinate the decisions of different agents to ensure consistency and coordination. For example, if the path planned by the vehicle agent does not match the signal time of the traffic signal agent, the central control module can adjust the vehicle's path planning or notify the traffic signal agent to adjust the signal time to avoid conflict. Make optimization decisions for the entire system based on the comprehensive analysis results. For example, when traffic congestion in a certain area is serious, adjust the cycle and phase of the traffic signal to guide vehicle diversion; or adjust the status of road facilities to provide a better driving environment for vehicles. Establish a feedback mechanism to feed back the coordinated and optimized decisions to each agent.

[0063] System operation process

[0064] 1. Information collection and processing:

[0065] Vehicle agent: The vehicle agent collects information about the surrounding environment through the sensor module, and performs preprocessing and feature extraction through the information processing module. For example, after the laser radar data is filtered by the median filter to remove noise points, more accurate distance and angle information of the target object can be obtained; after the camera image is smoothed by the Gaussian filter, the convolutional neural network is used to extract features, and feature information of target objects such as vehicles, pedestrians, and traffic signs can be obtained.

[0066] Traffic signal agent: The traffic signal agent receives vehicle information and traffic flow data through the data receiving module, and uses the traffic condition evaluation module to evaluate the current traffic conditions. For example, according to the LWR model, the traffic density is estimated by the number of vehicles and road length, and the traffic flow is determined based on the traffic density and vehicle speed.

[0067] Road facility agent: The road facility agent collects and processes its own location and status information through sensor modules. For example, the light sensor detects the light intensity and determines the brightness status of the street light; the position sensor determines the location information of the traffic sign.

[0068] 2. Information interaction:

[0069] Vehicle agent: The vehicle agent sends the extracted key features to the communication network through the communication module to share with other agents. For example, the vehicle's location, speed, obstacle information, etc.

[0070] Traffic signal agent: The traffic signal agent determines the optimal switching time and cycle of the traffic light based on the received vehicle information and traffic flow data, and shares this information with the vehicle agent through the communication module. For example, it sends the status information of the traffic light (such as the current color of the traffic light, the remaining time, etc.) and the optimal switching time and cycle information.

[0071] Road Facility Agent: The road facility agent shares its location and status information with the vehicle agent through the communication module to provide navigation and safety tips for the vehicle. For example, it sends the brightness status of street lights and the location information of traffic signs.

[0072] Central control module: The central control module receives information from all intelligent agents, adapts different communication protocols through the communication adaptation module, then the information parsing module extracts key information, and then the information fusion module performs information fusion.

[0073] 3. Collaborative decision-making:

[0074] Vehicle Agent: The vehicle agent uses intelligent decision-making algorithms (such as deep reinforcement learning algorithms) to plan paths and make speed decisions based on shared information and its own goals (such as reaching the destination, complying with traffic rules, etc.). For example, the vehicle agent determines the best driving path and speed based on the location and speed of surrounding vehicles, traffic signal status, road facility information, etc.

[0075] Traffic signal agent: The traffic signal agent dynamically adjusts the signal time according to the distribution of vehicles and traffic flow to optimize traffic flow and reduce congestion. For example, when the queue of vehicles in a certain direction is long, the green light time in that direction is extended.

[0076] Road Facility Agent: The road facility agent adjusts its own state according to the vehicle's position and driving direction to provide a better driving environment for the vehicle. For example, when the vehicle approaches a street lamp, the brightness of the street lamp is increased.

[0077] Central Control Module: The central control module coordinates and optimizes the decisions of all agents to ensure the efficient operation of the overall system. For example, if the path planned by the vehicle agent does not match the signal time of the traffic signal agent, the central control module can adjust the vehicle's path planning or notify the traffic signal agent to adjust the signal time.

[0078] 4. Execution and feedback:

[0079] Vehicle agent: The vehicle agent performs operations such as acceleration, deceleration, and steering according to the decision results to achieve unmanned driving. At the same time, it continuously monitors its own status and changes in the surrounding environment, and sends feedback information to the communication network. For example, the vehicle agent performs corresponding operations based on the path planning and speed decision results, and monitors its own position, speed, acceleration and other status information and changes in the surrounding environment, such as the appearance of obstacles, through sensors, and feeds this information back to other agents and the central control module.

[0080] Traffic signal agent: The traffic signal agent adjusts the status of the traffic lights according to the decision, continuously monitors the changes in traffic flow, and sends feedback information to the communication network. For example, the traffic signal agent adjusts the display status of the traffic lights according to the decision results, monitors the changes in traffic flow at the intersection through traffic flow monitoring equipment, and feeds back the actual status of the traffic lights and traffic flow monitoring information to other agents and the central control module.

[0081] Road facility agent: The road facility agent performs corresponding adjustment operations, monitors vehicle feedback, and sends information to the communication network. For example, the road facility agent adjusts the brightness of street lights and the display status of traffic signs according to the location and driving direction of the vehicle, and receives feedback from vehicle agents, such as the demand for street light brightness and the evaluation of traffic sign clarity, and feeds back the actual status of road facilities and vehicle feedback information to other agents and the central control module.

[0082] Central control module: The central control module receives feedback from each intelligent agent and makes real-time adjustments and optimizations. For example, the central control module evaluates the operating status of the entire system based on the feedback from each intelligent agent, adjusts and optimizes the collaborative control strategy, and feeds back the coordinated and optimized decision to each intelligent agent.

[0083] In summary, the vehicle agent uses a variety of sensors to collect environmental information. After preprocessing and feature extraction, it uses a deep reinforcement learning algorithm to perform path planning and speed decisions, achieving precise control of the vehicle. The traffic signal agent uses a macro traffic flow model and optimization algorithm to determine the optimal switching time and cycle of the traffic light based on vehicle information and traffic flow data, effectively optimizing traffic flow and reducing congestion. The road facility agent collects its own state information through sensors, and adjusts its state according to the vehicle's position and driving direction, providing a better driving environment for the vehicle. The central control module ensures the consistency and coordination between the various agents through information fusion and decision coordination, and realizes the efficient operation of the entire system.

[0084] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. The vehicle unmanned driving control system based on multi-agent cooperative control technology is characterized by: The system comprises a vehicle agent, a traffic signal agent, a road facility agent and a central control module. The vehicle agent, the traffic signal agent and the road facility agent are respectively connected to the central control module in communication. The vehicle agent comprises a sensor module, an information processing module, a decision-making planning module and a communication module. The traffic signal agent comprises a data receiving module, a traffic condition evaluation module, a signal light time adjustment module and a communication module. The road facility agent comprises a sensor module, a state adjustment module and a communication module. The central control module comprises a communication adaptation module, an information analysis module, an information fusion module and a decision coordination module. The system realizes unmanned driving of vehicles through multi-agent collaborative control technology. The multi-agent collaborative control technology comprises three main steps: information interaction, collaborative decision-making and execution feedback, which are specifically: ① Information interaction: The vehicle agent will preliminarily process and analyze the collected environmental information, extract key features, and send them to the communication network to share with other agents; The traffic signal agent comprehensively analyzes the current traffic conditions based on the received vehicle information and traffic flow data, determines the optimal switching time and cycle of the traffic lights, and shares this information with the vehicle agent; the road facility agent shares its own location and status information with the vehicle agent, providing navigation and safety tips for the vehicle; the central control module receives information from all agents, performs fusion and comprehensive analysis, and provides a global perspective for collaborative control; ② Collaborative decision-making: The vehicle agent uses intelligent decision-making algorithms to plan routes and make speed decisions based on shared information and its own goals; the traffic signal agent dynamically adjusts the signal time based on the distribution of vehicles and traffic flow; the road facility agent adjusts its own state based on the location and driving direction of the vehicle; the central control module coordinates and optimizes the decisions of all agents; ③ Execution feedback: The vehicle agent performs driving operations according to the decision results to achieve unmanned driving, continuously monitors its own status and changes in the surrounding environment, and sends feedback information to the communication network; the traffic signal agent adjusts the status of the traffic light according to the decision, continuously monitors changes in traffic flow and sends feedback information to the communication network; the road facility agent performs corresponding adjustment operations, monitors vehicle feedback and sends information to the communication network; the central control module receives feedback information for real-time adjustment and optimization, continuously monitors the decision execution of each agent, evaluates system performance, processes feedback information and optimizes decisions; In the traffic signal agent, the data receiving module is used to receive vehicle information and traffic flow data, the traffic condition evaluation module is used to evaluate the current traffic condition, the signal light time adjustment module is used to determine the optimal switching time and cycle of the signal light, and the communication module is used to exchange information with other agents; wherein: The traffic condition assessment module estimates traffic density and flow using a macro traffic flow model, determines traffic conditions based on vehicle information and traffic flow data, and the signal light time adjustment module uses an optimization algorithm to minimize the average vehicle delay time as the objective function to determine the optimal signal light switching time and cycle; The traffic condition assessment module uses the LWR (Lighthill-Whitham-Richards) model to assess the current traffic condition. The basic equation of the LWR model is: ,in represents the traffic density, Indicates the traffic flow, Indicates time, Indicates spatial location; The optimal switching time and cycle of the traffic light are determined as follows: Assuming that minimizing the average delay time of vehicles is the objective function, the fitness function is expressed as: ,in Indicates the green light time. Indicates yellow light time. Indicates the signal light cycle, Indicates the number of vehicles, Indicates The delay time of the vehicle.

2. The vehicle unmanned driving control system based on multi-agent cooperative control technology according to claim 1 is characterized in that: In the vehicle agent, the sensor module is used to collect information about the vehicle's surrounding environment, the information processing module is used to preprocess and extract features from the sensor data, the decision-making and planning module is used to make path planning and speed decisions based on shared information and its own goals, and the communication module is used to exchange information with other agents; wherein: The sensor module includes a laser radar, a camera, and a millimeter-wave radar. The laser radar measures the distance and angle of the target object, the camera obtains image information, and the millimeter-wave radar detects the relative speed of the target. The information processing module uses a median filter to remove noise points in the laser radar data, uses a Gaussian filter to smooth the camera image, uses a convolutional neural network to extract camera image features, uses a clustering algorithm to extract target features from the laser radar point cloud data, extracts target relative speed features from the millimeter-wave radar data, and performs data compression; The decision-making planning module adopts a deep reinforcement learning algorithm for path planning and speed decision-making. The state space is defined as a vector composed of the vehicle's current position, speed, direction, and surrounding environment information. The action space includes acceleration, deceleration, and steering operations. The reward function design includes the weighted sum of rewards for approaching the target, rewards for complying with traffic rules, and rewards for avoiding collisions.

3. The vehicle unmanned driving control system based on multi-agent cooperative control technology according to claim 2 is characterized in that: The reward function is specifically designed as follows: a positive reward is given when the vehicle approaches the target location, obeys traffic rules, keeps a safe distance from other vehicles, and drives in an environment with good lighting and clear traffic signs, and a negative reward is given otherwise; the reward function It is expressed as the weighted sum of multiple reward items, namely: = + + in , , , is the weight coefficient, is the reward for approaching the target, designed to Rewards for obeying traffic rules; It is a reward for avoiding collisions, based on the distance to surrounding vehicles; It is a driving environment reward.

4. The vehicle unmanned driving control system based on multi-agent cooperative control technology according to claim 1 is characterized in that: In the road facility agent, the sensor module is used to collect its own position and state information, the state adjustment module is used to adjust its own state according to the vehicle position and driving direction, and the communication module is used to exchange information with other agents; wherein: The state adjustment module adjusts the brightness of the street lamp according to the vehicle position and light intensity, and adjusts the display angle and brightness of the traffic sign according to the vehicle driving direction and distance.

5. The vehicle unmanned driving control system based on multi-agent cooperative control technology according to claim 4 is characterized in that: The street light brightness is adjusted by using a photosensitive sensor to detect the ambient light intensity, and the street light brightness is adjusted according to the vehicle position and light intensity; Set the light intensity to , the threshold is , the street light brightness is , then use the following rules to adjust the street light brightness: If < And when the vehicle approaches, the brightness of the street light is increased, which is expressed as +Δ , where Δ Adjust the step size for brightness; The traffic sign display adjustment: ensure that the traffic sign is displayed clearly, and adjust the display angle and brightness of the traffic sign according to the driving direction and distance of the vehicle; If the vehicle is driving towards the traffic sign directly, increase the brightness of the sign; if the vehicle is approaching the traffic sign from the side, adjust the display angle of the sign.

6. The vehicle unmanned driving control system based on multi-agent cooperative control technology according to claim 1 is characterized in that: In the central control module, the communication adaptation module is used to adapt the communication protocols of different intelligent agents, the information analysis module is used to analyze the information of each intelligent agent, the information fusion module is used to fuse the information of different intelligent agents, and the decision coordination module is used to coordinate and optimize the decisions of all intelligent agents; wherein: The information fusion module uses Bayesian estimation to perform information fusion, and the fused information is sent to each intelligent agent after being displayed in a map or chart. The decision coordination module coordinates the decisions of different intelligent agents, optimizes the decisions of the entire system based on the comprehensive analysis results, and establishes a feedback mechanism to feed back the coordinated and optimized decisions to each intelligent agent.

7. The vehicle unmanned driving control system based on multi-agent cooperative control technology according to claim 6 is characterized in that: Using Bayesian estimation for information fusion, for two information sources and , and their probability distributions are and , according to Bayes' theorem, the probability distribution after fusion is .

Images