An intelligent vehicle flow trajectory optimization method and system based on AI visual analysis

Abstract

The application discloses a kind of wisdom car flow trajectory optimization method and system based on AI vision analysis, it is related to intelligent traffic control technical field, the present application is by fusing city-level AI vision perception network and emergency vehicle real-time state data, constructs the dynamic decision dataset of whole element, high precision, provides reliable basis for traffic scheme formulation;The present application is by asymmetric phase compensation, safety redundancy constraint and fast phase recovery mechanism, the influence of signal adjustment is accurately controlled in necessary space-time range.Combining the hierarchical induction strategy from far to near, effectively guide the peripheral traffic diversion, avoid the secondary congestion caused by emergency traffic in a large range, long time, maintain the overall operation efficiency of road network.The present application constructs complete perception-decision-execution-feedback closed-loop control system, ensure the continuous and reliable of emergency corridor under changing environment, guarantee the uninterrupted execution of emergency task.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic control technology, specifically to a method and system for optimizing intelligent traffic flow trajectories based on AI visual analysis. Background Technology

[0002] With the acceleration of urbanization and the continuous growth of motor vehicle ownership, urban traffic congestion is becoming increasingly serious, posing a severe challenge to the rapid passage of emergency vehicles such as fire trucks, ambulances, and police vehicles. Ensuring priority passage for emergency vehicles in complex traffic environments is crucial to improving urban emergency rescue efficiency and ensuring public safety.

[0003] Existing technologies, such as invention patent applications related to intelligent transportation with publication numbers CN115691110B, CN116524715A, and CN118135824A, reveal that technical solutions for prioritizing emergency vehicle passage mainly fall into two categories: The first category is passive signal priority control based on pre-set fixed routes. This type of solution typically involves the command center manually or semi-automatically planning a fixed route for the emergency vehicle before departure or during its journey, and remotely intervening in traffic lights along the route, such as implementing a continuous green light or inserting a dedicated phase. However, this method has significant limitations: First, route planning relies on historical experience or static road network information, making it unable to respond in real-time to dynamic traffic conditions such as road construction, accidents, and chronic congestion, potentially causing the pre-set route to fail; second, the signal control strategy is relatively crude, often employing a full-line green light or prolonged phase occupation, which, while ensuring the passage of emergency vehicles, easily causes widespread and prolonged congestion of other vehicles along the route, creating a negative effect of opening up one area only to block another, resulting in excessively high overall traffic costs.

[0004] The second type of solution usually relies solely on the location data of the emergency vehicle itself, lacking a refined perception of the surrounding traffic flow status such as lane occupancy and queue length; its guidance information is usually a uniform broadcast prompt such as please give way, failing to generate differentiated driving behavior instructions based on the relative position of social vehicles and emergency vehicles and lane information.

[0005] Finally, such solutions often treat path planning, signal control, and vehicle guidance as independent components, lacking a unified decision-making model capable of coordinating spatiotemporal resource optimization. Information asynchrony and control disharmony between systems make it difficult to form efficient and smooth mobile green wave traffic corridors. In summary, existing technologies struggle to minimize overall disruption to urban transportation systems while ensuring emergency vehicles can travel at high speeds with zero delays. The fundamental reason lies in the lack of a closed-loop intelligent control system that integrates high-precision holographic perception, dynamic path planning, spatiotemporal optimization, and real-time feedback adjustments. Summary of the Invention

[0006] To address the aforementioned technical shortcomings, the present invention aims to provide a method and system for optimizing intelligent traffic flow trajectories based on AI visual analysis.

[0007] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: The first aspect of the present invention provides a smart traffic flow trajectory optimization method based on AI visual analysis, including: S1, acquiring real-time location, speed, destination data features of emergency vehicles and traffic flow status data features on the predetermined path ahead of the emergency vehicles in real time through a city-level traffic AI visual perception network and an emergency vehicle positioning and status vehicle unit, and constructing a dynamic decision dataset for emergency passage corridors.

[0008] S2. Perform multi-dimensional path planning and spatiotemporal impact range analysis on the dynamic decision dataset of emergency passage corridors to generate corridor contingency plan processing results that include estimated passage time, a list of affected intersections, and key bottleneck points.

[0009] S3. Input the corridor contingency plan processing results into the pre-constructed spatiotemporal resource dynamic allocation optimization model, and output the phase switching timing scheme for a series of traffic lights along the route, as well as the graded guidance suggestions generated for social vehicles in the corridor.

[0010] S4. Based on the judgment rules of real-time traffic flow evolution, the hierarchical guidance suggestion results are dynamically adjusted and judged. When it is determined that the corridor is blocked, the subsequent backup route planning and guidance scheme update processing are triggered.

[0011] S5. The subsequent update processing results are simultaneously sent out, real-time control commands are sent to the signal control system to form a mobile green wave, and avoidance prompts are pushed to roadside units or mobile Internet terminals to generate priority passage corridors for emergency vehicles.

[0012] A second aspect of the present invention provides a system for executing the AI-based intelligent traffic flow trajectory optimization method, comprising: a holographic perception layer consisting of AI cameras, radar, and roadside computing units covering urban main roads, used to construct a high-precision traffic perception base.

[0013] Edge computing layer: Deployed near the intersection, it is responsible for real-time processing of visual feature streams, target recognition and local traffic status analysis.

[0014] Cloud Control Center: Includes a path planning engine, a spatiotemporal resource optimization model library, and a multi-source data integration bus, responsible for the generation and adjustment of global corridors.

[0015] Execution layer: includes traffic signal controller control interface, roadside V2X communication unit and mobile travel service backend, used for instruction issuance and feedback closed loop.

[0016] The beneficial effects of the present invention are as follows: (1) In view of the problem that emergency vehicles are easily affected by dynamic road conditions in the prior art, the present invention constructs a dynamic decision dataset by integrating holographic perception data, and generates a moving green wave and graded guidance instructions that match the vehicle speed in real time based on the spatiotemporal resource optimization model, which can ensure that emergency vehicles can pass continuously in the dynamic traffic environment in the shortest time approximating the theoretical minimum.

[0017] (2) In view of the problem that emergency priority in the prior art can easily cause large-scale social traffic congestion, the present invention introduces the goal of minimizing the total delay of social vehicles in signal control through a two-layer optimization model, and combines a graded guidance strategy from far to near to accurately control the impact of signal adjustment within the necessary time and space range, effectively avoiding secondary congestion.

[0018] (3) The perception-decision-execution-feedback closed-loop control system constructed by the present invention has the ability to dynamically replan based on real-time traffic flow status. When the preset corridor is blocked, it can quickly switch to the backup path, ensuring the continuous and reliable execution of emergency passage tasks. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the implementation steps of the method of the present invention.

[0021] Figure 2 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0022] 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.

[0023] Reference Figure 1As shown, the first aspect of the present invention provides a smart traffic flow trajectory optimization method based on AI visual analysis, including: S1, acquiring real-time location, speed, destination data features of emergency vehicles and traffic flow status data features on the predetermined path ahead of the emergency vehicles in real time through a city-level traffic AI visual perception network and an emergency vehicle positioning and status vehicle unit, and constructing a dynamic decision dataset for emergency passage corridors.

[0024] In a specific embodiment of the present invention, the step of constructing a dynamic decision dataset for emergency passage corridors further includes: using panoramic cameras and edge computing nodes in a city-level traffic AI visual perception network to perform semantic segmentation and target tracking on the area covered by the emergency vehicle travel path, and extracting traffic flow status data features including lane occupancy rate, average queue length, social vehicle density distribution, and non-motorized vehicle interference coefficient.

[0025] Specifically, the system first deploys and activates a city-level traffic AI visual perception network, utilizing panoramic cameras distributed along the predetermined routes of emergency vehicles to continuously collect video footage of road traffic conditions. This video data stream is transmitted in real-time to roadside edge computing nodes for processing. Through semantic segmentation algorithms, the images are analyzed into different elements such as lanes, vehicles, pedestrians, and non-motorized vehicles. Then, target tracking algorithms are used to continuously track each social vehicle and non-motorized vehicle within the field of view. Based on this, the system can quantify and extract a series of refined traffic flow state data features. For example, it calculates lane occupancy by statistically analyzing the proportion of time vehicles occupy the detection loop area per unit time; it obtains the average queue length by identifying and counting the number of vehicles waiting behind the stop line at intersections; it obtains the social vehicle density distribution by analyzing the distribution of vehicles within lanes; and it generates a non-motorized vehicle interference coefficient by assessing the impact of non-motorized vehicle crossings on motorized vehicle flow.

[0026] Specifically, the non-motorized vehicle interference coefficient is obtained by tracking algorithms to acquire non-motorized vehicle travel trajectories, real-time speed and spacing data of motorized vehicle flow, quantitatively analyze the impact of non-motorized vehicle travel on motorized vehicle speed reduction, traffic conflicts, etc., and calculate the non-motorized vehicle interference coefficient according to a preset weighted model. The non-motorized vehicle interference coefficient takes a value of 0-1, and the larger the value, the more significant the interference.

[0027] The real-time location, speed, destination, vehicle type identification, mission urgency, instantaneous acceleration, heading angle, and estimated arrival time of emergency vehicles are obtained through the vehicle positioning and status onboard unit. Combined with high-precision map information, the dynamic coordinates of emergency vehicles are mapped to the lane-level road network topology.

[0028] Specifically, the positioning and status onboard unit mounted on the emergency vehicle continuously reports its dynamic information to the system, including vehicle type identification, mission urgency level set by the command center, and operational parameters such as instantaneous acceleration and heading angle acquired by sensors, as well as the estimated arrival time calculated based on the destination. After obtaining the geographical coordinates of the emergency vehicle, the system immediately combines them with pre-set high-precision map information to accurately match them to the specific lane, completing the mapping of dynamic coordinates in the lane-level road network topology.

[0029] A multi-source data fusion algorithm is used to synchronize and register visual perception data with vehicle-mounted unit data in time and space, eliminating perception blind spots and forming a dynamic decision dataset covering all elements of vehicle-road-environment.

[0030] Specifically, to ensure data integrity and consistency, the system employs a multi-source data fusion algorithm to process visual perception data from the city-level traffic AI visual perception network and vehicle-mounted data from the emergency vehicle positioning and status onboard unit. Time synchronization is achieved through a unified timestamp, and spatial registration is performed using a high-precision map as a unified spatial reference. This effectively solves the problems caused by differences in data acquisition frequency and coordinate systems between different devices, and fills in perception blind spots caused by occlusion or limited viewing angles of individual cameras. Ultimately, this forms a comprehensive, accurate, and all-encompassing dynamic decision-making dataset for emergency passage corridors.

[0031] This method constructs a multi-dimensional, high-precision dynamic decision-making dataset by deeply fusing holographic visual perception data with emergency vehicle status data. It not only overcomes the information limitations and delays inherent in traditional methods relying on single data sources, but also enables the system to perceive micro-level traffic conflicts and potential bottlenecks through refined traffic flow characterization and precise lane-level mapping of vehicle positions. This vehicle-road-environment data fusion, eliminating perception blind spots, provides a solid and reliable data foundation for subsequent route planning, signal optimization, and guidance strategy development. It enhances the predictability, scientific rigor, and timeliness of emergency traffic corridor generation schemes, thereby ensuring the efficient execution of emergency tasks.

[0032] S2. Perform multi-dimensional path planning and spatiotemporal impact range analysis on the dynamic decision dataset of emergency passage corridors to generate corridor contingency plan processing results that include estimated passage time, a list of affected intersections, and key bottleneck points.

[0033] In a specific embodiment of the present invention, the step of performing multi-dimensional path planning and spatiotemporal impact range analysis further includes: using an improved dynamic A* algorithm or Dijkstra algorithm, with the constraints of minimum travel time, most balanced traffic load, and construction avoidance, to plan no fewer than three candidate alternative paths.

[0034] Specifically, in order to generate a travel plan, the system uses an improved dynamic A* algorithm or Dijkstra's algorithm to search the road network for a path, generating no fewer than three candidate alternative paths.

[0035] Example 1: An improved dynamic A* algorithm is used to plan candidate alternative routes. The three constraints of minimizing travel time, maximizing traffic load balance, and avoiding construction are transformed into quantifiable computational logic for the algorithm. This ensures that the planned routes meet emergency traffic needs and generate at least three differentiated candidate routes. The specific implementation process is as follows: First, the urban road network is abstracted as a directed weighted graph, where road network nodes correspond to intersections and road segment endpoints, and road network edges correspond to actual road segments. Three types of core weights are assigned to each edge: For the construction avoidance constraint, the travel cost of the edge of the construction segment is set to infinity, forcing the algorithm to skip this type of road segment; For the minimum travel time constraint, the travel time of a single road segment is calculated based on the real-time length of the road segment, average vehicle speed, and traffic light waiting time, and the total travel time of the path is obtained by summing them; For the maximum traffic load balance constraint, the road segment load is quantified by the ratio of the real-time lane occupancy rate to the maximum capacity occupancy rate, and the load balance cost is calculated by the sum of the squared deviations between the load of each road segment and the regional average load. Secondly, the traditional A algorithm is improved by upgrading the cost function to a multi-objective fusion form. Travel time is assigned a weight of 0.6, and load balancing cost a weight of 0.4, while retaining the heuristic function to ensure algorithm convergence efficiency. Finally, a K-shortest path generation strategy is adopted. By dynamically adjusting the segment weights of selected paths, redundant planning is avoided. After generating initial candidate paths, redundant paths with a similarity exceeding 80% are eliminated. Ultimately, at least three candidate paths that satisfy all constraints and whose path orientation differs significantly from travel characteristics are selected, providing a foundation for subsequent optimal path evaluation.

[0036] Example 2: This embodiment also provides a candidate path planning implementation based on an improved Dijkstra algorithm. It uses minimum travel time, optimal traffic load balance, and construction avoidance as constraints to ensure the generation of at least three differentiated candidate paths. The specific process is as follows: First, the urban road network is abstracted into a directed weighted graph model. Road network nodes correspond to key locations such as intersections and road segment endpoints, and road network edges correspond to actual connected road segments. Three types of constraint-related weight parameters are configured for each edge: For construction avoidance constraints, information on construction road segments is obtained from the traffic control database, and edges corresponding to such road segments are marked as prohibited from passage, with the passage cost set to infinity, forcing the algorithm to directly avoid them during path search; For minimum travel time constraints, the travel time of a single road segment is calculated based on the real-time length, average speed, and traffic light phase data of the road segment in the dynamic decision dataset, and the total path travel time is accumulated; For optimal traffic load balance constraints, the load of a single road segment is quantified by the ratio of the real-time lane occupancy rate to the maximum capacity occupancy rate, and the load balance cost is calculated by the sum of the squared deviations between the load of each road segment and the regional average load. Secondly, the traditional Dijkstra algorithm is improved by upgrading the single distance cost function to a multi-objective fusion cumulative cost function, i.e., path cumulative cost = 0.6 × cumulative travel time + 0.4 × load balancing cost. This function quantifies the overall merits of the path. Finally, a K-shortest path strategy with multi-round search and dynamic weight adjustment is adopted. After the algorithm obtains the overall optimal path in the first run, the weights of the road segments traversed by the path are slightly increased, such as increasing the travel time weight by 12%, to avoid the algorithm repeatedly generating the same source path. After multiple iterations of the algorithm, candidate paths with a path similarity of less than 80%, all meeting construction avoidance requirements, and whose travel time and load balancing meet the constraint thresholds are selected. Finally, no less than three differentiated paths are retained to provide sufficient alternatives for subsequent optimal path evaluation and selection.

[0037] Establish a spatiotemporal conflict prediction model, analyze the impact of the physical space occupied by emergency vehicles in each future time slice on the flow rate of surrounding social vehicles based on the predicted trajectory of emergency vehicles, and identify key bottlenecks that may cause traffic congestion.

[0038] Specifically, a spatiotemporal conflict prediction model is established for each candidate route. This model analyzes the impact on the surrounding traffic flow based on the predicted trajectory of emergency vehicles, i.e., the road space positions that the vehicles will occupy at different points in the future. Through simulation, the system can identify traffic nodes that are prone to traffic congestion due to emergency vehicle intervention, i.e., those with severe congestion, due to narrow lanes, complex intersection designs, or huge traffic volumes, and mark them as key bottlenecks. The specific implementation process is as follows: First, based on the road network topology data of the candidate routes and the preset driving parameters of emergency vehicles, such as speed, acceleration, and heading angle, the predicted trajectory of emergency vehicles is generated. This trajectory is divided into 1-second time slices, and the road space positions that emergency vehicles will occupy in each time slice include the area occupied by the vehicle body and the safety protection area of ​​≥1.5 meters on both sides, which is mapped to the lane-level coordinate system of the high-precision map. Secondly, real-time traffic flow data of social vehicles along the candidate path and within a 150-meter radius within each time slice is obtained, in units of vehicles / minute / lane. Road network characteristic parameters such as number of lanes, road width, intersection turning design, and channelization structure are also included. The changes in the driving behavior of social vehicles after emergency vehicles occupy the target space are simulated through a micro-traffic simulation model, including lane change frequency, speed reduction, and capacity reduction. The inhibitory effect of emergency vehicles on the traffic flow of social vehicles is quantified by weighted fusion. Finally, a bottleneck point determination threshold is set: when a traffic node, such as a road segment or intersection, experiences a decrease in social vehicle flow rate of ≥50%, queue length of ≥50 meters, or congestion duration of ≥2 minutes in the simulation, or when the lane narrows (e.g., ≤2 lanes in one direction), the intersection design is complex (e.g., no signal control + multiple turning traffic convergence), or the initial traffic volume is huge (e.g., flow rate ≥120 vehicles / minute, lane-induced conflict frequency ≥3 times / minute), the node is marked as a critical bottleneck point, and its spatiotemporal information, including the time of occurrence, the road segment, and the scope of impact, is recorded simultaneously, providing core decision-making basis for risk assessment of subsequent candidate paths and selection of the optimal path.

[0039] The affected area after the corridor is generated is calculated based on traffic flow fluctuation theory. The impact level of the affected intersection is defined, and the impact level is weighted and evaluated according to the intersection saturation, turning ratio and pedestrian crossing demand.

[0040] Specifically, to determine the impact boundary of the traffic corridor on the entire traffic system, the system uses traffic flow fluctuation theory for calculation. This theory describes the propagation process of traffic congestion, and the system calculates the extent of the traffic pressure wave after the corridor is generated. Based on this, the system defines an impact level for each affected intersection. This impact level is not a single indicator, but a comprehensive score obtained by weighting and evaluating the intersection's real-time traffic saturation, the turning ratio in each direction, and pedestrian crossing demand. This provides a basis for subsequent signal control priority ranking. The specific implementation process is as follows: First, the impact range is calculated using the shock wave model in traffic flow fluctuation theory: After the emergency traffic corridor is generated, the priority passage of emergency vehicles will cause sudden changes in the traffic flow state of surrounding lanes, such as a sudden drop in vehicle speed and traffic compression, forming a backward-propagating traffic wave. Based on the initial traffic density of the road segment, the speed of emergency vehicles, and the number of lanes occupied by the corridor in the dynamic decision dataset, the impact range is calculated using the formula... , The speed of traffic wave propagation is derived from the relationship between traffic density and vehicle speed. To determine the duration of emergency vehicles passing through the road segment, the maximum distance of traffic wave propagation is calculated. All intersections within this distance are defined as affected intersections, thus clarifying the spatial boundaries of the affected area. Secondly, for each affected intersection, a weighted assessment is performed based on three indicators: intersection saturation, turning ratio, and pedestrian crossing demand, to determine the impact level. Intersection saturation is quantified as follows: , This represents the real-time total traffic flow at the intersection, in vehicles per minute. The traffic capacity of the intersection is designed based on the number of lanes and signal cycle, with a value ranging from 0 to 1. The closer the value is to 1, the more severe the congestion. Turning ratio is quantified. , This refers to the total turning traffic flow at the intersection, which includes both left-turning and right-turning traffic. The time frame is determined to be a significant turning conflict, with a value range of 0-1; pedestrian crossing demand is quantified as follows: , The number of pedestrians crossing the street within a unit of time, such as 1 minute. Green light duration for pedestrians crossing the street The time frame is determined to be high pedestrian crossing demand, with a value range of 0-1. Finally, the weights of the three indicators are set, pre-determined based on measured data from the intersection: for example, intersection saturation weight is 0.4, turning ratio weight is 0.3, and pedestrian crossing demand weight is 0.3. The overall impact score is then calculated. The impact level is divided into four levels based on the score: Level 1, meaning slight impact. Level 2, meaning moderate impact. Level 3, meaning severely impacted. Level 4 indicates extremely severe impact. The level result is directly used to formulate subsequent traffic signal control strategies. For example, at level 3-4 intersections, the green light phase for emergency vehicles needs to be extended and the release sequence of turning traffic needs to be optimized.

[0041] It should be noted that the above method enhances the scientific rigor and robustness of emergency corridor solutions by combining multi-constraint path planning with forward-looking spatiotemporal impact analysis. It moves beyond simply selecting the shortest path; instead, it pre-plans multiple alternative paths and conducts in-depth evaluations, ensuring the solution's resilience in the face of unforeseen circumstances. Utilizing spatiotemporal conflict prediction models, potential traffic congestion risks can be transformed from post-event discovery to pre-event warnings, enabling control strategies to proactively avoid or mitigate congestion at key bottlenecks. Simultaneously, the precise definition of the impact range and level based on traffic flow fluctuation theory allows the system to achieve differentiated and refined management of traffic resources, avoiding unnecessary traffic disturbances caused by a one-size-fits-all approach. This ensures priority passage for emergency vehicles while reducing negative impacts on general traffic.

[0042] S3. Input the corridor contingency plan processing results into the pre-constructed spatiotemporal resource dynamic allocation optimization model, and output the phase switching timing scheme for a series of traffic lights along the route, as well as the graded guidance suggestions generated for social vehicles in the corridor.

[0043] In a specific embodiment of the present invention, the workflow of the spatiotemporal resource dynamic allocation optimization model further includes: screening target candidate paths from candidate alternative paths, and quantifying the time resources required by emergency vehicles, namely the green light phase adaptation window, green light adjustment range, and spatial resources, namely the number of temporary dedicated lanes and bottleneck avoidance range, based on the emergency passage corridor dynamic decision dataset and the spatiotemporal information of the target candidate paths.

[0044] For example, the selection of target candidate paths is based on the generation of no less than three candidate alternative paths. A multi-dimensional evaluation system is constructed, and the number of key bottleneck points, the impact level of intersections, and the comprehensive optimization cost, such as 0.6 × travel time + 0.4 × traffic load balancing cost, are selected as evaluation indicators. Among them, the number of key bottleneck points has a weight of 0.3, the impact level of intersections has a weight of 0.3, and the comprehensive optimization cost has a weight of 0.4. The comprehensive score of each candidate path is calculated, and the path with the highest comprehensive score and satisfying the following conditions is selected as the target candidate path: ≤2 key bottleneck points, intersection impact level is mainly 1-2, and comprehensive optimization cost ≤ the optimal path cost × 1.1.

[0045] Time resource quantification is based on the dynamic decision-making dataset of emergency passage corridors and the spatiotemporal information of target candidate routes. By using the real-time location, speed, and route length of emergency vehicles, as well as the current phase data of traffic lights, the estimated time window for emergency vehicles to arrive at each intersection is calculated, i.e., the green light phase adaptation window. Then, combined with the deviation between this window and the current green light period, the green light adjustment range is calculated, i.e., the lengthening or shortening of the green light duration, and the adjustment range is controlled within 30 seconds to avoid excessive impact on social traffic. Finally, spatial resource quantification is based on the lane topology data, traffic flow density, and key bottleneck point distribution of the target candidate routes, combined with the road coordinates of high-precision maps, to determine the number of temporary dedicated lanes, i.e., ≤1 lane in one direction. The lane with the lowest traffic flow density is selected first. At the same time, the bottleneck point avoidance range radiating outward ≤150 meters is defined with the road coordinates of the key bottleneck point as the center. The start and end points of the control are accurately marked by road markers to ensure that the dedicated passage space for emergency vehicles does not conflict with the driving area of ​​social vehicles.

[0046] Construct a two-level multi-objective optimization function with the objective of minimizing the increase in total delays of social vehicles and the zero-delay passage of emergency vehicles.

[0047] For example, the core objective boundaries of the two-layer optimization are clearly defined. The upper-layer objective focuses on global coordination, with zero-delay passage for emergency vehicles as the core, such as a quantitative standard of ≤2 minutes for the entire passage delay of emergency vehicles, while also considering the efficiency of green wave coordination at intersections. The lower-layer objective focuses on local adaptation, with minimizing the increase in total delay for social vehicles as the core, such as a quantitative standard of ≤20% increase in average delay for social vehicles, while also satisfying the constraint of a ≥50% reduction in the frequency of bottleneck point conflicts. A mathematical optimization function is constructed: the upper-layer optimization function is... ,in The cumulative delays in passage for emergency vehicles The sum of the green wave phase deviations at each intersection is represented by w1=0.7 and w2=0.3, which are weights determined by fitting cross-intersection collaborative test data.

[0048] The lower-level optimization function is ,in This represents the increase in total delays for social vehicles. The intersection conflict frequency is represented by weights w3=0.6 and w4=0.4, which are dynamically adjusted based on the intersection's impact level. Finally, hard constraints are embedded in the function: construction zone space resources are unavailable, and pedestrian green light duration is ≥15 seconds, ensuring that the optimization process does not exceed safety and compliance limits. This two-layer multi-objective optimization function provides a unified quantitative calculation basis for subsequent upper-layer green wave planning and lower-layer intersection fine-tuning, achieving a balance between global efficiency and local adaptation.

[0049] The upper-level optimization model calculates the center time offset and bandwidth of the moving green wave based on the estimated time window of emergency vehicles arriving at each intersection, generates an asymmetric phase compensation scheme, and simultaneously outputs the activation time of the temporary dedicated lane, the lane number occupied, and the spatial control range of the bottleneck point.

[0050] The spatial control range of the bottleneck point is defined by road coordinates.

[0051] Specifically, the model calculates the offset of the center time when the green light needs to be turned on at each intersection, i.e., the amount by which the green light start time is advanced or delayed relative to a reference time, to ensure that the center time of the green wave is precisely aligned with the arrival time of emergency vehicles. Simultaneously, the model also calculates the bandwidth of the green wave, i.e., the duration of the green light, to ensure it is sufficient for emergency vehicles to pass smoothly. This series of calculations forms an asymmetric phase compensation scheme, serving as the top-level guiding strategy for the coordinated operation of all traffic lights along the route. Asymmetric phase compensation reflects the dynamic adaptability of the scheme, solving the poor adaptability problem caused by the traditional one-size-fits-all approach to green wave switching.

[0052] Based on the estimated arrival time windows of emergency vehicles at each intersection, and combined with the goal of zero-delay passage of emergency vehicles in the multi-objective optimization function, the center time offset of the moving green wave is calculated. This is done using the intersection traffic light reference phase period as a reference, through the formula... Obtain the center time offset of the moving green band. Estimate the arrival time of emergency vehicles. As the baseline green light start time, if A positive value delays the green light start, while a negative value starts it earlier, ensuring precise alignment between the green wave center time and vehicle arrival time. Simultaneously, based on the green wave coordination efficiency target in the optimization function and the emergency vehicle's length and speed, the green wave bandwidth is calculated. , The time it takes for a vehicle to completely pass through the intersection. For safety redundancy, a 2-second buffer is used to ensure unimpeded passage for emergency vehicles. Secondly, the asymmetric phase compensation scheme is generated by balancing emergency efficiency and social traffic impact through a multi-objective optimization function: for intersections with impact levels of 1-2 (i.e., low traffic volume), the bandwidth is controlled at 10-15 seconds, and the green light start time offset adjustment is ≤10 seconds; for intersections with impact levels of 3-4 (i.e., high traffic volume), the bandwidth is extended to 20-25 seconds, and the green light start time offset adjustment is ≤15 seconds, avoiding delays exceeding the ≤20% constraint threshold caused by green wave adjustments. Finally, the synchronized output of spatial resource information, namely the activation time of the temporary dedicated lane, the occupied lane number, and the bottleneck control range, is also guided by the optimization function. The activation time is synchronized with the offset of the green wave center time, that is, it is activated 3 seconds before the green light starts. The occupied lane number prioritizes the lane with the lowest social traffic density, which is in line with the goal of minimizing social vehicle delays. The bottleneck spatial control range is precisely delineated according to the requirement of a ≥50% reduction in conflict frequency in the optimization function, combined with high-precision map coordinates, to ensure that the allocation of spatiotemporal resources is completely consistent with the multi-objective optimization goal, and to achieve the collaborative optimization of time green wave and space-specific features.

[0053] The lower-level optimization model fine-tunes the signal timing for specific intersections by inserting dedicated phases, extending the current green light, or shortening the conflict phase red light, thereby achieving on-demand allocation of spatiotemporal resources and simultaneously outputting temporary turning restriction instructions for intersections, boundaries of priority passage areas for emergency vehicles, and guidance for yield lanes for social vehicles.

[0054] Specifically, the asymmetric phase compensation scheme is deployed to the lower-level optimization model, which is responsible for fine-tuning the signal timing at each specific intersection. Based on the current traffic conditions at the intersection, three strategies are selected to achieve on-demand allocation of spatiotemporal resources. These strategies include temporarily inserting a dedicated green light phase for the direction of emergency vehicles, extending the current green light time in the direction of an emergency vehicle as it approaches, or shortening the red light time in conflicting directions to clear traffic flow in advance. These three strategies cover different scenarios, ensuring that adjustments at each intersection are precisely adapted, rather than a uniform operation, reflecting the practicality of the technical solution. Furthermore, the two-layer architecture of upper-level global and lower-level local is one of the inventive aspects of this invention, distinguishing it from the single-level optimization schemes in existing technologies.

[0055] For example, the model first acquires real-time status data of the current intersection, including the current signal light phase, traffic density in each direction, proportion of traffic in conflicting directions, and intersection impact level. Combining this with the core objectives of minimizing the total increase in social vehicle delays and reducing bottleneck conflict frequency by ≥50% in the multi-objective optimization function, the model adaptively selects the optimal solution from three signal control fine-tuning strategies: if there is no corresponding traffic phase when an emergency vehicle arrives, and the proportion of traffic in conflicting directions is ≤30%, then a dedicated phase insertion strategy is triggered. The duration of the dedicated phase = emergency vehicle passage time + 2 seconds of safety redundancy. The estimated arrival time of emergency vehicles is calculated based on vehicle length and driving speed, ranging from 8 to 12 seconds. If the estimated arrival time of emergency vehicles deviates from the current green light end time by ≤5 seconds, and the current queue length in the lane is ≤30 meters, then the current green light will be extended. The extension time = deviation time + 1 second safety redundancy, and the total extension time shall not exceed 15 seconds to avoid exceeding the constraint that the increase in delays of social vehicles is ≤20%. If the proportion of traffic flow in the conflict direction is ≤20%, and the remaining red light time is ≥8 seconds, then the red light strategy for shortening the conflict phase will be adopted. The shortening amount = emergency vehicle arrival lead time - 3 seconds clearance time to ensure that the conflict area is cleared in advance. Secondly, the spatial resource instructions output synchronously are all generated around the optimization function objectives: temporary turning restriction instructions prioritize the turns with the highest conflict frequency; for example, if the left-turn conflict ratio is ≥40%, left turns are prohibited, and the restriction duration is completely synchronized with the signal control fine-tuning cycle; the boundary of the emergency vehicle priority passage area is delineated based on the lane coordinates of the high-precision map, such as the center lane from K1+250 to K1+300, ensuring that it is consistent with the direction of passage of the green light phase after fine-tuning; the guidance for social vehicles to avoid lanes prioritizes recommending the lanes with the lowest traffic density, such as density ≤20 vehicles / km, guiding vehicles to change lanes to the sides to avoid occupying the emergency priority area. The entire process uses a multi-objective optimization function to quantitatively evaluate the impact of each operation on social vehicle delays and conflict frequency, ensuring that signal control fine-tuning and spatial instructions meet the needs of rapid passage of emergency vehicles without exceeding the social traffic delay constraint threshold, and achieving on-demand adaptation of spatiotemporal resources at local intersections.

[0056] Introducing safety redundancy interval constraints ensures that the intersection clearing time meets the minimum safe driving distance requirement during rapid phase switching, and outputs a timing table for the coordinated execution of space resource occupancy and time resource adjustment.

[0057] It should be noted that the safety redundancy interval, or buffer time reserved during phase switching (e.g., 2 seconds), ensures that vehicles from the previous phase have completely exited the intersection before vehicles from the next phase enter. The intersection clearing time refers to the time from the issuance of the phase switching command to the exit of the last vehicle. Specifically, the implementation logic is that the safety redundancy interval is automatically included in the calculation when generating the phase adjustment plan. For example, if the original requirement was to shorten the red light of the conflicting phase by 5 seconds, due to safety constraints, the shortening is adjusted to 3 seconds, with a 2-second clearing time reserved. The safety redundancy interval constraint is a complete supplement to the technical solution, addressing the shortcomings of traditional emergency traffic control that prioritize efficiency over safety, ensuring the solution can be directly implemented. The duration of this safety redundancy interval constraint is calculated based on the intersection width W, the average speed v of vehicles, and the braking deceleration a. .

[0058] It should be noted that this method resolves the conflict between emergency vehicle traffic efficiency and social traffic order by establishing a two-layer multi-objective optimization model. The upper-layer model, from a global perspective, plans the optimal moving green wave, ensuring the macroscopic continuity and overall efficiency of the emergency corridor. The lower-layer model delves into the microscopic level of each intersection, making refined adjustments based on local conditions, achieving precise allocation of spatiotemporal resources. This hierarchical collaborative optimization mechanism not only ensures uninterrupted passage of emergency vehicles with the highest priority but also proactively and minimizes interference with other vehicles, avoiding the widespread traffic congestion caused by the extensive control methods in traditional emergency modes. Simultaneously, the introduction of a safety redundancy interval constraint, a hard safety constraint, adds a safety lock to the entire high-speed dynamic control process, ensuring that the improvement in traffic efficiency is carried out under safe conditions, reflecting the high completeness and reliability of the system design.

[0059] It should also be noted that in the process of solving the aforementioned spatiotemporal resource dynamic allocation optimization model, the safety redundancy interval constraint is a hard constraint with the highest priority. If the optimization solution cannot simultaneously meet the requirements of zero-delay passage for emergency vehicles and the safety redundancy interval, the safety interval will be prioritized, and this conflict event will be recorded and compensated for by appropriately adjusting the green wave bandwidth or activation time.

[0060] In a specific embodiment of the present invention, the step of generating graded guidance suggestion results further includes: dividing the guidance suggestion into three levels: warning zone, guidance zone and clearing zone, based on the distance between the location of the emergency vehicle and the social vehicles ahead.

[0061] In the warning zone, voice prompts are broadcast to vehicles within a 3-kilometer radius via mobile internet apps, advising them to plan detours in advance to reduce pressure on the core area of ​​the corridor.

[0062] In the guidance area, roadside variable message signs are used to prompt vehicles to maintain their speed or divert to the sides, reserving a center lane or emergency avoidance space for emergency vehicles.

[0063] In the clearing zone, such as within 100 meters behind and 500 meters ahead of emergency vehicles, specific driving behavior instructions are generated, including changing lanes to the left, slowing down and moving to the right, or passing through intersections to clear the queue of vehicles waiting to be transported, and these instructions are delivered within seconds via narrowband IoT or V2X communication.

[0064] This method establishes a tiered guidance system, progressing from distant to near, and from macro to micro levels, enabling refined and differentiated management of social vehicle behavior. It changes the traditional emergency passage system's reliance on delayed information dissemination, ambiguous scope, and singular instructions. Remote early warning and diversion effectively reduce traffic density in core corridors, mid-range guidance systematically shapes passage space, and precise near-end instructions ensure immediate clearing and avoidance. This progressive, multi-pronged guidance strategy not only improves the cooperation and avoidance efficiency of social vehicles, reducing panic and secondary accident risks caused by information asymmetry, but also creates a safer, smoother, and more reliable passage environment for emergency vehicles.

[0065] S4. Based on the judgment rules of real-time traffic flow evolution, the results of the graded guidance suggestions are dynamically adjusted and judged. When it is determined that there is unexpected congestion ahead that causes the corridor to be blocked, the subsequent backup route planning and guidance scheme update processing are triggered.

[0066] In a specific embodiment of the present invention, the step of dynamically adjusting and determining the graded guidance suggestion result based on the judgment rule of real-time traffic flow evolution further includes: setting a traffic flow anomaly monitoring threshold, comparing the preset traffic state with the real state of visual perception feedback in real time, and triggering an early warning logic when a sudden accident, secondary congestion or signal failure occurs on the target road segment, resulting in a decrease in traffic efficiency exceeding the traffic efficiency decline rate, such as 30%, and the state remains stable for more than a preset time.

[0067] The detection of sudden accidents, secondary congestion, and signal failures on the target road section, as well as the quantitative judgment of a traffic efficiency reduction of more than 30%, described in this step, are all achieved by relying on existing mature technology systems such as city-level traffic AI visual perception network, traffic signal controller status monitoring system, and road network traffic flow big data analysis platform. The detection logic of each abnormal state and the quantitative calculation method of traffic efficiency are conventional technical means in this field and will not be described in detail here.

[0068] The dynamic programming algorithm is used to reconstruct the expected arrival time window of emergency vehicles in real time, and to evaluate the coverage of the current signal phase scheme for the new time window.

[0069] Specifically, based on the current location and speed of the emergency vehicle and the new conditions of the blocked road ahead, the estimated arrival time window of the emergency vehicle at each subsequent intersection is reconstructed and updated in real time. The coverage of the currently implemented signal phase scheme for this new time window is evaluated, that is, whether the original green wave scheme can still guarantee that the emergency vehicle can pass smoothly within the newly predicted arrival time.

[0070] The coverage rate refers to the proportion of the overlap between the new time window and the current green light period to the total duration of the new time window. A coverage rate of 100% means that the new time window falls entirely within the green light period, and the existing solution can guarantee zero-delay passage for emergency vehicles. Otherwise, it is determined that the solution is not compatible.

[0071] If the current plan cannot meet the requirement that emergency vehicles can pass continuously at the set target speed, the backup route recalculation process will be initiated, and the signal coordination control parameters of all affected intersections along the route will be updated simultaneously.

[0072] It should be noted that this method, by introducing a closed-loop real-time monitoring and dynamic adjustment mechanism, greatly enhances the robustness and adaptability of emergency passage corridors in complex and ever-changing traffic environments. It enables the system to self-correct, shifting from passive response to proactive prediction and rapid response. When the pre-planned solution fails due to an emergency, it can quickly switch to the optimal backup plan, avoiding the risk of emergency vehicles getting stuck in unexpected congestion. This dynamic reconstruction and collaborative updating based on real-time data feedback ensures that the emergency passage corridor always maintains its most efficient operating state regardless of changes in road conditions, providing uninterrupted and reliable support for the successful completion of emergency missions.

[0073] S5. The subsequent update processing results are simultaneously sent out, real-time control commands are sent to the signal control system to form a mobile green wave, and avoidance prompts are pushed to roadside units or mobile Internet terminals to generate priority passage corridors for emergency vehicles.

[0074] In a specific embodiment of the present invention, the step of sending real-time control commands to the signal control system to form a moving green wave further includes: based on the real-time speed feedback of the emergency vehicle, using an adaptive predictive control algorithm to dynamically adjust the phase sequence and phase difference of the green wave band, thereby achieving a vehicle-moving wave-following green wave control.

[0075] Specifically, once the updated processing result is confirmed and issued, the system begins sending real-time control commands to the signal control system to form a moving green wave. The core of this process is based on real-time speed feedback from emergency vehicles, dynamically adjusting the green wave band by running an adaptive predictive control algorithm. This algorithm continuously predicts the precise arrival time of emergency vehicles at the next intersection and accordingly fine-tunes the phase sequence and phase difference of the traffic lights at that intersection—that is, the order in which the green lights turn on and their offset time relative to the standard cycle. This achieves a following-type green wave control effect, ensuring that the green light always turns on precisely before the emergency vehicle arrives.

[0076] It should also be noted that the adjustment logic of the adaptive predictive control algorithm is consistent with the two-layer multi-objective optimization objective of the dynamic allocation optimization model of spatiotemporal resources mentioned above. While achieving the smooth flow of emergency vehicles, it also takes into account the constraint that the increase in delay of social vehicles should be ≤20%. In multi-intersection collaborative control, a cascading triggering mechanism is adopted. When an emergency vehicle leaves the current intersection and enters the next road segment, the current intersection immediately executes the phase restoration logic. By reducing the passage time lost in the early stage by subsequent non-conflicting phases, the traffic flow can be rapidly self-healed.

[0077] Specifically, when sensors or visual perception confirm that an emergency vehicle has safely left the current intersection and entered the next road segment, the current intersection will immediately be triggered to execute phase recovery logic. This logic compensates for the time previously occupied by prioritizing emergency vehicle passage by intelligently reducing the green light duration of subsequent non-conflicting direction phases, thereby achieving rapid self-healing of traffic flow and minimizing the impact.

[0078] It should also be noted that the intelligent reduction of the green light duration for subsequent non-conflict direction phases is specifically based on the compensation benchmark of the extra green light duration previously occupied for emergency vehicles. Relying on parameters such as lane occupancy rate and vehicle queue length in non-conflict direction monitored in real time at the intersection, the duration is reduced slightly phase by phase based on the original green light timing, with a reduction of no more than 5 seconds per phase, and ensuring that the green light duration for a single non-conflict direction is not lower than the standard minimum timing in the field of traffic management. At the same time, the traffic flow status in this direction is monitored in real time, and if the queue worsens, the reduction is stopped immediately, ensuring that while completing the time compensation, new congestion in non-conflict directions is avoided, and traffic flow can quickly heal itself.

[0079] The control commands include phase-locking commands, forced jump commands, and queuing dissipation control commands based on oversaturated flow.

[0080] Specifically, the phase lock command is used to keep the green light in the direction of emergency vehicles constantly on, the forced jump command is used to immediately cut off the current phase and switch to the designated phase, and the queue dissipation control command is used to issue control commands specifically for quickly clearing queued vehicles in response to oversaturated flow caused by avoidance during the traffic recovery phase.

[0081] It should be noted that this method, through adaptive control that follows vehicle movement, upgrades the traditional fixed green wave to a mobile green wave that is tightly coupled with the real-time dynamics of emergency vehicles. This enables on-demand allocation of spatial and temporal resources, ensuring high efficiency and zero delays for emergency vehicle passage. Simultaneously, its unique cascaded triggering and phase recovery mechanism limits the impact of traffic control to the instant the emergency vehicle passes, and rapidly initiates a self-repair procedure after the vehicle has passed, shortening the time for traffic order to recover and avoiding secondary traffic congestion caused by prioritizing emergency response. Thus, a balance is achieved between ensuring emergency tasks and maintaining social traffic stability.

[0082] In a specific embodiment of the present invention, the step of pushing avoidance prompts to roadside units or mobile Internet terminals further includes: establishing a geofence-based precise information push mechanism, and issuing instructions only to specific vehicle MAC addresses within the potential influence range of the corridor.

[0083] Using an HMI (Human-Machine Interface) to display the real-time relative position and planned trajectory of emergency vehicles to drivers of other vehicles in the form of AR navigation, the panic and misoperation of drivers caused by information asymmetry can be reduced.

[0084] By integrating real-time yielding feedback data through the public travel information service platform, digital incentive records are created for yielding vehicles, and the yielding behavior data is fed back to the control system to optimize subsequent guidance strategies.

[0085] It should be noted that avoidance behavior data, such as the vehicle's response time and the selected avoidance method, will be fed back to the control system's backend in real time to continuously optimize and iterate subsequent guidance strategies, making them more in line with actual driving behavior habits.

[0086] It should be noted that this method improves the efficiency and safety of collaborative avoidance by other vehicles during emergency passage by establishing a precise, intuitive information interaction system with a feedback loop. This avoids interference from irrelevant information to other vehicles and enhances the effectiveness of information transmission. The visualization provided by AR navigation improves drivers' perception of emergencies and decision-making efficiency, reducing safety risks during avoidance. Furthermore, the introduction of digital incentives and a data feedback loop not only increases the participation and cooperation of other vehicles through positive incentives but also endows the system with self-learning and evolution capabilities, enabling the entire emergency guidance system to continuously optimize and become more intelligent and user-friendly.

[0087] Reference Figure 2 As shown, the second aspect of the present invention provides a system for executing the AI-based visual analysis-based intelligent traffic flow trajectory optimization method, comprising: a holographic perception layer: composed of AI cameras, radar, and roadside computing units covering urban main roads, used to construct a high-precision traffic perception base.

[0088] It should be noted that by deploying AI cameras, radar and other sensors along the main urban roads, and in conjunction with roadside computing units, the road traffic is monitored around the clock and in all areas. These devices work together to build a high-precision traffic perception base that can accurately reflect the real traffic world.

[0089] Edge computing layer: Deployed near the intersection, it is responsible for real-time processing of visual feature streams, target recognition and local traffic status analysis.

[0090] It should be noted that the computing nodes are physically deployed near the intersection. Their core task is to receive and process massive visual feature streams from the holographic perception layer in real time, performing target recognition, trajectory tracking, and local traffic status analysis locally, thereby reducing the burden on the cloud and ensuring low-latency response. The processed structured data is then aggregated to the cloud control center.

[0091] Cloud Control Center: Includes a path planning engine, a spatiotemporal resource optimization model library, and a multi-source data integration bus, responsible for the generation and adjustment of global corridors.

[0092] It should be noted that the cloud control center integrates a path planning engine responsible for path selection, a model library that stores and runs dynamic allocation optimization models for spatiotemporal resources, and a multi-source data integration bus for integrating various types of information. This center is responsible for performing global optimal calculations and generating and dynamically adjusting the overall plan for emergency passage corridors.

[0093] Execution layer: includes traffic signal controller control interface, roadside V2X communication unit and mobile travel service backend, used for instruction issuance and feedback closed loop.

[0094] The execution layer is equipped with an instruction coordination arbitration module to ensure that signal control instructions and vehicle guidance instructions are issued synchronously based on a unified time stamp, and to prioritize the immediate delivery of instructions to key areas such as clearing zones.

[0095] It should be noted that the execution layer is responsible for translating the decisions of the cloud control center into actions in the physical world. This layer includes control interfaces that communicate directly with traffic signals, variable message signs and roadside V2X communication units for disseminating information to vehicles and infrastructure, and a mobile travel service backend responsible for pushing messages to public mobile applications. Through these components, instructions are accurately issued and feedback information is collected, forming a complete control and feedback closed loop.

[0096] It should also be noted that the system organically combines the high real-time sensing capabilities of front-end devices with the powerful global computing capabilities of the cloud platform by constructing a four-layer collaborative architecture of perception-edge-cloud-execution. The introduction of the edge computing layer improves the system's response speed to local traffic events, while the cloud control center ensures the global optimality of corridor planning and the scientific nature of the strategy. This layered, decoupled, and collaborative system design not only solves the bottleneck problem of massive data transmission and processing but also achieves end-to-end closed-loop control from data acquisition, analysis, decision-making to command execution, providing solid hardware and software support for the stable, efficient, and reliable operation of the entire emergency vehicle dynamic priority passage corridor generation and control method.

[0097] The technical principle of this invention lies in constructing a data-driven closed-loop dynamic collaborative control system. First, this method integrates city-level holographic visual perception network data with real-time status data of emergency vehicles to construct a comprehensive and accurate dynamic decision-making dataset, providing a solid foundation for all subsequent decisions. Based on this dataset, the system does not simply select a path, but performs multi-dimensional planning and pre-analysis to predict potential traffic bottlenecks and their impact range, generating scientific corridor plans. Its core is to decompose the complex traffic control problem into two levels—phase timing optimization for traffic lights and behavioral guidance optimization for social vehicles—through a spatiotemporal resource dynamic allocation optimization model, and then solve them collaboratively. Most importantly, this method is not a one-time static planning, but introduces a dynamic judgment and adjustment mechanism based on real-time traffic flow evolution, enabling the entire system to self-correct and iterate its solutions according to changes in actual road conditions, thus forming a complete and adaptive control closed loop from perception, decision-making, execution to feedback.

[0098] This invention establishes a clear priority passage corridor for emergency vehicles. Through precise mobile green wave control and proactive guidance for other vehicles, it ensures that emergency vehicles reach their destinations quickly and safely, improving the efficiency and success rate of emergency response. Secondly, this method effectively reduces the impact of emergencies on urban traffic networks. Through tiered guidance and refined control, it minimizes the impact on other vehicles, avoiding widespread and prolonged traffic congestion caused by harsh regulations, achieving a harmonious coexistence of emergency priority and social traffic order. Furthermore, the system's dynamic adjustment capability gives it high robustness. Even in the event of unexpected accidents or other unforeseen circumstances along the route, it can quickly plan backup plans, ensuring uninterrupted execution of emergency tasks. Overall, it enhances the intelligence level of the urban traffic emergency management system and its ability to cope with complex situations.

[0099] The various thresholds involved in this invention, such as the 30% traffic efficiency reduction threshold and the weights in the non-motorized vehicle interference coefficient weighted model, are all based on statistical analysis of historical traffic flow big data of the target city, multiple iterative tests on the micro traffic simulation platform, and calibration combined with the experience of traffic engineering experts to ensure their effectiveness and rationality in practical applications.

[0100] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented in software, the above embodiments can be implemented, in whole or in part, as a computer program product. Those skilled in the art will recognize that the algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0101] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0102] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0103] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A method for optimizing intelligent traffic flow trajectories based on AI visual analysis, characterized in that, include: S1. By using a city-level traffic AI visual perception network and an emergency vehicle positioning and status vehicle unit, the real-time location, speed, destination data features of emergency vehicles, as well as traffic flow status data features on the predetermined path ahead of the emergency vehicles are obtained in real time, and an emergency passage corridor dynamic decision dataset is constructed. S2. Perform multi-dimensional path planning and spatiotemporal impact range analysis on the dynamic decision dataset of emergency passage corridors to generate corridor contingency plan processing results that include estimated passage time, a list of affected intersections and key bottlenecks. S3. Input the corridor contingency plan processing results into the pre-constructed spatiotemporal resource dynamic allocation optimization model, and output the phase switching timing scheme for a series of traffic lights along the route, as well as the graded guidance suggestion results generated for social vehicles in the corridor. S4. The results of the graded guidance suggestions are dynamically adjusted and determined based on the judgment rules of real-time traffic flow evolution. When it is determined that the corridor is blocked, the subsequent backup route planning and guidance scheme update processing are triggered. S5. The subsequent update processing results are simultaneously sent out, real-time control commands are sent to the signal control system to form a mobile green wave, and avoidance prompts are pushed to roadside units or mobile Internet terminals to generate priority passage corridors for emergency vehicles.

2. The intelligent traffic flow trajectory optimization method based on AI visual analysis according to claim 1, characterized in that, The step of constructing the dynamic decision dataset for emergency access corridors further includes: By utilizing panoramic cameras and edge computing nodes in a city-level traffic AI visual perception network, semantic segmentation and target tracking are performed on the area covered by the emergency vehicle's driving path, and traffic flow status data features including lane occupancy, average queue length, social vehicle density distribution, and non-motorized vehicle interference coefficient are extracted. The real-time location, speed, destination, vehicle type identification, mission urgency, instantaneous acceleration, heading angle, and estimated arrival time of emergency vehicles are obtained through the vehicle positioning and status onboard unit. Combined with high-precision map information, the dynamic coordinates of emergency vehicles are mapped to the lane-level road network topology. A multi-source data fusion algorithm is used to synchronize and register visual perception data with vehicle-mounted unit data in time and space, eliminating perception blind spots and forming a dynamic decision dataset covering all elements of vehicle-road-environment.

3. The intelligent traffic flow trajectory optimization method based on AI visual analysis according to claim 1, characterized in that, The steps for performing multi-dimensional path planning and spatiotemporal impact range analysis further include: Using the improved dynamic A* algorithm or Dijkstra algorithm, with the constraints of minimum travel time, most balanced traffic load, and construction avoidance, no fewer than three candidate alternative routes are planned. Establish a spatiotemporal conflict prediction model, analyze the impact of the physical space occupied by emergency vehicles in each future time slice on the flow rate of surrounding social vehicles based on the predicted trajectory of emergency vehicles, and identify key bottlenecks that may cause traffic congestion. The affected area after the corridor is generated is calculated based on traffic flow fluctuation theory. The impact level of the affected intersection is defined, and the impact level is weighted and evaluated according to the intersection saturation, turning ratio and pedestrian crossing demand.

4. The intelligent traffic flow trajectory optimization method based on AI visual analysis according to claim 1, characterized in that, The workflow of the spatiotemporal resource dynamic allocation optimization model further includes: Select target candidate routes from candidate alternative routes, and quantify the time and space resources required by emergency vehicles based on the dynamic decision dataset of emergency access corridors and the spatiotemporal information of target candidate routes. Construct a two-level multi-objective optimization function with the objective of minimizing the increase in total delays of social vehicles and the zero-delay passage of emergency vehicles. The upper-level optimization model calculates the center time offset and bandwidth of the moving green wave based on the estimated time window of emergency vehicles arriving at each intersection, generates an asymmetric phase compensation scheme, and simultaneously outputs the activation time of the temporary dedicated lane, the lane number occupied, and the spatial control range of the bottleneck point. The lower-level optimization model fine-tunes the signal timing for specific intersections by inserting dedicated phases, extending the current green light, or shortening the conflicting phase red light, thereby achieving on-demand allocation of spatiotemporal resources and simultaneously outputting temporary turning restriction instructions for intersections, boundaries of priority passage areas for emergency vehicles, and guidance for yield lanes for social vehicles. Introducing safety redundancy interval constraints ensures that the intersection clearing time meets the minimum safe driving distance requirement during rapid phase switching, and outputs a timing table for the coordinated execution of space resource occupancy and time resource adjustment.

5. The intelligent traffic flow trajectory optimization method based on AI visual analysis according to claim 1, characterized in that, The step of generating hierarchical induction recommendation results further includes: Based on the distance between the emergency vehicle's location and other vehicles ahead, guidance suggestions are divided into three levels: warning zone, guidance zone, and clearing zone. In the warning zone, voice prompts are broadcast to vehicles via mobile internet apps, advising them to plan detour routes in advance to reduce pressure on the core area of ​​the corridor. In the guidance area, roadside variable message signs are used to prompt vehicles to maintain their speed or divert to the sides, and to reserve the center lane or emergency avoidance space for emergency vehicles. In the clearing zone, specific driving behavior instructions are generated, including changing lanes to the left, slowing down and moving to the right, or passing through intersections to clear the queue of vehicles waiting to be transported, and these instructions are delivered within seconds via narrowband IoT or V2X communication.

6. The intelligent traffic flow trajectory optimization method based on AI visual analysis according to claim 1, characterized in that, The step of dynamically adjusting the tiered guidance suggestion results based on the judgment rule of real-time traffic flow evolution further includes: Set traffic flow anomaly monitoring thresholds, compare the preset traffic status with the actual status of visual perception feedback in real time, and trigger the warning logic when a sudden accident, secondary congestion or signal failure occurs on the target road segment, causing the traffic efficiency to decrease by more than the preset traffic efficiency decline rate threshold and the status remains stable for more than the preset time. The dynamic programming algorithm is used to reconstruct the expected arrival time window of emergency vehicles in real time, and to evaluate the coverage of the current signal phase scheme for the new time window. If the current plan cannot meet the requirement that emergency vehicles can pass continuously at the set target speed, the backup route recalculation process will be initiated, and the signal coordination control parameters of all affected intersections along the route will be updated simultaneously.

7. The intelligent traffic flow trajectory optimization method based on AI visual analysis according to claim 1, characterized in that, The step of sending real-time control commands to the signal control system to form a moving green wave further includes: Based on the real-time speed feedback of emergency vehicles, the phase sequence and phase difference of the green wave are dynamically adjusted using an adaptive predictive control algorithm to achieve vehicle-moving wave-following green wave control. In multi-intersection collaborative control, a cascading triggering mechanism is adopted. When an emergency vehicle leaves the current intersection and enters the next road segment, the current intersection immediately executes the phase restoration logic to compensate for the previous lost passage time by reducing the subsequent non-conflicting phases. The control commands include phase-locking commands, forced jump commands, and queuing dissipation control commands based on oversaturated flow.

8. The intelligent traffic flow trajectory optimization method based on AI visual analysis according to claim 1, characterized in that, The step of pushing avoidance prompts to roadside units or mobile internet terminals further includes: Establish a geofence-based precision information push mechanism to send instructions only to specific vehicle MAC addresses within the potential influence range of the corridor; Using the HMI interface, AR navigation is used to display the real-time relative position and planned trajectory of emergency vehicles to drivers of other vehicles. By integrating real-time yielding feedback data through the public travel information service platform, digital incentive records are created for yielding vehicles, and the yielding behavior data is fed back to the control system to optimize subsequent guidance strategies.

9. A system for executing the intelligent traffic flow trajectory optimization method based on AI visual analysis as described in any one of claims 1-8, characterized in that, include: Holographic perception layer: Composed of AI cameras, radar, and roadside computing units covering the main urban roads, used to build a high-precision traffic perception base; Edge computing layer: Deployed near the intersection, responsible for real-time processing of visual feature streams, target recognition and local traffic status analysis; Cloud Control Center: Includes a path planning engine, a spatiotemporal resource optimization model library, and a multi-source data integration bus, responsible for the generation and adjustment of global corridors; Execution layer: includes traffic signal controller control interface, roadside V2X communication unit and mobile travel service backend, used for instruction issuance and feedback closed loop.

Images