Intelligent traffic intervention system and method based on multi-source data fusion and unmanned aerial vehicle cooperation

Abstract

The invention discloses an intelligent traffic intervention system and method based on multi-source data fusion and unmanned aerial vehicle cooperation, and belongs to the intelligent traffic and low-altitude economy fusion technology. According to the intervention method, firstly, through real-time collection and fusion processing of multi-source heterogeneous traffic data, data sources are integrated, a fusion feature vector constructed by combining space-time alignment and data cleaning operation can comprehensively and truly reflect the traffic running state, and high-quality and all-around data support is provided for follow-up traffic event diagnosis; secondly, a hybrid architecture diagnosis model is adopted, the reliability and accuracy of event diagnosis are considered, an accurate basis is provided for unmanned aerial vehicle scheduling, and waste of disposal resources is avoided; then, based on the event diagnosis result and the real-time state of the unmanned aerial vehicle resource pool, a scheduling optimization model is constructed, an optimal flight path is planned at the same time, and refined and dynamic scheduling of unmanned aerial vehicle resources is achieved; finally, a field intervention mode is innovated, and the traffic control flexibility and effectiveness are improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent transportation and low-altitude economy integration technology, and more specifically, it relates to an intelligent traffic intervention system and method based on multi-source data fusion and UAV collaboration. Background Technology

[0002] With the continuous growth of urban traffic flow, the efficiency of handling sudden congestion and traffic accidents has become a key measure of urban management. Current technological systems exhibit multiple breaks in the "perception-decision-execution" chain when dealing with such dynamic events.

[0003] At the event perception level, existing technologies mainly rely on fixed monitoring equipment and passive alarms. Fixed cameras have blind spots, floating car data is not representative enough at low penetration rates, and various data sources are independent of each other, lacking effective intelligent fusion and analysis methods. This results in insufficient diagnostic accuracy and significant response delays in determining event type, severity, and scope of impact.

[0004] In terms of incident response, traditional methods face structural limitations. Fixed traffic lights and signs cannot dynamically adjust their positions according to sudden road conditions, making them ineffective in providing early warnings behind accidents or congestion points, a major reason for frequent secondary accidents. Mobile temporary facilities require manual deployment, with response times of up to tens of minutes, failing to meet the needs of rapid response. Police resources are limited by coverage area and workload, making it difficult to achieve precise, 24 / 7 traffic management.

[0005] Although drones have begun to be used in traffic patrols, current applications are relatively rudimentary, primarily serving as "aerial cameras" for post-incident evidence collection. Their mission triggering, flight control, and intelligent coordination with traffic command systems are insufficient, resulting in limited functionality and failing to realize their potential as proactive intervention "intelligent mobile terminals." The structural defects in existing technologies—the disconnect between front-end perception and back-end decision-making, and between command decisions and dynamic execution methods—remain unresolved. There is an urgent need for innovative systems to bridge these gaps and achieve an automated and intelligent closed loop for traffic incident handling. Summary of the Invention

[0006] This invention addresses the technical problems existing in the prior art by providing an intelligent traffic intervention system and method based on multi-source data fusion and UAV collaboration.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: This invention first provides an intelligent traffic intervention method based on multi-source data fusion and UAV collaboration, comprising the following steps: S1. Real-time acquisition and fusion processing of multi-source heterogeneous traffic data; The system receives real-time traffic data from multiple heterogeneous data sources via an application programming interface (API). The data sources include at least one of the following: GPS trajectory and speed information reported by the vehicle navigation terminal, vehicle identity and location information collected by the roadside unit, and video stream data collected by the smart camera. The data is then spatiotemporally aligned and cleaned to construct a fused feature vector. S2. Intelligent diagnosis and classification of traffic incidents based on fused feature vectors; The fused feature vector obtained in S1 is input into the pre-trained event diagnosis model, which outputs a structured diagnosis report. The diagnosis model adopts a hybrid architecture that combines a rule-based expert system with a lightweight neural network. The diagnosis logic of the diagnosis model includes: event type determination and severity level quantification. S3. Drone mission planning and dynamic scheduling based on diagnostic results; Based on the event type and level in the diagnostic report, the task-payload mapping library is queried to determine the required payload combination. Combined with the real-time status information of the UAV resource pool, a scheduling optimization model is constructed with the goal of minimizing the global response time. After solving the model, the appropriate UAV is assigned to the event, and the optimal flight path is generated. S4. Adaptive intervention of drones as dynamic air traffic facilities in the field; After the dispatched drones arrive at the target airspace according to the dispatch instructions, they perform dual-cooperative adaptive intervention tasks. The dual collaborations are: generation and adaptive adjustment of dynamic traffic signals in the air, and micro-guidance information interaction for connected vehicles.

[0008] Preferably, the fused feature vector X ( t )for: X ( t ) = [ v _ avg ( t ), σ _ v ( t ),o ccupancy ( t ), ΔQ ( t ), I _ obj ( t )] In the formula, v _ avg ( t )for t Average vehicle speed on the target road segment at any given time; σ _ v ( t () represents the standard deviation of vehicle speed; occupancy ( t () represents lane occupancy rate; ΔQ ( t () represents the rate of change of queue length; I _ obj ( t ) represents the encoded value of the target detection result based on video.

[0009] Preferably, the diagnostic logic of the diagnostic model includes: event type determination and severity level quantification; Event type determination: If I _ obj ( t ='accident vehicle' and v _ avg ( t )exist Δt Internal drop exceeds threshold α If so, it is determined to be a traffic accident. T _ accident If only v _ avg ( t )< β and occupancy ( t If the level remains high, it is considered chronic congestion. T _ congestion ; Severity level quantification: Classified according to the impact index S; S = w 1*(1- v _ avg ( t ) / v _ free ) + w 2*( ΔQ ( t ) / L _ segment ) + w 3* N _ lanes _ blocked In the formula, v _ free For free flow velocity; L _ segment The length of the road segment; N _ lanes _ blocked Number of lanes affected; w 1. w 2. w 3 represents the weighting coefficient; according to S The event severity level is divided into L1 (minor), L2 (moderate), and L3 (serious) based on the preset range into which the value falls.

[0010] The preferred scheduling optimization model is:

[0011]

[0012] In the formula, U A collection of available drones; T _ comm _ i To issue commands to drones i Communication latency; T _ flight _ i For drones i Arrival at the location of the incident D . loc The estimated flight time; Payload _ i For drones i The set of loads; Battery _ i Its remaining battery power; E _ min The minimum power threshold required to perform the task.

[0013] Preferably, step S4 includes: S4.1: Generation and adaptive adjustment of dynamic air traffic signals; The dispatched UAV_k drone generates visual traffic signals in the air that match ground road conditions using an onboard foldable LED array or high-brightness projection module; the signal display parameters are dynamically adjusted according to the environment, and the display brightness is as follows:

[0014] In the formula, L _ base Basic brightness; Ambient _ Light This is the normalized value for ambient light; Traffic _ Density Traffic density; k 1. k 2 is the adjustment coefficient; S4.2: Microscopic guidance information interaction for connected vehicles; UAV_k establishes a direct link with specific connected vehicles entering its communication range via C-V2X or DSRC protocols. Based on the UAV's real-time observation of on-site traffic flow, it generates and pushes customized guidance information Msg for specific vehicles. Its data structure is: Msg={Vehicle_ID,Action,Details,Validity_Time}.

[0015] Preferred options also include: S5. Real-time evaluation of intervention effects and iterative optimization of strategies; During the drone's mission, it continuously transmits live video streams and vehicle trajectory data. The backend system quantifies the intervention effect in real time by comparing key indicators before and after the intervention, and stores the complete data tuple of the mission into the historical database to update the diagnostic model parameters in S2 and the scheduling weights in S3, forming a self-iterative and optimized technical closed loop.

[0016] Furthermore, the present invention also provides an intelligent traffic intervention system based on multi-source data fusion and UAV collaboration, which is used to implement the above-mentioned intervention method. The intervention system includes: The multi-source data fusion and intelligent diagnosis subsystem includes a data acquisition interface unit, a data fusion processing unit, and an event diagnosis unit; The UAV resource scheduling and mission management subsystem includes a UAV resource database, mission planning unit, and airspace management interface; The aerial dynamic intervention and interaction subsystem is deployed on the dispatched UAV and includes a mission execution control unit, an aerial signal generation unit, and a V2X communication unit. The effect evaluation and model optimization subsystem includes an effect monitoring unit and a model iteration unit.

[0017] Preferably, the data acquisition interface unit receives real-time data streams from the vehicle navigation system, roadside unit, and traffic monitoring camera; the data fusion processing unit maps multi-source heterogeneous data to a unified spatiotemporal coordinate system through a spatiotemporal alignment algorithm to construct a fused feature vector; and the event diagnosis unit performs event type determination and severity level quantification through a hybrid diagnosis model and outputs a structured diagnostic report.

[0018] Preferably, the UAV resource library dynamically maintains the status tuples of each UAV; the task planning unit queries the task-payload mapping relationship database based on the diagnostic report, runs the scheduling optimization model, and generates a scheduling scheme; the airspace management interface is used to submit flight paths and operational airspaces to the urban UAV air traffic management system for real-time airspace conflict verification and flight permit application. The mission execution control unit controls the UAV to fly to the target airspace and adaptively adjusts its hovering position based on the received scheduling scheme and real-time airborne sensor data. The airborne signal generation unit includes a physical folding mechanism, a high-brightness LED array, and a digital light processing projection module. According to the received instructions and environmental sensor input, it dynamically controls the traffic signal content and visual parameters displayed in the air. The V2X communication unit is used to establish links with connected vehicles entering the communication radius and encapsulates and sends customized guidance information in a structured format.

[0019] Preferably, the airborne signal generation unit can automatically deploy an LED array or projection screen after the drone takes off and automatically retract it before landing; When sending guidance information to connected vehicles, the V2X communication unit prioritizes processing information on vehicles in key impact locations, including: the first and last vehicles in a congested convoy, large freight vehicles, emergency vehicles, and vehicles where abnormal driving behavior is detected.

[0020] Compared with the prior art, the present invention has the following beneficial effects: The intelligent traffic intervention method based on multi-source data fusion and UAV collaboration provided in this invention breaks through the limitations of traditional traffic intervention methods, which rely on fixed ground facilities, static resources, and a single perception dimension. Through intelligent design across the entire process, it achieves precision, efficiency, and flexibility in traffic incident handling. First, the intervention method integrates data sources through real-time collection and fusion processing of multi-source heterogeneous traffic data. The fused feature vector constructed by combining spatiotemporal alignment and data cleaning operations comprehensively and accurately reflects the traffic operation status, providing high-quality, comprehensive data support for subsequent traffic incident diagnosis. Second, a hybrid architecture diagnostic model combining a rule-based expert system and a lightweight neural network is adopted, balancing the reliability and accuracy of incident diagnosis and providing precise basis for UAV scheduling, avoiding waste of resources. Then, based on the incident diagnosis results and the real-time status of the UAV resource pool, a scheduling optimization model is constructed, and the optimal flight path is planned, achieving refined and dynamic scheduling of UAV resources. Finally, an innovative on-site intervention mode is implemented, improving the flexibility and effectiveness of traffic control. This embodiment forms a complete closed loop of multi-source data acquisition and fusion, intelligent event diagnosis and classification, dynamic drone scheduling, and on-site adaptive intervention, realizing intelligent handling of traffic incidents from perception, diagnosis, scheduling to intervention, constructing a full-process intelligent closed loop, and promoting the upgrading of traffic control models. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 This is a schematic diagram of the overall architecture and data flow of the intervention system according to one embodiment of the present invention; Figure 2 This is a schematic diagram of a multi-source data fusion and event diagnosis process according to one embodiment of the present invention; Figure 3 This is a schematic diagram of the drone scheduling and mission planning logic according to one embodiment of the present invention; Figure 4 This is a schematic diagram of the on-site working status of a drone as a dynamic air traffic facility according to one embodiment of the present invention. Detailed Implementation

[0023] To make the technical problems, solutions, and beneficial effects addressed by this application clearer, the following detailed description, in conjunction with the accompanying drawings and embodiments, provides a more comprehensive explanation of an intelligent traffic intervention system and method based on multi-source data fusion and UAV collaboration. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0024] Example 1 This invention provides an intelligent traffic intervention method based on multi-source data fusion and UAV collaboration, comprising the following steps: S1. Real-time acquisition and fusion processing of multi-source heterogeneous traffic data; The system receives real-time traffic data from multiple heterogeneous data sources via an application programming interface (API). The data sources include at least one of the following: GPS trajectory and speed information reported by the vehicle navigation terminal, vehicle identity and location information collected by the roadside unit, and video stream data collected by the smart camera. The data is then spatiotemporally aligned and cleaned to construct a fused feature vector. S2. Intelligent diagnosis and classification of traffic incidents based on fused feature vectors; The fused feature vector obtained in S1 is input into the pre-trained event diagnosis model, which outputs a structured diagnosis report. The diagnosis model adopts a hybrid architecture that combines a rule-based expert system with a lightweight neural network. The diagnosis logic of the diagnosis model includes: event type determination and severity level quantification. S3. Drone mission planning and dynamic scheduling based on diagnostic results; Based on the event type and level in the diagnostic report, the task-payload mapping library is queried to determine the required payload combination. Combined with the real-time status information of the UAV resource pool, a scheduling optimization model is constructed with the goal of minimizing the global response time. After solving the model, the appropriate UAV is assigned to the event, and the optimal flight path is generated. S4. Adaptive intervention of drones as dynamic air traffic facilities in the field; After the dispatched drones arrive at the target airspace according to the dispatch instructions, they perform dual-cooperative adaptive intervention tasks. The dual collaborations are: generation and adaptive adjustment of dynamic traffic signals in the air, and micro-guidance information interaction for connected vehicles.

[0025] The intelligent traffic intervention method based on multi-source data fusion and UAV collaboration provided in this invention breaks through the limitations of traditional traffic intervention methods, which rely on fixed ground facilities, static resources, and a single perception dimension. Through intelligent design across the entire process, it achieves precision, efficiency, and flexibility in traffic incident handling. First, the intervention method integrates data sources through real-time collection and fusion processing of multi-source heterogeneous traffic data. The fused feature vector constructed by combining spatiotemporal alignment and data cleaning operations comprehensively and accurately reflects the traffic operation status, providing high-quality, comprehensive data support for subsequent traffic incident diagnosis. Second, a hybrid architecture diagnostic model is adopted, balancing the reliability and accuracy of incident diagnosis, providing precise basis for UAV scheduling and avoiding waste of resources. Then, based on the incident diagnosis results and the real-time status of the UAV resource pool, a scheduling optimization model is constructed, simultaneously planning the optimal flight path to achieve refined and dynamic scheduling of UAV resources. Finally, an innovative on-site intervention mode is implemented, improving the flexibility and effectiveness of traffic control. This embodiment forms a complete closed loop of multi-source data acquisition and fusion, intelligent event diagnosis and classification, dynamic drone scheduling, and on-site adaptive intervention, realizing intelligent handling of traffic incidents from perception, diagnosis, scheduling to intervention, constructing a full-process intelligent closed loop, and promoting the upgrading of traffic control models.

[0026] Specifically, the intelligent traffic intervention method based on multi-source data fusion and UAV collaboration provided in this embodiment includes the following steps: S1. Real-time acquisition and fusion processing of multi-source heterogeneous traffic data.

[0027] Real-time traffic data is received through an application programming interface (API) from at least three heterogeneous data sources, including at least one of the following: GPS trajectory and speed information reported by the vehicle navigation terminal, vehicle identity and location information collected by the roadside unit (RSU), and video stream data collected by the smart camera.

[0028] The above data is spatiotemporally aligned and cleaned to construct a fused feature vector. X ( t ).

[0029] Fusion feature vectors X ( t )for: X ( t ) = [ v _ avg ( t ), σ _ v ( t ),o ccupancy ( t ), ΔQ ( t ), I _ obj ( t )] In the formula, v _ avg ( t )for t Average vehicle speed on the target road segment at any given time; σ _ v ( t () represents the standard deviation of vehicle speed; occupancy ( t () represents lane occupancy rate; ΔQ ( t () represents the rate of change of queue length; I _ obj ( t ) is the encoded value of the video-based target detection results (such as accident vehicles, debris).

[0030] S2. Intelligent diagnosis and classification of traffic incidents based on fused feature vectors.

[0031] The fused feature vector obtained in S1 X ( t Inputting the data into a pre-trained event diagnosis model automatically outputs a structured diagnostic report D.

[0032] The diagnostic model adopts a hybrid architecture that combines rule-based expert systems with lightweight neural networks. The core diagnostic logic of the model includes: event type determination and severity level quantification.

[0033] Event type determination: For example, if I _ obj ( t ='accident vehicle' and v _ avg ( t )exist Δt Internal drop exceeds threshold α (like α If the percentage is 70%, then it is classified as a traffic accident. T _ accident If only v _ avg ( t )< β (like β =20 km / h) and occupancy ( t If the level remains high, it is considered chronic congestion. T _ congestion .

[0034] Severity level quantification: Classified according to the impact index S.

[0035] S = w1*(1- v _ avg ( t ) / v _ free ) + w 2*( ΔQ ( t ) / L _ segment ) + w 3* N _ lanes _ blocked In the formula, v _ free For free flow velocity; L _ segment The length of the road segment; N _ lanes _ blocked Number of lanes affected; w 1. w 2. w 3 represents the weighting coefficient.

[0036] according to S The event severity level is divided into L1 (minor), L2 (moderate), and L3 (serious) based on the preset range into which the value falls.

[0037] S3. Drone mission planning and dynamic scheduling based on diagnostic results.

[0038] Based on the event type and level in the diagnostic report D, query the task-load mapping library to determine the required task-load combination P (e.g., P={HD camera, LED projection}).

[0039] Based on the real-time status information of the drone resource pool (such as location, battery level, payload type, etc.), a scheduling optimization model is constructed with the goal of minimizing the global response time.

[0040] The scheduling optimization model is as follows:

[0041]

[0042] In the formula, U A collection of available drones; T _ comm _ i To issue commands to drones i Communication latency; T _ flight _ i For drones i Arrival at the location of the incident D . loc The estimated flight time; Payload _ i For drones i The set of loads; Battery _ i Its remaining battery power; E _ min The minimum power threshold required to perform the task.

[0043] Solve the model, assign an appropriate UAV_k to the event, and generate the optimal flight path.

[0044] S4. Adaptive intervention of drones as dynamic air traffic facilities in the field.

[0045] After the dispatched UAV_k arrives at the target airspace according to the scheduling instructions, it executes a dual-cooperative adaptive intervention mission, which specifically includes two cooperative sub-steps: S4.1: Generation and adaptive adjustment of dynamic air traffic signals; The dispatched UAV_k drone generates visual traffic signals in the air that match the ground road conditions through its onboard foldable LED array or high-brightness projection module.

[0046] Signal display parameters are dynamically adjusted according to the environment; display brightness:

[0047] In the formula, L _ base Basic brightness; Ambient _ Light This is the normalized value for ambient light; Traffic _ Density Traffic density; k 1. k 2 represents the adjustment factor. For example, at night or during periods of heavy traffic, the brightness and flashing frequency are automatically increased to enhance the warning effect.

[0048] S4.2: Microscopic guidance information interaction for connected vehicles; UAV_k establishes direct links with specific connected vehicles entering its communication range via C-V2X or DSRC protocols. Based on real-time observation of traffic flow by the UAV, it generates and pushes customized guidance information (Msg) for specific vehicles (such as the first vehicle in a queue or large vehicles). The data structure is: Msg={Vehicle_ID,Action,Details,Validity_Time}. For example, Msg={“Veh_12345”,“LANE_CHANGE”,“Immediately change lanes to the right, accident ahead”,“2023-10-27T10:05:00Z”}.

[0049] S5. Real-time evaluation of intervention effects and iterative optimization of strategies.

[0050] During the mission, the drone continuously transmits live video streams and vehicle trajectory data. The backend system compares changes in key indicators before and after intervention (such as the recovery rate of average traffic speed). The effect of this intervention, E, is quantified in real time (queue dissipation rate). The complete data tuple (X(t), D, UAV_k, E) of this task is stored in the historical database for periodic updates and optimizations of the diagnostic model parameters in S2 and the scheduling weights in S3, forming a self-iterative optimization technical closed loop.

[0051] Example 2 This invention provides an intelligent traffic intervention system based on multi-source data fusion and UAV collaboration, which is used to implement the above-mentioned intervention method. The intervention system includes: a multi-source data fusion and intelligent diagnosis subsystem, a UAV resource scheduling and task management subsystem, an aerial dynamic intervention and interaction subsystem, and an effect evaluation and model optimization subsystem.

[0052] Specifically, in this embodiment, the multi-source data fusion and intelligent diagnosis subsystem includes a data acquisition interface unit, a data fusion processing unit, and an event diagnosis unit.

[0053] The data acquisition interface unit receives real-time data streams from the vehicle navigation system, roadside unit (RSU), and traffic monitoring cameras through a standardized API gateway; The data fusion processing unit is equipped with a spatiotemporal alignment algorithm to map multi-source heterogeneous data to a unified spatiotemporal coordinate system and construct fused feature vectors. X ( t ); The event diagnosis unit has a built-in hybrid diagnosis model, which includes a rule-based inference engine and a lightweight neural network classifier. The hybrid diagnosis model performs event type determination and severity level quantification, and outputs a structured diagnosis report D.

[0054] In this embodiment, the UAV resource scheduling and mission management subsystem includes a UAV resource library, a mission planning unit, and an airspace management interface.

[0055] The drone resource library dynamically maintains the status tuples [ID, Position, Battery, Payload, Status] for each drone. The task planning unit can query the task-payload mapping relationship database based on the diagnostic report D, and run the scheduling optimization model to generate a scheduling scheme that includes the target UAV ID, flight path, operational airspace coordinates, and core task instructions. The airspace management interface is used to submit flight paths and operational airspace to the Urban Unmanned Air Traffic Management (UTM) system for real-time airspace conflict verification and flight permit application.

[0056] In this embodiment, the aerial dynamic intervention and interaction subsystem is deployed on the dispatched UAV and is an airborne system, including a mission execution control unit, an airborne signal generation unit, and a V2X communication unit.

[0057] The mission execution control unit, based on the received scheduling scheme and real-time airborne sensor data, controls the UAV to fly to the target airspace and adaptively adjusts the hovering position; The airborne signal generation unit integrates a physical folding mechanism, a high-brightness LED array, and a digital light processing projection module. Its controller adjusts the brightness according to received commands and environmental sensor inputs, based on a brightness adjustment formula. Dynamically control the content and visual parameters of traffic signals displayed in the air; The V2X communication unit supports the C-V2X Uu interface and PC5 direct communication protocol. It is used to establish links with connected vehicles entering the communication radius and encapsulate and send customized guidance information in the structured format Msg={Vehicle_ID, Action,Details,Validity_Time}.

[0058] In this embodiment, the effect evaluation and model optimization subsystem is a background data analysis system, which includes an effect monitoring unit and a model iteration unit.

[0059] The effect monitoring unit is used to receive on-site video and vehicle trajectory data transmitted back by the drone, calculate quantitative indicators such as traffic speed recovery rate, and generate an effect evaluation report. The model iteration unit stores the full-process data (X(t), D, UAV_k, E) of each intervention task into the case database, and periodically uses the accumulated data to perform offline training and optimization on the parameters of the hybrid diagnostic model in the event diagnosis unit and the scheduling weights in the task planning unit, so as to achieve continuous self-improvement of system performance.

[0060] Furthermore, as a preferred embodiment, the airborne signal generation unit can automatically deploy an LED array or projection screen after the drone takes off and automatically retract it before landing to reduce flight drag and storage space.

[0061] Furthermore, as a preferred embodiment, when the V2X communication unit sends guidance information to connected vehicles, it prioritizes processing information on vehicles at key influencing locations. These vehicles include: the first and last vehicles in a congested convoy, large freight vehicles, emergency vehicles, and vehicles where abnormal driving behavior (such as frequent lane changes or sudden braking) is detected.

[0062] Example 3 This embodiment provides an intelligent intervention process in a highway traffic accident scenario. Taking a typical highway two-vehicle rear-end collision as a background, this embodiment elaborates on the workflow of the system and method provided by this invention.

[0063] Specifically, the intervention process is as follows.

[0064] S1. System Initialization and Data Awareness See Figure 1 The intelligent traffic intervention system provided in this embodiment of the invention has been deployed in urban traffic command centers. Wherein: The multi-source data fusion and intelligent diagnosis subsystem, the UAV resource scheduling and mission management subsystem, and the effect evaluation and model optimization subsystem run on a cloud server cluster.

[0065] The aerial dynamic intervention and interaction subsystem is pre-installed in each duty drone.

[0066] The system maintains connectivity with roadside units (RSUs), high-definition cameras, and vehicle-mounted network platforms via a dedicated network.

[0067] At 10:00 AM on a certain day, a rear-end collision occurred between two vehicles at kilometer marker 10+300 on the G92 expressway (the leftmost lane in the eastbound direction). Almost simultaneously, the following multi-source data began to converge, such as... Figure 1 The following is an example of the "Roadside and Vehicle-Mounted Sensing Layer": Vehicle data: Approximately 20 vehicles equipped with navigation terminals behind the accident experienced a sudden drop in speed from 100 km / h to less than 20 km / h within 30 seconds, with some vehicles triggering emergency braking events and reporting them.

[0068] Roadside Unit (RSU) data: The RSU located at K10+200 detected that the average speed v_avg of passing vehicles decreased by more than 70% within the time window Δt=1min, and the vehicle density increased sharply.

[0069] Video data: The camera at K10+350 captured a vehicle parked at an abnormal angle through image recognition and detected the hazard warning lights.

[0070] S2, Multi-source data fusion and intelligent diagnosis The aforementioned heterogeneous data streams are transmitted in real time to the multi-source data fusion and intelligent diagnostic subsystem. (See also...) Figure 2 The system performs the following steps: S2.1 Data Preprocessing: The subsystem performs time alignment (unified to UTC timestamp) and spatial mapping (unified all coordinates to the highway mileage marker coordinate system) on the received GPS data, RSU packets, and video streams.

[0071] S2.2 Feature Vector Construction: The system extracts key features and constructs the fused feature vector X(t) at the current time t:

[0072] In the formula, I _ obj ( t The results are obtained in real time from the camera video stream analyzed by a lightweight convolutional neural network (CNN) model.

[0073] S2.3 Hybrid Model Diagnosis: X(t) is input into a pre-trained hybrid diagnostic model. The rule-based reasoning engine in the model is triggered first: since I_obj(t) = "accident vehicle" and v_avg(t) drops beyond the threshold α (α = 65%), the engine initially judges it as a "traffic accident". This initial judgment, along with X(t), is fed into a lightweight neural network classifier for accurate authentication and quantitative evaluation.

[0074] The classifier outputs the event type as T_accident (probability 98%) and calculates the severity index S = 0.8, corresponding to level L2 (moderately severe).

[0075] S2.4 Generating a Diagnostic Report: The system ultimately generates a structured diagnostic report. D = {Type: T _ accident ,grade: L 2 core position :[ K 10+300, Leftmost lane ], Scope of influence : K 10+200 To K 10+500, Estimated processing time 25 minute} This report, D, became the sole and precise basis for triggering all subsequent actions.

[0076] S3, Intelligent Scheduling and Mission Planning for Unmanned Aerial Vehicles Diagnostic report D was immediately sent to the UAV resource scheduling and mission management subsystem. (See also...) Figure 3 The scheduling decision-making process is as follows: S3.1 Task Requirement Analysis: Based on the event type T_acciden in D, the subsystem queries the built-in "Event-Task-Load" mapping library to determine the necessary load combination for handling this incident: P = { High-definition zoom gimbal camera ( Used for exploration ), Folded LED Array / Projection module ( Used for drainage )}.

[0077] S3.2 Resource Matching and Optimization Decision: The subsystem retrieves the currently available UAV resource pool, such as... Figure 3 As shown in the list on the left.

[0078] Assume the UAV-02 drone is in the following state: {Location: K9+800 patrolling, Battery: 78%, Payload: [camera, projection array], Status: interruptible}.

[0079] The scheduling optimization model is initiated and evaluated with the goal of minimizing the overall response time.

[0080] The model calculates the estimated time T_flight for UAV-02 to reach Loc_E(K10+300) to be 2.5 minutes. Considering the perfect load matching and sufficient power, UAV-02 is determined to be the optimal solution.

[0081] S3.3 Generating a scheduling scheme: The system generates detailed scheduling instructions and applies to the Low Altitude Traffic Management System (UTM) for flight permission for UAV-02 from the current airspace to the target airspace (centered on K10+300, with a radius of 100 meters and an altitude of 30-150 meters) through the airspace management interface.

[0082] After UTM approval, the dispatch instruction was issued to UAV-02.

[0083] The instructions include: flight waypoints, target hovering coordinates [H1] (directly above the accident site, used for reconnaissance), second hovering coordinates [H2] (150 meters behind the accident site, used for traffic control), and core mission sequence.

[0084] S4, UAV on-site adaptive intervention The UAV-02 autonomously flew to the scene, and its onboard airborne dynamic intervention and interaction subsystem began operation. (See also...) Figure 4 The following adaptive intervention occurred on-site: S4.1 On-site reconnaissance and situation update: UAV-02 first flew to coordinates [H1] and used a high-definition zoom camera to perform a 360° scan of the accident scene. It automatically identified that the accident involved two cars, with no one lying on the ground, but debris was blocking part of the middle lane. This image was transmitted back in real time, and the command center confirmed that no emergency medical rescue was needed and updated the response strategy.

[0085] S4.2, as a dynamic air traffic control facility: The UAV-02 then flies to coordinates [H2] (safe distance behind the accident site), and its bottom-mounted folding LED array automatically unfolds. At this time, it is daytime, and the ambient light is strong; the system adjusts its brightness according to the formula:

[0086] In the formula, Ambient_Light is estimated at 0.8 and Traffic_Density is 0.85, and the high brightness value is calculated.

[0087] The LED array then displayed a large red "prohibition" symbol and a green "left arrow," as shown. Figure 4 As shown by the drone, it clearly instructs vehicles behind to "Do not enter the rightmost lane, please change lanes to the left." At the same time, its projection module projects several green laser lines onto the ground, forming a temporary and conspicuous "virtual lane line" between the middle lane and the rightmost lane, guiding traffic flow through the bottleneck area in an orderly manner.

[0088] S4.3 Micro-interaction with connected vehicles: During UAV-02 operations, a connected SUV at the end of a congested queue enters its V2X communication range. The UAV-02's communication unit actively establishes a direct link (PC5 interface) with the vehicle. Based on the analysis of the overall traffic flow, the UAV-02 sends a customized guidance message to the SUV's in-vehicle infotainment system, such as... Figure 4 As shown in the bubble box: Msg={Vehicle_ID:“SUV_XYZ”, Action:“LANE_CHANGE_GUIDANCE”, Details:“Accident 1 km ahead. Please follow the green arrow on the drone at the front and change lanes to the left in advance. Estimated time to pass: 4 minutes.”, Validity_Time:“2023-10-27T10:08:00Z”}.

[0089] Drivers can operate based on these intuitive and personalized instructions, effectively preventing further deterioration of traffic flow caused by sudden braking and hesitation.

[0090] S5. Effectiveness Evaluation and Closed Loop During the mission, UAV-02 continuously transmitted data such as traffic flow speed and queue length. The effect evaluation and model optimization subsystem calculated that, after the UAV intervention was initiated, the average traffic speed recovery rate η for this road segment reached 70% within 15 minutes, which is better than the average level (50%) of similar historical incidents. The complete data chain of this mission (X(t), D, UAV-02, η=0.7) was stored in the case library for subsequent iterative optimization of the diagnostic model and scheduling parameters, thereby enabling the system's self-learning and continuous improvement.

[0091] This embodiment thus fully demonstrates the complete closed-loop process of the present invention, from automatic perception and diagnosis to intelligent scheduling and decision-making, then to dynamic aerial traffic management and personalized interaction using drones, and finally to the optimization of results through feedback. This system significantly improves the timeliness, accuracy, and intelligence of handling traffic emergencies, effectively solving the fundamental problem of the inability to dynamically relocate fixed facilities.

[0092] This invention provides an intelligent traffic intervention system and method that utilizes multi-source traffic data for intelligent event diagnosis and automatically dispatches drones as dynamic execution terminals for on-site intervention. It breaks through the limitations of traditional traffic intervention methods, which rely on fixed ground facilities, static resources, and a single perception dimension. Through a fully intelligent design, it achieves precise, efficient, and flexible traffic event handling. First, the intervention method integrates data sources through real-time acquisition and fusion processing of multi-source heterogeneous traffic data. The fused feature vector constructed by combining spatiotemporal alignment and data cleaning operations comprehensively and accurately reflects the traffic operation status, providing high-quality and comprehensive data support for subsequent traffic event diagnosis. Second, it employs a hybrid architecture diagnostic model combining a rule-based expert system and a lightweight neural network, balancing the reliability and accuracy of event diagnosis and providing precise basis for drone dispatch, avoiding waste of resources. Then, based on the event diagnosis results and the real-time status of the drone resource pool, a dispatch optimization model is constructed, simultaneously planning the optimal flight path to achieve refined and dynamic dispatch of drone resources. Finally, it innovates the on-site intervention mode, improving the flexibility and effectiveness of traffic control. This embodiment forms a complete closed loop of multi-source data acquisition and fusion, intelligent event diagnosis and classification, dynamic drone scheduling, and on-site adaptive intervention, realizing intelligent handling of traffic incidents from perception, diagnosis, scheduling to intervention, constructing a full-process intelligent closed loop, and promoting the upgrading of traffic control models.

[0093] In the description of this invention, it should be understood that terms such as “length”, “width”, “upper”, “lower”, “front”, “rear”, “left”, “right”, “vertical”, “horizontal”, “top”, “bottom”, “inner”, and “outer” indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0094] Furthermore, in the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0095] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A smart traffic intervention method based on multi-source data fusion and UAV collaboration, characterized in that, Includes the following steps: S1. Real-time acquisition and fusion processing of multi-source heterogeneous traffic data; Real-time traffic data is received from multiple heterogeneous data sources via an application programming interface (API). These data sources include at least one of the following: GPS trajectory and speed information reported by an in-vehicle navigation terminal; vehicle identity and location information collected by a roadside unit; and video stream data collected by a smart camera. The data is then spatiotemporally aligned and cleaned to construct a fused feature vector. S2. Intelligent diagnosis and classification of traffic incidents based on fused feature vectors; The fused feature vector obtained in S1 is input into the pre-trained event diagnosis model, which outputs a structured diagnosis report. The diagnosis model adopts a hybrid architecture that combines a rule-based expert system with a lightweight neural network. The diagnosis logic of the diagnosis model includes: event type determination and severity level quantification. S3. Drone mission planning and dynamic scheduling based on diagnostic results; Based on the event type and level in the diagnostic report, the task-payload mapping library is queried to determine the required payload combination. Combined with the real-time status information of the UAV resource pool, a scheduling optimization model is constructed with the goal of minimizing the global response time. After solving the model, the appropriate UAV is assigned to the event, and the optimal flight path is generated. S4. Adaptive intervention of drones as dynamic air traffic facilities in the field; After the dispatched drones arrive at the target airspace according to the dispatch instructions, they perform dual-cooperative adaptive intervention tasks. The dual collaborations are: generation and adaptive adjustment of dynamic traffic signals in the air, and micro-guidance information interaction for connected vehicles.

2. The intelligent traffic intervention method based on multi-source data fusion and UAV collaboration according to claim 1, characterized in that, The fused feature vector X ( t )for: X ( t ) = [ v _ avg ( t ), σ _ v ( t ),o ccupancy ( t ), ΔQ ( t ), I _ obj ( t )] In the formula, v _ avg ( t )for t Average vehicle speed on the target road segment at any given time; σ _ v ( t () represents the standard deviation of vehicle speed; occupancy ( t () represents lane occupancy rate; ΔQ ( t () represents the rate of change of queue length; I _ obj ( t ) represents the encoded value of the target detection result based on video.

3. The intelligent traffic intervention method based on multi-source data fusion and UAV collaboration according to claim 1, characterized in that, The diagnostic model's diagnostic logic includes: event type determination and severity level quantification; Event type determination: If I _ obj ( t ='accident vehicle' and v _ avg ( t )exist Δt Internal drop exceeds threshold α If so, it is determined to be a traffic accident. T _ accident If only v _ avg ( t )< β and occupancy ( t If the level remains high, it is considered chronic congestion. T _ congestion ; Severity level quantification: Classified according to the impact index S; S = w 1*(1- v _ avg ( t ) / v _ free ) + w 2*( ΔQ ( t ) / L _ segment ) + w 3* N _ lanes _ blocked In the formula, v _ free For free flow velocity; L _ segment The length of the road segment; N _ lanes _ blocked Number of lanes affected; w 1. w 2. w 3 represents the weighting coefficient; according to S The event severity level is divided into L1 (minor), L2 (moderate), and L3 (serious) based on the preset range into which the value falls.

4. The intelligent traffic intervention method based on multi-source data fusion and UAV collaboration according to claim 1, characterized in that, The scheduling optimization model is as follows: In the formula, U A collection of available drones; T _ comm _ i To issue commands to drones i Communication latency; T _ flight _ i For drones i Arrival at the location of the incident D . loc The estimated flight time; Payload _ i For drones i The set of loads; Battery _ i Its remaining battery power; E _ min The minimum power threshold required to perform the task.

5. The intelligent traffic intervention method based on multi-source data fusion and UAV collaboration according to claim 1, characterized in that, Step S4 includes: S4.1: Generation and adaptive adjustment of dynamic air traffic signals; The dispatched UAV_k drone generates visual traffic signals in the air that match ground road conditions using an onboard foldable LED array or high-brightness projection module; the signal display parameters are dynamically adjusted according to the environment, and the display brightness is as follows: In the formula, L _ base Basic brightness; Ambient _ Light This is the normalized value for ambient light; Traffic _ Density Traffic density; k 1. k 2 is the adjustment coefficient; S4.2: Microscopic guidance information interaction for connected vehicles; UAV_k establishes a direct link with specific connected vehicles entering its communication range via C-V2X or DSRC protocols. Based on the UAV's real-time observation of on-site traffic flow, it generates and pushes customized guidance information Msg for specific vehicles. Its data structure is: Msg={Vehicle_ID,Action,Details,Validity_Time}.

6. A smart traffic intervention method based on multi-source data fusion and UAV collaboration according to any one of claims 1-5, characterized in that, Also includes: S5. Real-time evaluation of intervention effects and iterative optimization of strategies; During the mission, the drone continuously transmits live video streams and vehicle trajectory data. The back-end system quantifies the intervention effect in real time by comparing key indicators before and after the intervention, and stores the complete data tuple of the task into the historical database to update the diagnostic model parameters in S2 and the scheduling weights in S3, forming a self-iterative optimization technical closed loop.

7. An intelligent traffic intervention system based on multi-source data fusion and UAV collaboration, characterized in that, It is used to implement the intervention method according to any one of claims 1-6, the intervention system comprising: The multi-source data fusion and intelligent diagnosis subsystem includes a data acquisition interface unit, a data fusion processing unit, and an event diagnosis unit; The UAV resource scheduling and mission management subsystem includes a UAV resource database, mission planning unit, and airspace management interface; An aerial dynamic intervention and interaction subsystem, deployed on the dispatched UAV, includes a mission execution control unit, an aerial signal generation unit, and a V2X communication unit; The effect evaluation and model optimization subsystem includes an effect monitoring unit and a model iteration unit.

8. The intelligent traffic intervention system based on multi-source data fusion and UAV collaboration according to claim 7, characterized in that, The data acquisition interface unit receives real-time data streams from the vehicle navigation system, roadside unit, and traffic monitoring camera; the data fusion processing unit uses a spatiotemporal alignment algorithm to map multi-source heterogeneous data to a unified spatiotemporal coordinate system and construct a fused feature vector; the event diagnosis unit uses a hybrid diagnosis model to perform event type determination and severity level quantification and output a structured diagnosis report.

9. The intelligent traffic intervention system based on multi-source data fusion and UAV collaboration according to claim 7, characterized in that, The UAV resource database dynamically maintains the status tuples of each UAV; the task planning unit queries the task-payload mapping database based on the diagnostic report and runs the scheduling optimization model to generate a scheduling scheme; the airspace management interface is used to submit flight paths and operational airspaces to the urban UAV air traffic management system for real-time airspace conflict verification and flight permit application. The task execution control unit controls the UAV to fly to the target airspace and adaptively adjusts the hovering position based on the received scheduling scheme and real-time airborne sensor data; the airborne signal generation unit includes a physical folding mechanism, a high-brightness LED array and a digital light processing projection module, and dynamically controls the traffic signal content and visual parameters displayed in the air according to the received instructions and environmental sensor input. The V2X communication unit is used to establish a link with connected vehicles entering the communication radius and to encapsulate and send customized guidance information in a structured format.

10. The intelligent traffic intervention system based on multi-source data fusion and UAV collaboration according to claim 7, characterized in that, The airborne signal generation unit can automatically deploy an LED array or projection screen after the drone takes off and automatically retract it before landing. When the V2X communication unit sends guidance information to connected vehicles, it prioritizes processing information on vehicles at key impact locations. These vehicles include the first and last vehicles in a congested convoy, large freight vehicles, emergency vehicles, and vehicles where abnormal driving behavior is detected.

Images