Traffic conflict point active collision avoidance control method and system

Abstract

The invention relates to the technical field of traffic avoidance, in particular to a traffic conflict point active collision avoidance control method and system, and the method comprises the steps: carrying out the dynamic conflict point recognition of a road intersection, and obtaining a real-time traffic conflict point distribution diagram; performing behavior feature extraction on the traffic participants according to the real-time traffic conflict point distribution map to obtain dynamic state data of the traffic participants; inputting the dynamic state data of the traffic participants into a traffic conflict risk assessment network for risk quantification processing to obtain a traffic conflict risk assessment result; and performing conflict development trend prediction processing on the traffic conflict risk assessment result to obtain a traffic conflict evolution prediction result. Through cooperative control with infrastructures such as a traffic signal control system and a traffic guidance system, the traffic light time length, lane distribution and the like can be intelligently optimized, the efficiency of a traffic management system is improved, and the problems of traffic bottleneck and congestion are avoided.

Description

Technical Field

[0001] This invention relates to the field of traffic avoidance technology, specifically to an active collision avoidance control method and system for traffic conflict points. Background Technology

[0002] Traffic conflict refers to a situation in a traffic environment where two or more traffic participants (such as vehicles, pedestrians, non-motorized vehicles, etc.) have a potential collision or interference in the same or adjacent space. It usually refers to a conflict caused by poor traffic behavior, traffic signals, road conditions or other factors, which has a certain degree of danger, but does not necessarily lead to a traffic accident immediately. The occurrence of traffic conflict usually indicates the potential risk of traffic accident.

[0003] Currently, traditional methods typically rely on fixed traffic flow data and predetermined traffic patterns, lacking real-time monitoring and analysis of the dynamic status of traffic participants. As a result, traditional methods often fail to effectively predict and intervene in potential risks before conflicts occur, leading to a failure to respond to traffic conflicts in a timely manner in some cases, thus increasing the probability of accidents.

[0004] Furthermore, traditional methods typically optimize for a single traffic participant or traffic signal, lacking a mechanism for multi-objective coordinated control. Therefore, in complex traffic environments, traditional methods struggle to effectively coordinate the behavior of multiple traffic participants, easily leading to vehicle conflicts and mutual interference, thus reducing traffic flow and safety. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution: a method for active collision avoidance control at traffic conflict points, comprising: Dynamic conflict point identification is performed at road intersections to obtain a real-time traffic conflict point distribution map; Based on the real-time traffic conflict point distribution map, behavioral features of traffic participants are extracted to obtain dynamic state data of traffic participants, wherein the dynamic state data of traffic participants includes velocity vector, acceleration trend and motion trajectory; The dynamic status data of the traffic participants are input into the traffic conflict risk assessment network for risk quantification to obtain the traffic conflict risk assessment result; the traffic conflict risk assessment result is then processed to predict the conflict development trend to obtain the traffic conflict evolution prediction result, wherein the traffic conflict evolution prediction result includes the risk spread speed and the scope of impact. The traffic conflict evolution prediction results are subjected to collision avoidance strategy priority processing to obtain a priority sequence; the priority sequence is then subjected to multi-objective coordinated control processing to obtain active collision avoidance control decision results. The real-time traffic conflict point distribution map, the dynamic status data of traffic participants, and the active collision avoidance control decision results are processed in real time to obtain the active collision avoidance control execution results for traffic conflict points.

[0006] Preferably, dynamic conflict point identification is performed at road intersections to obtain a real-time traffic conflict point distribution map, including: Traffic data is collected, and the traffic data is subjected to spatiotemporal calibration processing to obtain multi-source sensing data; The multi-source sensing data is subjected to feature extraction processing to obtain a basic sensing feature map, wherein the basic sensing feature map includes a vehicle detection box, a speed estimate, and a motion direction angle; The basic perception feature map is subjected to feature fusion processing to obtain a traffic scene feature representation; The traffic scene feature representation is subjected to traffic participant relationship modeling processing to obtain a traffic scene relationship graph, wherein the traffic scene relationship graph includes the interaction relationship between vehicles; The potential conflict points are calculated and processed on the traffic scene relationship diagram to obtain a real-time traffic conflict point distribution map.

[0007] Preferably, behavioral features of traffic participants are extracted based on the real-time traffic conflict point distribution map to obtain dynamic state data of traffic participants, including: Based on the real-time traffic conflict point distribution map, target tracking processing is performed on traffic participants to obtain continuous movement trajectory data for each traffic participant; The motion trajectory data is smoothed and filtered to remove noise interference and fill in data gaps, resulting in an optimized trajectory sequence. The velocity vectors of traffic participants are calculated based on the optimized trajectory sequence, wherein the velocity vectors include the velocity magnitude and the direction angle of motion; The acceleration trend of traffic participants, including longitudinal acceleration and lateral acceleration components, is calculated based on the velocity vector. The acceleration trend is subjected to feature analysis to identify abnormal driving behavior patterns, including rapid acceleration, rapid deceleration and sharp turns. The velocity vector, acceleration trend, motion trajectory, and abnormal behavior pattern are integrated and processed to obtain dynamic state data of traffic participants.

[0008] Preferably, the dynamic status data of the traffic participants is input into a traffic conflict risk assessment network for risk quantification to obtain traffic conflict risk assessment results, including: The dynamic status data of the traffic participants are input into the spatiotemporal correlation analysis unit for conflict pattern recognition processing to obtain a preliminary risk assessment vector. The preliminary risk assessment vector is subjected to multi-factor risk fusion processing to obtain a comprehensive risk assessment matrix; The comprehensive risk assessment matrix is ​​input into the cascaded risk level classification unit for risk level determination, and the traffic conflict risk assessment result is obtained.

[0009] Preferably, the traffic conflict risk assessment results are processed to predict the conflict development trend, resulting in a traffic conflict evolution prediction result, including: Acquire dynamic behavioral characteristic data of traffic participants and road network topology data, wherein the dynamic behavioral characteristic data of traffic participants includes driving behavior pattern characteristics associated with vehicle driving trajectory, spatiotemporal distribution characteristics of traffic conflict points, and traffic flow state evolution characteristics. Based on the spatiotemporal distribution characteristics of the traffic conflict points, high-incidence areas of traffic conflicts are determined, and based on the driving behavior pattern characteristics, abnormal driving behavior identification criteria are determined. The high-incidence areas of traffic conflicts are input into a preset spatiotemporal correlation feature extraction network to obtain a dynamic risk assessment map; The dynamic risk assessment map is optimized to obtain an optimized dynamic risk assessment map; Based on the optimized dynamic risk assessment map, timing optimization parameters for the intelligent traffic signal control system and path planning adjustment coefficients for the traffic guidance system are generated, resulting in traffic conflict evolution prediction results.

[0010] Preferably, the traffic conflict evolution prediction results are subjected to collision avoidance strategy priority processing to obtain a priority sequence, including: The traffic conflict evolution prediction results are converted into an initial strategy representation space; High-risk conflict scenarios are identified using a conflict risk assessor, and the feasibility of strategies for these high-risk conflict scenarios is verified. Establish a strategy evaluation model and combine it with dynamic weight adjustment to solve for the optimal strategy combination under multiple constraints; The optimal strategy combination is sorted to obtain a priority sequence.

[0011] Preferably, the priority sequence is subjected to multi-objective coordinated control processing to obtain active collision avoidance control decision results, including: The traffic participants are prioritized according to the priority sequence to determine the key vehicles that need to be controlled first. Vehicle control commands are generated for the key vehicle, wherein the control parameters include the desired acceleration, steering angle, and braking pressure; The vehicle's safe driving trajectory is calculated based on the vehicle control commands to ensure driving stability while avoiding collisions; For non-critical vehicles, cooperative collision avoidance suggestions are generated, including auxiliary control commands such as deceleration prompts, lane keeping, and following distance adjustments. Based on the traffic signal control system, a signal timing optimization scheme is generated to adjust the duration and phase sequence of traffic lights; It generates control commands for variable lanes, speed limit signs, and warning devices to achieve coordinated control of infrastructure and vehicles; The vehicle control commands, the safe driving trajectory, the cooperative collision avoidance suggestions, and the infrastructure control commands are fused together to obtain the active collision avoidance control decision result.

[0012] Preferably, the real-time traffic conflict point distribution map, the dynamic state data of traffic participants, and the active collision avoidance control decision results are processed in real-time to obtain the active collision avoidance control execution results for traffic conflict points, including: The conflict point location information in the real-time traffic conflict point distribution map is synchronized with the location trajectory information in the dynamic status data of traffic participants on the time axis. Spatial range matching is performed based on the calibrated location trajectory information and the conflict point location information; Based on the conflict-related object identifier in the active collision avoidance control decision result, the association binding is completed to obtain the spatiotemporal association matching result; The spatiotemporal correlation matching results and the active collision avoidance control decision results are sorted by conflict priority to obtain a priority sorting result. The instruction execution levels are divided according to the priority level in the priority sorting result; the control objectives of each level are decomposed into instruction elements, namely, execution timing requirements, execution parameter thresholds and execution feedback requirements; and the collision avoidance control sub-instructions corresponding to each level are generated by combining the decomposed instruction elements.

[0013] Preferably, the real-time control execution processing of the real-time traffic conflict point distribution map, the dynamic state data of traffic participants, and the active collision avoidance control decision results to obtain the active collision avoidance control execution results for traffic conflict points further includes: The collision avoidance control sub-commands are dynamically allocated by the execution unit to obtain the allocation result; The allocation results are subjected to real-time execution status monitoring to obtain feedback data of the collision avoidance control execution process; The execution deviation information in the feedback data of the collision avoidance control execution process is correlated with the priority weight in the priority ranking result; the collision avoidance control sub-instructions that have not been executed are adjusted according to the analysis results; the execution results of the adjusted sub-instructions are integrated with the execution results that have been completed; and the active collision avoidance control execution results for traffic conflict points are generated according to the integration results.

[0014] A traffic conflict point active collision avoidance control system, applicable to the aforementioned traffic conflict point active collision avoidance control method, includes: The conflict point identification unit is used to dynamically identify conflict points at road intersections and obtain a real-time traffic conflict point distribution map. The feature extraction unit is used to extract behavioral features of traffic participants based on the real-time traffic conflict point distribution map to obtain dynamic state data of traffic participants, wherein the dynamic state data of traffic participants includes velocity vector, acceleration trend and motion trajectory; The risk assessment unit is used to input the dynamic status data of the traffic participants into the traffic conflict risk assessment network for risk quantification processing to obtain the traffic conflict risk assessment result; and to perform conflict development trend prediction processing on the traffic conflict risk assessment result to obtain the traffic conflict evolution prediction result, wherein the traffic conflict evolution prediction result includes the risk spread speed and the scope of impact. The strategy hierarchical unit is used to prioritize the traffic conflict evolution prediction results for collision avoidance strategies to obtain a priority sequence; and to perform multi-objective coordinated control processing on the priority sequence to obtain active collision avoidance control decision results. The collision avoidance control unit is used to perform real-time control execution processing on the real-time traffic conflict point distribution map, the dynamic status data of traffic participants, and the active collision avoidance control decision results, so as to obtain the active collision avoidance control execution results for traffic conflict points.

[0015] Compared with the prior art, the beneficial effects of the present invention are: (1) This invention quantifies the dynamic state of participants through a traffic conflict risk assessment network and combines it with conflict development trend prediction to accurately identify and quantify traffic conflict risks, enabling the system to intervene before problems occur and minimize damage. Furthermore, through multi-objective coordinated control, it can optimize the collision avoidance decisions of various traffic participants, coordinate the behavior of different vehicles, reduce mutual interference, and improve overall traffic flow and safety. (2) By coordinating with infrastructure such as traffic signal control system and traffic guidance system, this invention can intelligently optimize traffic light duration, lane allocation, etc., improve the efficiency of traffic management system, and avoid traffic bottlenecks and congestion problems; and by real-time monitoring and feedback of the execution process, it can promptly adjust the unfinished collision avoidance control sub-instructions, ensure the execution accuracy of collision avoidance control decisions, and avoid safety hazards caused by execution errors. Attached Figure Description

[0016] Figure 1 This is a schematic flowchart of the overall method in one embodiment of the present invention; Figure 2This is a schematic diagram of the overall system architecture in one embodiment of the present invention.

[0017] In the diagram: 1. Conflict point identification unit; 2. Feature extraction unit; 3. Risk assessment unit; 4. Strategy classification unit; 5. Collision avoidance control unit. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1, please refer to Figure 1 This invention provides a technical solution: a method for active collision avoidance control at traffic conflict points, comprising: S1. Dynamically identify conflict points at road intersections to obtain a real-time traffic conflict point distribution map; S2. Based on the real-time traffic conflict point distribution map, the behavioral characteristics of traffic participants are extracted to obtain the dynamic state data of traffic participants, which includes velocity vector, acceleration trend and motion trajectory. S3. Input the dynamic status data of traffic participants into the traffic conflict risk assessment network for risk quantification to obtain the traffic conflict risk assessment results; perform conflict development trend prediction processing on the traffic conflict risk assessment results to obtain the traffic conflict evolution prediction results, including the risk diffusion speed and impact range. S4. Prioritize the traffic conflict evolution prediction results for collision avoidance strategies to obtain a priority sequence; perform multi-objective coordinated control processing on the priority sequence to obtain the active collision avoidance control decision results. S5. Perform real-time control execution processing on the real-time traffic conflict point distribution map, traffic participant dynamic status data, and active collision avoidance control decision results to obtain the active collision avoidance control execution results for traffic conflict points.

[0020] It should be noted that traffic flow information at intersections is collected in real time using sensors (such as cameras, radar, lidar, etc.) or other data sources. The dynamic state of the traffic flow is analyzed, and traffic conflict points are automatically identified. Conflict points refer to locations where different traffic participants (such as pedestrians, vehicles, cyclists, etc.) may collide at traffic intersections. Dynamic state data for each traffic participant is extracted from the real-time data, including: velocity vectors representing the direction and speed of movement in space; acceleration trends representing changes in the participant's speed, which can help determine if there are sudden braking or acceleration actions; and motion trajectories describing the participant's path over a certain period. The risk of traffic conflict is assessed through the analysis of the dynamic state data of traffic participants. This dynamic state data is then input into a dedicated traffic conflict risk assessment network, which quantifies the risk of traffic conflict and outputs a specific risk assessment result. Based on the current risk assessment result, the possible development trend of traffic conflict is predicted. The predictive model uses the behavioral characteristics of traffic participants and real-time data to predict the expansion or escalation of conflicts. Specifically, this includes: the speed at which conflict risk spreads at intersections or roads; predicting the area or size of the potential conflict impact; identifying the traffic participants most in need of action to reduce conflict risk based on the conflict evolution prediction results; assessing which traffic participants are most likely to collide based on the predicted conflict development trend and ranking them according to the urgency of conflict risk to generate a priority sequence; implementing coordinated control strategies based on the priority sequence to ensure all traffic participants can avoid collisions as much as possible; calculating the optimal control decision (e.g., adjusting vehicle speed, path, and traffic light signals) by comprehensively considering the collision avoidance needs of each participant through a multi-objective coordinated control algorithm; executing the calculated control decisions in real time to ensure effective handling of traffic conflict points; and conveying collision avoidance instructions to traffic participants in real time through traffic lights, road signs, onboard systems, or other control means, such as guiding traffic flow by adjusting traffic lights, audible warnings, or other methods to avoid collisions. Through this series of analyses and control processes, the system can identify conflict points and predict conflict trends in real-time traffic, and prevent conflicts from occurring through reasonable control measures, thereby improving traffic safety.

[0021] In an optional embodiment, dynamic conflict point identification is performed on road intersections to obtain a real-time traffic conflict point distribution map, including: Traffic data is collected and spatiotemporally calibrated to obtain multi-source sensing data; Feature extraction processing is performed on multi-source sensing data to obtain a basic sensing feature map, which includes vehicle detection boxes, speed estimates, and motion direction angles. The basic perception feature map is subjected to feature fusion processing to obtain the feature representation of the traffic scene; Traffic scene feature representation is modeled to obtain a traffic scene relationship graph, which includes the interaction relationships between vehicles. The potential conflict points are calculated and processed on the traffic scene relationship diagram to obtain a real-time traffic conflict point distribution map.

[0022] It should be noted that the process involves collecting real-time traffic data (such as traffic cameras, radar, and sensors) and performing spatiotemporal calibration on this data. The collected traffic data may come from different sensing sources and may be asynchronous in time and space. Therefore, spatiotemporal calibration techniques are needed to align these multi-source data to ensure that the data can be analyzed within the same spatiotemporal framework, resulting in a consistent, multi-source sensing dataset. Important traffic features are extracted from the spatiotemporally calibrated multi-source data. Using computer vision and signal processing methods, the following are extracted from the traffic data: vehicle detection boxes (the positions of vehicles in images or videos, usually labeled as bounding boxes); speed estimates (the estimated speeds of each vehicle based on traffic data); and motion direction angles (the directions of vehicle movement, usually expressed as angles relative to due north or the road direction). Sensing features from different sources and types are combined to form a unified traffic scene feature representation. Through feature fusion techniques, vehicle detection boxes, speed estimates, motion direction angles, and other information are fused into a more comprehensive traffic scene feature. This step typically uses deep learning, image processing, or other data fusion techniques to merge different features into a comprehensive feature map representing the entire traffic scene. Establish a relationship model among various participants in a traffic scenario to identify interactions between vehicles. Based on traffic scenario characteristics, model the relationships between traffic participants (such as vehicles, pedestrians, cyclists, etc.), especially the interactions between vehicles; for example, which vehicles might interact by approaching, changing lanes, driving side-by-side, or crossing. This step uses graph neural networks (GNNs) or other relationship modeling methods to generate a traffic scenario relationship graph by analyzing the dependencies between various elements in the traffic scenario. The traffic scenario relationship graph shows the interactions between various traffic participants (vehicles, pedestrians, etc.); for example, the interaction between two vehicles can be represented by the edges of the graph, possibly indicating the relative position, speed, acceleration, and intersection paths of the vehicles. Based on the traffic scenario relationship graph, identify and predict potential traffic conflict points. By analyzing the interactions between traffic participants and predicting their future trajectories, potential conflict points are calculated; for example, two vehicles may collide at an intersection or clash while changing lanes. The process of calculating potential conflict points typically uses techniques such as conflict detection algorithms, path prediction, and time series analysis. Finally, a real-time traffic conflict point distribution map is generated, showing the potential conflict points in the traffic scenario. Based on the traffic scenario relationship diagram and potential conflict points obtained in the previous steps, a conflict point distribution map is generated. This map shows the location and density of potential conflicts at intersections or on roads in real time, providing a basis for subsequent collision avoidance control and decision-making.

[0023] In an optional embodiment, behavioral features of traffic participants are extracted based on a real-time traffic conflict point distribution map to obtain dynamic state data of traffic participants, including: Based on the real-time traffic conflict point distribution map, target tracking processing is performed on traffic participants to obtain continuous movement trajectory data for each traffic participant; The motion trajectory data is smoothed and filtered to remove noise interference and fill in data gaps, resulting in an optimized trajectory sequence. The velocity vectors of traffic participants are calculated based on the optimized trajectory sequence, where the velocity vectors include the velocity magnitude and the direction angle of motion. The acceleration trend of traffic participants is calculated based on the velocity vector, including longitudinal and lateral acceleration components; The acceleration trend is analyzed to identify abnormal driving behavior patterns, including rapid acceleration, rapid deceleration, and sharp turns. By integrating and processing velocity vectors, acceleration trends, motion trajectories, and abnormal behavior patterns, dynamic state data of traffic participants can be obtained.

[0024] It should be noted that, based on the real-time traffic conflict point distribution map, the system tracks the movement trajectory of each traffic participant (e.g., vehicles, pedestrians, etc.); each traffic participant is tracked in real time using target detection and tracking algorithms (such as Kalman filtering, multi-target tracking, etc.); the system identifies each participant in the traffic scene and records their movement trajectory, obtaining continuous position data of each participant over time; noise interference in the movement trajectory data is cleaned up, and any missing data is filled in; for trajectory data obtained from target tracking, noise or data loss may occur due to environmental interference, sensor errors, or other reasons; smoothing filtering algorithms (such as Kalman filtering, mean filtering, etc.) are used to reduce the impact of noise, and interpolation methods (such as linear interpolation or spline interpolation) are used to fill in missing data, resulting in a smooth and complete trajectory sequence; Based on the optimized trajectory data, the speed information of each traffic participant is calculated; the instantaneous speed of each participant is obtained by calculating the distance between adjacent trajectory points and dividing by the time interval; the motion direction angle of the traffic participant is calculated by analyzing the direction change of the trajectory; this angle describes the motion direction of the traffic participant relative to a certain reference direction (usually the road or north); the acceleration trend of each traffic participant is calculated based on the velocity vector; acceleration is calculated through the rate of change of velocity, and is divided into: longitudinal acceleration, which describes the acceleration or deceleration of the vehicle along the direction of travel; it is usually represented by the change in the magnitude of velocity; lateral acceleration, which describes the acceleration or deceleration of the vehicle when turning or deviating from the original trajectory; it reflects the acceleration of the vehicle perpendicular to the direction of travel. By analyzing acceleration trends, abnormal driving behavior patterns of traffic participants are identified. Acceleration trends are used to detect several common abnormal driving behaviors: when longitudinal acceleration exceeds a certain threshold, it indicates sudden acceleration, possibly for overtaking or other reasons; when longitudinal acceleration is negative and changes significantly, it indicates sudden deceleration, possibly for emergency braking or obstacle avoidance; when lateral acceleration increases significantly, it indicates the vehicle is making a sharp turn, potentially indicating a risk of loss of control. Velocity vectors, acceleration trends, motion trajectories, and abnormal behavior patterns are integrated to generate dynamic state data of traffic participants. All this information (speed, acceleration, motion trajectory, and abnormal behavior) is combined to form a comprehensive dynamic state dataset of traffic participants. This dataset contains the real-time state of each traffic participant and can be used for further traffic safety analysis, behavior prediction, or real-time traffic monitoring systems.

[0025] In an optional embodiment, dynamic status data of traffic participants are input into a traffic conflict risk assessment network for risk quantification to obtain traffic conflict risk assessment results, including: The dynamic status data of traffic participants are input into the spatiotemporal correlation analysis unit for conflict pattern recognition processing to obtain a preliminary risk assessment vector. The preliminary risk assessment vector is subjected to multi-factor risk fusion processing to obtain a comprehensive risk assessment matrix; The comprehensive risk assessment matrix is ​​input into the cascaded risk level division unit for risk level determination, resulting in the traffic conflict risk assessment.

[0026] It should be noted that, based on the dynamic state data of traffic participants (such as speed, acceleration, and trajectory), the system analyzes their temporal and spatial relationships to identify potential conflict patterns. In this stage, the system inputs the dynamic state data of traffic participants into the spatiotemporal correlation analysis unit and, based on the participants' motion trajectories, speeds, accelerations, and other information, analyzes their relative positions, relative speeds, and potential future collisions or conflicts. By comparing the relative positions and motion states of different traffic participants within a specific time period, the system identifies possible conflict patterns. For example, if two vehicles approach an intersection at high speed, or if one vehicle rapidly accelerates towards another, the system may identify a potential traffic conflict. Through this process, the system generates a preliminary risk assessment vector, representing the initial risk level of each conflict pattern, typically including information such as the potential collision probability and the severity of the conflict.

[0027] The system comprehensively assesses conflict risk by combining multiple factors (such as the speed and acceleration of traffic participants, road conditions, and traffic signals). Preliminary risk assessment vectors may not fully and accurately reflect conflict risk because they may overlook other crucial factors. To further improve the accuracy of risk assessment, the system integrates these preliminary risk assessment vectors with other potential risk factors (such as weather, traffic conditions, road surface friction coefficient, and traffic light status). Multi-factor risk fusion uses a weighted or mathematical model to comprehensively consider various risk factors, resulting in a more comprehensive risk assessment. The fused assessment data forms a comprehensive risk assessment matrix that not only reflects the risk level of each conflict mode but also incorporates a comprehensive consideration of multiple risk factors.

[0028] Based on the information in the comprehensive risk assessment matrix, risks are categorized into different levels, and a final risk assessment result is provided. At this stage, the system inputs the comprehensive risk assessment matrix into a cascaded risk level classification unit. This unit assigns different risk assessment values ​​to different risk levels according to preset risk classification criteria, such as low risk, medium risk, and high risk. The cascaded nature of this stage refers to the system's potential application of multiple judgment criteria or models to progressively refine the risk level. For example, it might first provide a preliminary risk assessment based on collision probability, and then further refine and adjust based on factors such as traffic signals and weather conditions. Ultimately, the system outputs a traffic conflict risk assessment result, which is a risk level assessment of traffic conflicts, indicating the magnitude of traffic conflict risk at a specific time and location, thereby assisting traffic management personnel in making appropriate decisions and taking corresponding countermeasures.

[0029] In an optional embodiment, the traffic conflict risk assessment results are processed to predict the conflict development trend, resulting in a traffic conflict evolution prediction result, including: Acquire dynamic behavioral characteristic data of traffic participants and road network topology data. The dynamic behavioral characteristic data of traffic participants includes driving behavior pattern characteristics associated with vehicle driving trajectory, spatiotemporal distribution characteristics of traffic conflict points, and traffic flow state evolution characteristics. Based on the spatiotemporal distribution characteristics of traffic conflict points, high-incidence areas of traffic conflicts are identified, and criteria for identifying abnormal driving behavior are determined based on driving behavior pattern characteristics. The areas with high incidence of traffic conflicts are input into a pre-set spatiotemporal correlation feature extraction network to obtain a dynamic risk assessment map; The dynamic risk assessment map is optimized to obtain the optimized dynamic risk assessment map. Based on the optimized dynamic risk assessment map, timing optimization parameters for the intelligent traffic signal control system and path planning adjustment coefficients for the traffic guidance system are generated, and traffic conflict evolution prediction results are obtained.

[0030] It should be noted that driving behavior pattern characteristics refer to driving behaviors related to the vehicle's trajectory, such as acceleration, braking, speeding, and lane changing. These characteristics allow the system to understand driver habits and identify behaviors that may trigger conflicts, such as sudden braking, rapid acceleration, or frequent lane changes. The spatiotemporal distribution characteristics of traffic conflict points refer to the distribution of locations and times where traffic conflicts (such as accidents and minor collisions) occur. Analyzing this data can identify high-risk spatiotemporal areas, i.e., which locations and time periods are more prone to traffic conflicts. Traffic flow state evolution characteristics describe changes in traffic volume and vehicle speed; for example, a gradual increase in traffic volume on a road segment, or a decrease in vehicle speed at certain times. These characteristics help understand the dynamic changes in traffic flow and potential congestion and risks. Road network topology data refers to geometric information about the road network, including road shape, location, intersections, and number of lanes. This information is crucial for understanding traffic flow, analyzing conflict areas, and predicting potential risks. High-incidence areas of traffic conflicts are identified by analyzing the spatiotemporal distribution of past traffic conflicts (accidents, etc.) to pinpoint areas and time periods with higher frequency of conflicts. These high-incidence areas typically include intersections, sharp bends, and road bottlenecks. A spatiotemporal correlation feature extraction network (SCRI) is a machine learning or deep learning-based network model used to extract meaningful features from spatiotemporal data and identify potential traffic conflict patterns. After inputting the identified high-incidence areas into this network, it analyzes spatiotemporal features and historical data to output a dynamic risk assessment map, which represents the traffic risk assessment value for each area and time period. This map helps identify the most dangerous time periods and road segments in the traffic system. After obtaining the initial risk assessment map, it may need to be optimized. The optimization process may include adjusting the assessment model, removing noisy data, and focusing more on high-risk areas. The goal is to make the risk assessment more accurate, reduce errors, and improve the reliability of predictions. With the optimized map, the system can more accurately reflect the potential risks of traffic conflicts, helping traffic management personnel make better decisions. Based on the optimized dynamic risk assessment map, the system generates timing optimization parameters for the intelligent traffic signal control system. This process involves adjusting traffic light timings (such as red and green light cycles and green light durations) to reduce the risk of traffic conflicts or alleviate problems caused by excessive traffic flow. Simultaneously, the system generates route planning adjustment coefficients for the traffic guidance system. Traffic guidance systems typically guide drivers to choose the best routes and avoid high-risk areas or congested sections through road signs, navigation systems, etc. By adjusting the route planning coefficients, the system can provide safer route recommendations based on the risk assessment map. Finally, through the above process, the system outputs traffic conflict evolution prediction results. These results describe the possible trends of traffic conflicts, predict which road sections or time periods will have a higher risk of conflict in the future, and how to reduce the probability of conflict occurrence by optimizing traffic signal timings and route planning.

[0031] In an optional embodiment, the traffic conflict evolution prediction results are processed to prioritize collision avoidance strategies, resulting in a priority sequence, including: The traffic conflict evolution prediction results are converted into an initial policy representation space; High-risk conflict scenarios are identified using a conflict risk assessor, and the feasibility of strategies for these scenarios is verified. Establish a strategy evaluation model and combine it with dynamic weight adjustment to solve for the optimal strategy combination under multiple constraints; The optimal strategy combinations are sorted to obtain a priority sequence.

[0032] It should be noted that transforming the aforementioned risk assessment data into a strategy representation space means mapping the prediction results into a strategy selection framework. This representation space can include different control parameters and schemes, such as adjustments to traffic signal control, route recommendations, and speed limits. For example, for high-risk intersections, adjusting signal timing or setting temporary speed limits could be considered; for peak-hour traffic flow, strategies to guide traffic flow diversion might be adopted. The conflict risk assessor is a tool that analyzes traffic risks based on prediction data and algorithms (such as machine learning and data mining), and can identify high-risk areas and time periods for traffic conflicts. High-risk conflict scenarios are areas with a high risk of conflict or accidents, calculated by the assessor; for example, these might be intersections with high traffic volume, sharp bends, or congested areas. Once high-risk scenarios are identified, they can be analyzed in depth to determine which strategies to adopt to reduce conflict risk. For each high-risk conflict scenario, it is necessary to verify whether the proposed strategy is practically feasible and can effectively reduce the probability of conflict. The strategy evaluation model is used to assess the effectiveness of different strategies, typically based on factors such as traffic flow changes, conflict reduction effects, cost, and time. This model can comprehensively consider multiple influencing factors to evaluate the overall impact of each strategy on the traffic system. During the strategy evaluation process, the weights of different factors may change; for example, during peak hours, more emphasis may be placed on reducing conflict risk, while during low-traffic periods, more attention may be paid to improving efficiency. Therefore, it is necessary to dynamically adjust the weights of each factor according to the actual situation. By adjusting the weights, the model can better address the actual needs under different times, locations, and traffic conditions. When optimizing strategies, multiple constraints are typically encountered, such as: minimum and maximum traffic signal timing requirements; road capacity and traffic flow limits; budget and time constraints for implementing the strategy. Under these constraints, the system needs to use optimization algorithms (such as linear programming and heuristic algorithms) to find an optimal combination of strategies that minimizes overall traffic risk, maximizes efficiency, and meets all constraints. For example, a possible strategy combination might be: increasing green light duration during high-traffic periods, adjusting vehicle speeds on some road sections, and guiding certain vehicles to detour. All optimized strategy combinations are ranked according to factors such as effectiveness, implementation difficulty, and cost, resulting in a priority sequence. The ranking may be based on maximizing the benefits of the strategy combination, the ease of implementation, and minimizing the risks. The final ranking result will clarify which strategies should be implemented first. For example, signal timing should be adjusted first during peak traffic hours, while route guidance systems should be optimized first during periods of less congestion. Through this prioritization, traffic managers can start implementing strategies with higher priority to maximize the safety and efficiency of the traffic system.

[0033] In an optional embodiment, multi-objective coordinated control processing is performed on the priority sequence to obtain an active collision avoidance control decision result, including: Traffic participants are prioritized according to a priority sequence to identify key vehicles that require priority control. Generate vehicle control commands for key vehicles, where control parameters include desired acceleration, steering angle, and braking pressure; Calculate the vehicle's safe driving trajectory based on vehicle control commands to ensure driving stability while avoiding collisions; For non-critical vehicles, cooperative collision avoidance suggestions are generated, including auxiliary control commands such as deceleration prompts, lane keeping, and following distance adjustments. Based on the signal timing optimization scheme generated by the traffic signal control system, the duration and phase sequence of traffic lights are adjusted. It generates control commands for variable lanes, speed limit signs, and warning devices to achieve coordinated control of infrastructure and vehicles; By integrating vehicle control commands, safe driving trajectories, cooperative collision avoidance suggestions, and infrastructure control commands, active collision avoidance control decision results are obtained.

[0034] It should be noted that after prioritizing, the system determines which vehicles are "critical vehicles," meaning those that need to be controlled first in the current traffic scenario. These vehicles are typically located in critical positions where traffic conflicts may occur, have high or low speeds, or are close to other vehicles. For these critical vehicles, the system generates control commands to guide how they should react to avoid conflict. Control parameters include: determining the vehicle's acceleration or deceleration strategy; guiding the vehicle to avoid obstacles or conflict areas by adjusting the steering wheel angle; and controlling the vehicle's deceleration process by adjusting the braking force to avoid collisions with other vehicles. These control commands ensure that critical vehicles can safely and quickly take measures to avoid conflict. After generating control commands, the system also needs to calculate a safe trajectory to ensure the vehicle maintains stable driving while avoiding collisions, preventing loss of control or other safety accidents. For example, the system may require the vehicle to travel along a specific trajectory to avoid collisions with other vehicles ahead and maintain vehicle stability and safety. For non-critical vehicles, i.e., those with a low current risk of conflict, the system will generate cooperative collision avoidance suggestions to guide these vehicles on how to cooperate with other vehicles as much as possible for safe driving without interfering with the main collision avoidance strategy. These cooperative control suggestions include: prompting the driver to slow down appropriately to avoid exacerbating traffic congestion or sudden braking that could lead to a larger conflict; advising the driver to stay in the current lane to avoid new traffic conflicts caused by lane changes; and reminding the driver to maintain an appropriate following distance to avoid chain collisions caused by excessively close following distances. Traffic light timing (green, red, yellow, etc.) directly affects vehicle driving. By analyzing traffic flow and conflict risks in real time, the system adjusts the duration and phase sequence of traffic lights (i.e., the timing of green and red light switching). For example, it may be necessary to extend the green light time for a certain direction to reduce the possibility of conflicts. On some traffic-intensive roads, the number of lanes may be variable; the system may adjust the opening or closing of variable lanes based on real-time traffic conditions. Speed ​​limit signs are dynamically adjusted based on traffic flow and safety risks to guide drivers to slow down or accelerate. In potentially dangerous areas, warning devices (such as flashing red lights, traffic lights, etc.) are activated. Warning signs, etc., alert drivers to danger; through the above control commands, traffic infrastructure (such as traffic lights, lane indicators, speed limit signs, etc.) and vehicles achieve coordinated control; for example, when the speed limit of a lane is adjusted, the timing of traffic lights may also change to coordinate the smoothness of the entire traffic system; finally, the system integrates vehicle control commands, collision avoidance trajectories, coordinated collision avoidance suggestions, and infrastructure control commands to obtain an active collision avoidance control decision; this decision is a comprehensive control scheme that can ensure the efficient and orderly operation of the entire traffic system while ensuring the avoidance of traffic conflicts.

[0035] In an optional embodiment, real-time control execution processing is performed on the real-time traffic conflict point distribution map, traffic participant dynamic state data, and active collision avoidance control decision results to obtain the active collision avoidance control execution results for traffic conflict points, including: The location information of conflict points in the real-time traffic conflict point distribution map is synchronized with the location trajectory information in the dynamic status data of traffic participants on the time axis. Spatial range matching is performed based on the calibrated location trajectory information and the conflict point location information; Based on the conflict-related object identifiers in the active collision avoidance control decision results, the association binding is completed to obtain the spatiotemporal association matching results; The spatiotemporal correlation matching results and the active collision avoidance control decision results are sorted by conflict priority to obtain the priority sorting results; The instruction execution levels are divided according to the priority level in the priority sorting results; the control objectives of each level are decomposed into instruction elements, namely, execution timing requirements, execution parameter thresholds, and execution feedback requirements; and the collision avoidance control sub-instructions corresponding to each level are generated by combining the decomposed instruction elements.

[0036] It should be noted that the location trajectory information in the dynamic status data of traffic participants refers to the real-time location information of traffic participants collected by sensors or other technologies (such as GPS), showing their movement trajectories on the road. Since the conflict point location information and the location trajectory information are collected from different data sources, they may have different timestamps. In order to accurately determine whether a traffic participant is approaching a conflict point, the timelines of this information need to be aligned or synchronized. The purpose of this step is to ensure that the two datasets are consistent in time for subsequent analysis. The calibrated location trajectory and conflict point are then matched spatially to determine whether the traffic participant's current position is close to or has entered a potential conflict area. If the trajectory information and the conflict point location match, it indicates that the current traffic participant is facing a possible conflict risk. Each traffic participant and each conflict point has a unique identifier; these identifiers help the system understand which traffic participant might be associated with which conflict point; location trajectory information, conflict point information, and conflict-associated object identifiers are bound together; the bound data forms a "spatiotemporal correlation matching result," which tells the system which traffic participants might face a conflict at a specific time and location; the system prioritizes conflicts based on the spatiotemporal correlation matching result and the active collision avoidance control decision result; the prioritization is based on factors such as the severity of the conflict, proximity, and the status of the traffic participants (e.g., speed, distance); high-priority conflicts are handled first to ensure that the most dangerous situations are controlled in a timely manner; according to the priority ranking result, the system divides instructions into different execution levels; high-priority conflicts are assigned to higher execution levels, and low-priority conflicts are assigned to lower levels; each level represents a set of conflict situations or control objectives that need to be handled; each level will have different control objectives, which correspond to different collision avoidance decision tasks; for example, level 1 may be responsible for the most urgent collision avoidance task, and level 2 may be responsible for lower-risk situations; The control objectives at each level are broken down into instruction elements. Instruction elements refer to the various components of a control instruction, mainly including: execution timing requirements, which indicate when the collision avoidance operation needs to be performed, possibly including time windows or time interval requirements; execution parameter thresholds, which set the specific parameters for executing the instruction, such as the numerical range of acceleration, deceleration, and steering angle; and execution feedback requirements, which specify how the system provides feedback on the execution results, such as whether the vehicle successfully avoided the conflict area, or whether the expected deceleration or steering effect was achieved. The various elements of the instruction (timing, parameter thresholds, and feedback requirements) are combined to form a complete collision avoidance control instruction. These sub-instructions are specific execution instructions generated for different levels of control objectives. For example, for emergency collision avoidance situations, sub-instructions might include rapid braking and sharp turns; for low-risk situations, it might simply be adjusting vehicle speed or changing lanes.

[0037] In an optional embodiment, real-time control execution processing is performed on the real-time traffic conflict point distribution map, traffic participant dynamic state data, and active collision avoidance control decision results to obtain the active collision avoidance control execution results for traffic conflict points, and the process further includes: The collision avoidance control sub-commands are dynamically allocated to the execution unit to obtain the allocation result; The allocation results are monitored and processed in real time to obtain feedback data on the collision avoidance control execution process. The system performs correlation analysis between the execution deviation information in the feedback data of the collision avoidance control execution process and the priority weight in the priority ranking result; adjusts the collision avoidance control sub-instructions that have not been executed in the process based on the analysis results; integrates the execution results of the adjusted sub-instructions with the completed execution results; and generates the active collision avoidance control execution results for traffic conflict points based on the integration results.

[0038] It should be noted that an execution unit refers to a specific hardware or software unit that performs collision avoidance operations, such as an onboard control system, braking system, or steering system. Each collision avoidance control command is dynamically assigned to a different execution unit. The assignment process considers factors such as the availability, current status, and load capacity of each unit to ensure that the command is executed by the appropriate unit at the appropriate time. For example, a braking command might be assigned to the braking system, and a steering command to the steering system. After assignment, the system needs to monitor the status of each execution unit in real time to ensure that the execution of the command is not delayed or fails. Monitoring includes tracking the working status, progress, and real-time feedback of each execution unit. The monitoring results generate feedback data, which may include information such as the execution progress, success or failure, execution delay, and abnormal status of each execution unit. This feedback data reflects the execution status of the collision avoidance control commands in real time. The feedback data contains information indicating whether there are deviations during execution, such as excessively long command execution time, system response lag, or execution results not meeting expectations. Each collision avoidance control sub-command has a priority; higher-priority tasks need to be executed first and maintain high execution accuracy. Priority weights refer to the importance and urgency of the tasks, reflected in the priority ranking. Based on the execution deviations in the feedback data, combined with the task priority weights, the system analyzes which tasks have not been completed as expected or have significant deviations. Through this analysis, the system can understand whether some low-priority tasks are consuming too many resources, leading to insufficient execution of high-priority tasks, or whether some commands have poor execution results, requiring adjustments. For commands that fail to complete due to resource constraints or execution failures, the system will adjust based on the analysis results. For example, it may reallocate more resources, speed up command execution, or change the execution order to ensure that the most urgent collision avoidance tasks can be completed smoothly. The results of successfully executed collision avoidance control commands will be recorded, which may include whether the vehicle has successfully decelerated or successfully avoided the conflict zone. The adjusted and reallocated command execution results will be integrated to form a complete set of execution results. The purpose of integrating the completed and adjusted execution results is to obtain a complete and optimized collision avoidance control execution status, ensuring that all commands are processed correctly at the right time. The execution results of active collision avoidance control at traffic conflict points represent the execution effect of the traffic conflict point collision avoidance control strategy. Through real-time feedback, adjustment, and optimization from all execution units, the system finally outputs a comprehensive and optimized collision avoidance control execution result, ensuring that traffic participants can effectively avoid potential conflict points.

[0039] Example 2, please refer to Figure 2 This invention provides a technical solution: a traffic conflict point active collision avoidance control system, applicable to the aforementioned traffic conflict point active collision avoidance control method, comprising: The conflict point identification unit 1 is used to dynamically identify conflict points at road intersections and obtain a real-time traffic conflict point distribution map. Feature extraction unit 2 is used to extract behavioral features of traffic participants based on the real-time traffic conflict point distribution map to obtain dynamic state data of traffic participants, including velocity vector, acceleration trend and motion trajectory. Risk assessment unit 3 is used to input dynamic status data of traffic participants into the traffic conflict risk assessment network for risk quantification and to obtain traffic conflict risk assessment results; and to perform conflict development trend prediction processing on the traffic conflict risk assessment results to obtain traffic conflict evolution prediction results, including the risk spread speed and impact range. Strategy hierarchical unit 4 is used to prioritize the collision avoidance strategies based on the traffic conflict evolution prediction results to obtain a priority sequence; and to perform multi-objective coordinated control processing on the priority sequence to obtain the active collision avoidance control decision results. The collision avoidance control unit 5 is used to perform real-time control execution processing on the real-time traffic conflict point distribution map, the dynamic status data of traffic participants, and the active collision avoidance control decision results, so as to obtain the active collision avoidance control execution results of traffic conflict points.

[0040] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for active collision avoidance control at traffic conflict points, characterized in that, include: Dynamic conflict point identification is performed at road intersections to obtain a real-time traffic conflict point distribution map; Based on the real-time traffic conflict point distribution map, behavioral features of traffic participants are extracted to obtain dynamic state data of traffic participants, wherein the dynamic state data of traffic participants includes velocity vector, acceleration trend and motion trajectory; The dynamic status data of the traffic participants are input into the traffic conflict risk assessment network for risk quantification to obtain the traffic conflict risk assessment result; the traffic conflict risk assessment result is then processed to predict the conflict development trend to obtain the traffic conflict evolution prediction result, wherein the traffic conflict evolution prediction result includes the risk spread speed and the scope of impact. The traffic conflict evolution prediction results are subjected to collision avoidance strategy priority processing to obtain a priority sequence; the priority sequence is then subjected to multi-objective coordinated control processing to obtain active collision avoidance control decision results. The real-time traffic conflict point distribution map, the dynamic status data of traffic participants, and the active collision avoidance control decision results are processed in real time to obtain the active collision avoidance control execution results for traffic conflict points.

2. The active collision avoidance control method for traffic conflict points according to claim 1, characterized in that, Dynamic conflict point identification is performed at road intersections to obtain a real-time traffic conflict point distribution map, including: Traffic data is collected, and the traffic data is subjected to spatiotemporal calibration processing to obtain multi-source sensing data; The multi-source sensing data is subjected to feature extraction processing to obtain a basic sensing feature map, wherein the basic sensing feature map includes a vehicle detection box, a speed estimate, and a motion direction angle; The basic perception feature map is subjected to feature fusion processing to obtain a traffic scene feature representation; The traffic scene feature representation is subjected to traffic participant relationship modeling processing to obtain a traffic scene relationship graph, wherein the traffic scene relationship graph includes the interaction relationship between vehicles; The potential conflict points are calculated and processed on the traffic scene relationship diagram to obtain a real-time traffic conflict point distribution map.

3. The active collision avoidance control method for traffic conflict points according to claim 2, characterized in that, Based on the real-time traffic conflict point distribution map, behavioral features of traffic participants are extracted to obtain dynamic state data of traffic participants, including: Based on the real-time traffic conflict point distribution map, target tracking processing is performed on traffic participants to obtain continuous movement trajectory data for each traffic participant; The motion trajectory data is smoothed and filtered to remove noise interference and fill in data gaps, resulting in an optimized trajectory sequence. The velocity vectors of traffic participants are calculated based on the optimized trajectory sequence, wherein the velocity vectors include the velocity magnitude and the direction angle of motion; The acceleration trend of traffic participants, including longitudinal acceleration and lateral acceleration components, is calculated based on the velocity vector. The acceleration trend is subjected to feature analysis to identify abnormal driving behavior patterns, including rapid acceleration, rapid deceleration and sharp turns. The velocity vector, acceleration trend, motion trajectory, and abnormal behavior pattern are integrated and processed to obtain dynamic state data of traffic participants.

4. The active collision avoidance control method for traffic conflict points according to claim 3, characterized in that, The dynamic status data of the traffic participants are input into the traffic conflict risk assessment network for risk quantification to obtain the traffic conflict risk assessment results, including: The dynamic status data of the traffic participants are input into the spatiotemporal correlation analysis unit for conflict pattern recognition processing to obtain a preliminary risk assessment vector. The preliminary risk assessment vector is subjected to multi-factor risk fusion processing to obtain a comprehensive risk assessment matrix; The comprehensive risk assessment matrix is ​​input into the cascaded risk level classification unit for risk level determination, and the traffic conflict risk assessment result is obtained.

5. The active collision avoidance control method for traffic conflict points according to claim 4, characterized in that, The traffic conflict risk assessment results are processed to predict the conflict development trend, resulting in a traffic conflict evolution prediction result, including: Acquire dynamic behavioral characteristic data of traffic participants and road network topology data, wherein the dynamic behavioral characteristic data of traffic participants includes driving behavior pattern characteristics associated with vehicle driving trajectory, spatiotemporal distribution characteristics of traffic conflict points, and traffic flow state evolution characteristics. Based on the spatiotemporal distribution characteristics of the traffic conflict points, high-incidence areas of traffic conflicts are determined, and based on the driving behavior pattern characteristics, abnormal driving behavior identification criteria are determined. The high-incidence areas of traffic conflicts are input into a preset spatiotemporal correlation feature extraction network to obtain a dynamic risk assessment map; The dynamic risk assessment map is optimized to obtain an optimized dynamic risk assessment map; Based on the optimized dynamic risk assessment map, timing optimization parameters for the intelligent traffic signal control system and path planning adjustment coefficients for the traffic guidance system are generated, resulting in traffic conflict evolution prediction results.

6. The active collision avoidance control method for traffic conflict points according to claim 5, characterized in that, The traffic conflict evolution prediction results are processed by collision avoidance strategy prioritization to obtain a priority sequence, including: The traffic conflict evolution prediction results are converted into an initial strategy representation space; High-risk conflict scenarios are identified using a conflict risk assessor, and the feasibility of strategies for these high-risk conflict scenarios is verified. Establish a strategy evaluation model and combine it with dynamic weight adjustment to solve for the optimal strategy combination under multiple constraints; The optimal strategy combination is sorted to obtain a priority sequence.

7. The active collision avoidance control method for traffic conflict points according to claim 6, characterized in that, The priority sequence is subjected to multi-objective coordinated control processing to obtain active collision avoidance control decision results, including: The traffic participants are prioritized according to the priority sequence to determine the key vehicles that need to be controlled first. Vehicle control commands are generated for the key vehicle, wherein the control parameters include the desired acceleration, steering angle, and braking pressure; The vehicle's safe driving trajectory is calculated based on the vehicle control commands to ensure driving stability while avoiding collisions; For non-critical vehicles, cooperative collision avoidance suggestions are generated, including auxiliary control commands such as deceleration prompts, lane keeping, and following distance adjustments. Based on the traffic signal control system, a signal timing optimization scheme is generated to adjust the duration and phase sequence of traffic lights; It generates control commands for variable lanes, speed limit signs, and warning devices to achieve coordinated control of infrastructure and vehicles; The vehicle control commands, the safe driving trajectory, the cooperative collision avoidance suggestions, and the infrastructure control commands are fused together to obtain the active collision avoidance control decision result.

8. The active collision avoidance control method for traffic conflict points according to claim 7, characterized in that, The real-time traffic conflict point distribution map, the dynamic status data of traffic participants, and the active collision avoidance control decision results are processed in real-time to obtain the active collision avoidance control execution results for traffic conflict points, including: The conflict point location information in the real-time traffic conflict point distribution map is synchronized with the location trajectory information in the dynamic status data of traffic participants on the time axis. Spatial range matching is performed based on the calibrated location trajectory information and the conflict point location information; Based on the conflict-related object identifier in the active collision avoidance control decision result, the association binding is completed to obtain the spatiotemporal association matching result; The spatiotemporal correlation matching results and the active collision avoidance control decision results are sorted by conflict priority to obtain a priority sorting result. The instruction execution levels are divided according to the priority level in the priority sorting result; the control objectives of each level are decomposed into instruction elements, namely, execution timing requirements, execution parameter thresholds and execution feedback requirements; and the collision avoidance control sub-instructions corresponding to each level are generated by combining the decomposed instruction elements.

9. The active collision avoidance control method for traffic conflict points according to claim 8, characterized in that, The real-time control execution processing of the real-time traffic conflict point distribution map, the dynamic status data of traffic participants, and the active collision avoidance control decision results to obtain the active collision avoidance control execution results for traffic conflict points also includes: The collision avoidance control sub-commands are dynamically allocated by the execution unit to obtain the allocation result; The allocation results are subjected to real-time execution status monitoring to obtain feedback data of the collision avoidance control execution process; The execution deviation information in the feedback data of the collision avoidance control execution process is correlated with the priority weight in the priority ranking result; the collision avoidance control sub-instructions that have not been executed are adjusted according to the analysis results; the execution results of the adjusted sub-instructions are integrated with the execution results that have been completed; and the active collision avoidance control execution results for traffic conflict points are generated according to the integration results.

10. A traffic conflict point active collision avoidance control system, applicable to the traffic conflict point active collision avoidance control method described in any one of claims 1-9, characterized in that, include: The conflict point identification unit is used to dynamically identify conflict points at road intersections and obtain a real-time traffic conflict point distribution map. The feature extraction unit is used to extract behavioral features of traffic participants based on the real-time traffic conflict point distribution map to obtain dynamic state data of traffic participants, wherein the dynamic state data of traffic participants includes velocity vector, acceleration trend and motion trajectory; The risk assessment unit is used to input the dynamic status data of the traffic participants into the traffic conflict risk assessment network for risk quantification processing to obtain the traffic conflict risk assessment result; and to perform conflict development trend prediction processing on the traffic conflict risk assessment result to obtain the traffic conflict evolution prediction result, wherein the traffic conflict evolution prediction result includes the risk spread speed and the scope of impact. The strategy hierarchical unit is used to prioritize the traffic conflict evolution prediction results for collision avoidance strategies to obtain a priority sequence; and to perform multi-objective coordinated control processing on the priority sequence to obtain active collision avoidance control decision results. The collision avoidance control unit is used to perform real-time control execution processing on the real-time traffic conflict point distribution map, the dynamic status data of traffic participants, and the active collision avoidance control decision results, so as to obtain the active collision avoidance control execution results for traffic conflict points.

Images