Lane passing unattended operation optimization method based on cloud edge collaboration

Abstract

The invention discloses a lane passing unattended operation optimization method based on cloud edge collaboration, and relates to the technical field of intelligent traffic, and the method comprises the steps: collecting multi-source heterogeneous data through an edge sensing unit, and forming a unified multi-dimensional passing parameter snapshot based on edge node preprocessing; performing basic characteristic disassembly and comprehensive analysis on the traffic flow to obtain a traffic characteristic comprehensive index; a dynamic threshold value is generated based on cloud historical data and real-time data, strategy generation is started through a conventional or emergency trigger mechanism, a cloud decision center generates a global strategy in combination with scene similarity matching and a multi-objective optimization algorithm, and verification is performed based on digital twinborn simulation; the execution instruction is analyzed based on the edge node, and meanwhile, multi-dimensional indexes are collected for effect verification and dynamic adjustment; the interference risk is effectively reduced, and the passing efficiency, order and safety under the unattended scene of the lane are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, specifically to an unattended lane passage optimization method based on cloud-edge collaboration. Background Technology

[0002] With the acceleration of urbanization and the surge in motor vehicle ownership, problems such as traffic congestion and low traffic efficiency have become increasingly prominent. Intelligent transportation, as a core area of ​​new infrastructure construction, has become a key direction for promoting the transformation and upgrading of the transportation industry. To promote the intelligent upgrading of transportation infrastructure and achieve unmanned management of key road sections and hubs, a cloud-edge collaborative architecture based on edge real-time perception processing and cloud global decision optimization has become the core technical path to solve the problems of high cost and slow response of manual management in scenarios such as highway entrances and exits and urban main roads. However, existing technical solutions still have many shortcomings: In traditional technologies, the quantitative integration logic of cloud-edge collaboration is insufficient. Existing analysis methods only constrain cloud-edge coordination through a single average traffic flow indicator, which cannot quantify the weight of vehicle model differences and vehicle following behavior on traffic efficiency in different scenarios. Moreover, the policy triggering thresholds are mostly fixed thresholds. When facing lane traffic control in a wide range of scenarios such as highway toll stations and park micro-entrances, the traffic efficiency, order and safety are low in unattended lane scenarios.

[0003] To address the aforementioned shortcomings, a technical solution is provided. Summary of the Invention

[0004] The purpose of this invention is to solve the problems of low traffic efficiency, order and safety in unattended lane scenarios, and to propose an optimization method for unattended lane traffic based on cloud-edge collaboration.

[0005] The objective of this invention can be achieved through the following technical solutions: The cloud-edge collaborative optimization method for unattended lane passage includes: S1. Traffic Parameter Perception: Based on edge perception units, basic parameters of lane traffic scenarios and individual vehicle characteristic parameters are obtained. Basic parameters include traffic flow density, average vehicle speed, vehicle type distribution, environmental visibility, and the original traffic efficiency baseline value of the lane. Individual vehicle characteristic parameters include vehicle identity features, vehicle behavior features, and vehicle status features. Multi-source heterogeneous data are preprocessed based on edge computing nodes to generate multi-dimensional traffic parameter snapshots. S2. Traffic Characteristics Analysis: Based on the multi-dimensional traffic parameter snapshots preprocessed by edge computing nodes, traffic flow is analyzed in a multi-dimensional hierarchical manner, including basic characteristic decomposition analysis and comprehensive traffic characteristics analysis. S3, Collaborative Strategy Generation: Dynamic thresholds generated by the cloud combining historical and real-time data are used for strategy triggering and condition judgment. Global strategies are generated based on the comprehensive index of traffic characteristics, combined with historical data from the cloud and real-time status of the edge. S4. Optimize instruction execution and effect verification: Edge computing nodes parse instruction packets and convert them into device-recognizable instructions; after each statistical period, multi-dimensional indicators are collected to verify the effect of the strategy.

[0006] Furthermore, the specific implementation process of the basic characteristic decomposition analysis includes: Traffic flow density, average vehicle speed, and timestamp information are extracted from multidimensional traffic parameter snapshots to construct a three-dimensional correlation matrix. The update interval of the three-dimensional correlation matrix is ​​used as the statistical period. The average vehicle flow is calculated based on the arithmetic mean of the average traffic flow density within the statistical period and the weighted average of the ratio of instantaneous speed to effective travel time of each vehicle. Based on the average vehicle flow and the lane's preset saturation flow benchmark, the current cycle is divided into time periods; based on multi-dimensional traffic parameter snapshots, the rate of change of flow between adjacent time periods is obtained; Based on the vehicle type distribution data, the number of vehicle types and the proportion of each type of vehicle in the total number of vehicles are combined with the equivalent conversion coefficient of each type of vehicle to obtain the comprehensive vehicle type influence coefficient; the equivalent conversion coefficient refers to the initial value of the vehicle type preset based on road type and lane function classification.

[0007] Furthermore, the specific implementation process of the basic characteristic decomposition analysis also includes: Based on vehicle behavior characteristics, the instantaneous speed of each following vehicle is extracted. Through formula Get the first The theoretical safe distance for a vehicle, among which, Indicates the first The instantaneous speed of the vehicle, Indicates the vehicle's maximum deceleration. Indicates the driver's reaction time; The safety distance compliance rate is obtained by comparing the actual vehicle headway with the theoretical safe distance. The percentage of smooth driving is obtained by comparing the vehicle's instantaneous acceleration with a preset stable acceleration range. A comprehensive interactive impact index was obtained based on the safe distance compliance rate and the proportion of smooth driving. Based on vehicle status characteristics, vehicles with surface temperatures exceeding a threshold are identified through infrared thermal imaging data, or vehicles with mechanical faults are identified through acoustic sensor feature matching and marked as abnormal vehicles. Through formula The abnormal interference index is obtained, where, This represents the total number of vehicles marked as abnormal within the statistical period. This represents the total number of all vehicles passing through within the statistical period. This indicates the average duration of a single interference.

[0008] Furthermore, the specific operational steps for the comprehensive analysis of traffic characteristics are as follows: The real-time average traffic flow is compared with the lane's preset saturation flow benchmark. At the same time, a comprehensive vehicle type influence coefficient is introduced for correction to obtain the effective traffic index. Based on the comprehensive interaction impact index, the traffic order coordination index is obtained by introducing the rate of change of traffic flow in adjacent time periods. A comprehensive interference index is obtained based on the abnormal interference index and the visibility of the actual lane environment. By obtaining the preprocessing latency of edge computing nodes and the historical data retrieval and decision response time of the cloud through edge sensing network units, and combining the effective index, the cloud-edge collaborative processing efficiency index is obtained. Based on the traffic effectiveness index, traffic order coordination index, comprehensive interference index, and cloud-edge collaborative processing efficiency index, a comprehensive traffic characteristic index is obtained after normalization and weighted formula.

[0009] Furthermore, the specific process for generating the dynamic threshold is as follows: Based on the cloud-based historical database, obtain the historical traffic characteristic comprehensive index sequence with the same spatiotemporal attributes as the current time; the spatiotemporal attributes include the same workday type, the same time period, and the same lane or area; select the traffic characteristic comprehensive index with the same spatiotemporal attributes within the past 3 months to construct a historical dataset; extract the traffic characteristic comprehensive index of several consecutive statistical cycles that have been completed within the current time period to construct a real-time data window; Sort the historical dataset from smallest to largest, and calculate the upper quartile and lower quartile as the initial upper threshold and initial lower threshold, respectively; The trend slope is calculated based on the comprehensive index of traffic characteristics of real-time window data and the total number of statistical periods. By introducing a correction coefficient to correct the trend slope, an adjustment factor is obtained. Combined with the standard deviation of the data in the real-time data window, the initial threshold is adjusted to obtain the dynamic upper limit threshold and the dynamic lower limit threshold.

[0010] Furthermore, the specific operational steps for triggering the strategy and determining the conditions are as follows: Regular trigger: Based on the real-time traffic characteristics comprehensive index and the dynamic upper and lower thresholds, the lane traffic status is determined, including: optimized status, maintained status and excellent status; Emergency Trigger: Emergency trigger conditions are monitored in parallel for abnormal changes in the traffic order coordination index and the comprehensive interference index. The system is triggered when any one of the conditions is met. Based on the traffic order coordination index and comprehensive interference index of the current statistical period, they are compared with the corresponding index of the previous statistical period. If the deviation value is greater than the preset threshold, the emergency triggering mechanism is triggered. When a regular or emergency trigger is detected, a cloud-based global policy is activated to optimize lane traffic flow.

[0011] Furthermore, the specific generation process of the global strategy is as follows: The cloud-based decision center parses the data packets uploaded by edge nodes and extracts key parameters, including the comprehensive index of traffic characteristics and multi-dimensional traffic parameter snapshots; at the same time, it retrieves historical data with the same spatiotemporal attributes. Taking the comprehensive traffic characteristic index as the core, a high-dimensional feature vector is constructed by integrating macro-dynamic characteristics of traffic flow, multi-vehicle type influence coefficient, vehicle interaction behavior influence and abnormal state interference, and then forming a dynamic traffic flow scenario set with all scenario vectors in the historical database. When identifying new traffic scenarios, the similarity between the current scenario and historical scenarios is calculated based on the multidimensional Euclidean distance formula; the R historical scenarios with the smallest similarity are selected as the candidate strategy set. For each candidate strategy set, the strategy type and strategy effect corresponding to the historical scenario are extracted. The strategy weight is calculated based on the strategy effect and the historical application frequency. Combined with the current scenario state, a multi-objective optimization algorithm is used to generate a global strategy, and simulation verification is performed. Once the verification is successful, the cloud-based decision center will package the optimization parameters into a standardized instruction package and distribute it to the edge computing nodes.

[0012] Furthermore, the specific implementation process of step S4 is as follows: Edge nodes perform security verification and parsing of command packets, extract control parameters, and convert them into device-recognizable commands, including signal control, lane function, and speed guidance commands. When a sudden accident is detected, the emergency control mode is activated, including sound and light alarms, adjusting timing to reserve emergency windows, pushing detour suggestions to surrounding vehicles, and uploading on-site video streams to the cloud. After each statistical period, traffic efficiency, traffic order, and safety indicators are collected and weighted to obtain a comprehensive performance score. The applicability of the strategy is determined based on the comparison between the comprehensive performance score and the preset threshold range.

[0013] Compared with the prior art, the beneficial effects of the present invention are: This invention achieves precise optimization of unattended lane passage through a cloud-edge collaborative architecture. It comprehensively collects multi-source heterogeneous data via edge sensing units and preprocesses it. Combined with multi-dimensional hierarchical analysis, it quantifies vehicle model differences, vehicle interactions, and the impact of abnormal interference. Dynamic thresholds and a dual-trigger mechanism adapt to different scenarios, avoiding policy lag or over-optimization. Cloud-based strategies are verified through digital twin simulation to ensure adaptability and effectiveness. After command execution, a closed loop is formed through comprehensive evaluation of multiple indicators, enabling dynamic policy iteration. An emergency mode provides rapid response to unforeseen circumstances, ensuring traffic safety. This significantly improves lane passage efficiency and order, reduces interference risks, adapts to complex traffic scenarios, and reduces manual management costs. Attached Figure Description

[0014] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0015] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. 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.

[0016] It should be understood that the terms “comprising” and “including” used in this disclosure and claims indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0017] It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure. As used in this disclosure and claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in this disclosure and claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations.

[0018] like Figure 1 As shown, the unattended lane passage optimization method based on cloud-edge collaboration includes passage parameter perception, passage characteristic analysis, collaborative strategy generation, optimization command execution, and effect verification.

[0019] S1. Traffic Parameter Perception: The basic parameters of the lane traffic scenario are acquired through the edge perception unit. The basic parameters of the lane traffic scenario include: traffic flow density, average vehicle speed, vehicle type distribution, environmental visibility, and the baseline value of the original lane traffic efficiency. The vehicle type distribution represents the proportion of each type of vehicle in the traffic flow, including small passenger cars, large passenger cars, small trucks, and large trucks. Simultaneously, the edge perception unit acquires individual characteristic parameters of vehicles in the lane. These individual characteristic parameters include: Vehicle identification features: The vehicle license plate information is obtained through a high-definition license plate recognition unit as the unique identifier of the vehicle. At the same time, the vehicle communicates with vehicles that support vehicle-road cooperation through a roadside communication unit to obtain an anonymous temporary ID provided by the corresponding vehicle terminal. Vehicle behavior characteristics: The video analysis unit continuously tracks the vehicle, extracts the motion trajectory, and obtains the vehicle's instantaneous speed and acceleration, motion direction angle, headway and distance between the vehicle and the vehicles in front and behind based on the trajectory data; Vehicle status characteristics: The infrared thermal imaging unit detects the surface temperature distribution of the vehicle and identifies abnormally high temperature points. The acoustic sensor array collects the sound characteristics of the vehicle as it passes by, which is used for auxiliary identification of the vehicle model or preliminary judgment of mechanical faults. The multi-source heterogeneous data acquired by the edge sensing unit are all transmitted to the local edge computing node. The edge node performs preprocessing operations on the multi-source heterogeneous data to form a unified, timestamp-aligned multidimensional access parameter snapshot. The preprocessing operations include time synchronization, coordinate alignment, invalid data filtering, and missing value imputation.

[0020] S2. Traffic Characteristics Analysis: Based on the multi-dimensional traffic parameter snapshots preprocessed from edge computing nodes, a multi-dimensional hierarchical analysis of traffic flow is performed. The specific analysis method is as follows: S201, Basic Characteristics Disassembly Analysis: Macroscopic dynamic characteristics of traffic flow: Traffic flow density, average vehicle speed, and timestamp information are extracted from multidimensional traffic parameter snapshots to construct a three-dimensional correlation matrix of density, speed, and time. The three-dimensional correlation matrix is ​​automatically updated with data points at fixed time intervals. Using the update time interval of the three-dimensional matrix as the statistical period, the average traffic flow density A and average vehicle speed B within the statistical period are extracted, normalized, and then entered into the formula. The average vehicle flow rate was calculated. ,in, This represents the arithmetic mean of the average traffic flow density over a statistical period. It represents the weighted average of the instantaneous speed and effective driving time of each vehicle within the same period; the effective driving time percentage refers to the ratio of the actual driving time of each vehicle within the statistical period to the total driving time of all vehicles within the same statistical period. Based on the average vehicle flow within the statistical period and combined with the lane's preset saturation flow benchmark, the current period is divided into three time periods: peak, off-peak, and low-peak. For example, when the average vehicle flow is greater than 80% of the lane's preset saturation flow benchmark, the current period is marked as a peak period; when the average vehicle flow is between 30% and 80% of the lane's preset saturation flow benchmark, it is an off-peak period; when the average vehicle flow is less than 30% of the lane's preset saturation flow benchmark, the current period is marked as a low-peak period. Simultaneously, based on the timestamp-aligned multi-dimensional traffic parameter snapshot, the rate of change of traffic between adjacent time periods is obtained. The smaller the absolute value of the rate of change, the smoother the transition between time periods. Multi-vehicle model impact coefficient: Based on vehicle type distribution data, combined with the preset equivalent conversion coefficients for various vehicle types, the equivalent conversion coefficients refer to the preset initial values ​​for vehicle types based on road type and lane function classification; For example, on highways: in the straight lane, the width is 1.0 for small passenger cars, 1.6 for large passenger cars, 1.3 for small trucks, and 2.2 for large trucks; the left-turn lane is increased by 0.1 based on the straight lane width. Urban roads: On straight-ahead lanes, the following vehicle sizes apply: 1.0 for small passenger vehicles, 1.5 for large passenger vehicles, 1.2 for small trucks, and 2.0 for large trucks. Through formula The comprehensive vehicle model impact coefficient is calculated, where, This indicates the total number of vehicle types. Indicates the first The percentage of different vehicle types in the total number of vehicles in a lane. Indicates the first Equivalent conversion factor for vehicle type; The higher the overall vehicle type influence coefficient, the greater the interference from vehicle type differences in the current lane; For example, the distribution of vehicle types in lanes during the current statistical period is as follows: Small passenger vehicles: , ; Large passenger buses: , 5; Small vans: , 2; Large trucks: , 2.0; Comprehensive vehicle model influence coefficient This indicates that the total interference intensity of the current traffic flow is 1.22 times that of pure small passenger cars; Impact of vehicle interaction behavior: Based on vehicle behavior characteristics, the instantaneous speed of each following vehicle is extracted. After normalization, the formula is entered. Calculation yields the first The theoretical safe distance for a vehicle, among which, Indicates the first The initial speed of the vehicle, This indicates the vehicle's maximum deceleration, specifically the maximum deceleration during emergency braking. This indicates the driver's reaction time, which is the time interval from when the driver notices a hazard to when they begin to brake. The safety distance compliance rate is obtained by comparing the actual headway distance with the theoretical safe distance and calculating the proportion of all vehicles whose actual headway distance is greater than the theoretical safe distance. Based on the instantaneous acceleration of the vehicle, and compared with the preset stable acceleration range, the proportion of stable driving is obtained by calculating the proportion of the total observation time when the acceleration data of all vehicles are in the preset stable acceleration range. The comprehensive interactive impact index is calculated based on the safe distance compliance rate and the proportion of stable driving, combined with a weighted formula. The higher the comprehensive interaction impact index, the greater the potential risks and efficiency losses caused by vehicle-to-vehicle interaction behavior; Abnormal state interference: Based on vehicle status characteristics, vehicles with surface temperatures exceeding a threshold are identified through infrared thermal imaging data, or vehicles with mechanical faults are identified based on acoustic sensor feature matching. Both types of vehicles are collectively referred to as abnormal vehicles. Through formula The abnormal interference index is calculated, where, This represents the total number of vehicles marked as abnormal within the statistical period. This represents the total number of all vehicles passing through within the statistical period. Indicates the average duration of a single interference; S202, Comprehensive Analysis of Traffic Characteristics: Real-time average vehicle flow The data is compared with the preset lane saturation flow benchmark, and a comprehensive vehicle type influence coefficient is introduced for correction, using the formula... The effective passage index is calculated. ;in, This indicates the preset saturation flow rate benchmark for the lane. This indicates the weighting factor for the overall vehicle model impact coefficient; Based on the comprehensive interaction impact index, the rate of change of traffic flow in adjacent time periods is introduced, and a weighted formula is used to obtain the traffic order coordination index. Based on the abnormal interference index and combined with the actual visibility of the lane environment collected by the edge perception unit, the normalized data is then input into the formula. The comprehensive interference index is calculated. ,in, Indicates actual environmental visibility. This represents the baseline value for lane design visibility under the same environmental conditions. Indicates the deviation of environmental visibility. , These represent the weighting factors for the environmental visibility deviation and the abnormal interference index, respectively. The preprocessing latency of edge computing nodes is obtained through edge sensing network units. This refers to the time interval from the point when the edge sensing unit collects raw data to the point when the edge node completes the traffic characteristic analysis and encapsulates the data packet to be uploaded; simultaneously, it obtains the time interval from when the edge node sends the data packet to when it receives the decision instruction from the cloud, thus obtaining the cloud historical data retrieval and decision response time. Combined with the prevailing effective index, through the formula The cloud-edge collaborative processing efficiency index was calculated. ; This indicates the maximum permissible processing delay for the current traffic scenario. This represents the actual improvement in the passability efficiency index, which is the improvement in the passability efficiency index relative to the state before the decision was made after the application of cloud-based decision-making. This represents the maximum increase in the historical effective index. Indicates time-sensitive components; These represent the weighting factors that influence the ratio of the actual increase in the timeliness component and the historical maximum increase in the effective index, respectively. Based on the traffic effectiveness index, traffic order coordination index, comprehensive interference index, and cloud-edge collaborative processing efficiency index, the comprehensive traffic characteristic index is calculated by normalizing and then combining it with a weighted formula.

[0021] S3, Cooperative Strategy Generation: Dynamic threshold calculation: Based on a cloud-based historical database, obtain a sequence of historical traffic characteristic comprehensive indexes that have the same spatiotemporal attributes as the current time. Spatiotemporal attributes include the same workday type, the same time period, and the same lane or area. Select the traffic characteristic comprehensive indexes of all statistical periods with the same spatiotemporal attributes within the past 3 months to construct a historical dataset. Meanwhile, from the real-time data stream uploaded by edge computing nodes, the comprehensive index of traffic characteristics for several consecutive statistical cycles completed within the current time period is extracted to construct a real-time data window; The comprehensive index of the passability characteristics of the historical dataset is sorted from smallest to largest, and the upper quartile and lower quartile are calculated as the initial upper thresholds. and initial lower threshold ; For example, the historical dataset for the past four Monday morning rush hours is: The historical dataset after sorting is ;but , ; Based on real-time window data from the past three months, using the formula The trend slope is calculated. , Indicates the first One statistical period, Indicates the first The comprehensive index of traffic characteristics corresponding to each statistical period. Indicates the total number of statistical periods; By introducing a preset correction coefficient k, the trend slope is corrected to obtain the adjustment factor. And based on the adjustment factor, combined with the data standard deviation of the real-time data window. The dynamic threshold is obtained by adjusting the initial threshold. Dynamic upper limit threshold: = Dynamic lower threshold: = in, This represents the fluctuation tolerance coefficient, an empirical parameter set manually based on business scenarios and experience, used to adjust the adaptability of the threshold under different traffic flow stability conditions; At the end of each complete time period, new data generated within that time period is automatically added to the historical database, and the initial threshold for subsequent time periods is recalculated. When it is detected that the comprehensive index of traffic characteristics deviates from the current threshold range continuously over multiple consecutive statistical periods, and the trend slope is not zero, the threshold recalculation process will be triggered in advance. Strategy triggering and condition judgment: Regular triggering: A comparison is made between a comprehensive index based on real-time traffic characteristics and dynamic upper and lower threshold values. When the comprehensive traffic characteristic index is lower than the dynamic lower limit threshold, it indicates that the current lane traffic status has deviated from the normal level, and the overall traffic efficiency, order or anti-interference ability tends to deteriorate. It is determined that the current strategy is no longer applicable, that is, the optimization state, triggering the complete strategy generation and optimization process. When the comprehensive traffic characteristic index is between the dynamic lower limit threshold and the dynamic upper limit threshold, it indicates that the current traffic status is within an acceptable normal range. In this case, the existing control strategy is maintained or slightly optimized, and continuous monitoring is carried out, i.e., the status is maintained. Light optimization refers to adjusting parameters. For example, based on the small changes in the average vehicle speed in the last few cycles, the green light duration is increased or decreased by a few seconds in the next signal cycle without initiating a global strategy replanning. When the comprehensive index of traffic characteristics is higher than the dynamic upper limit threshold, it indicates that the current traffic status is good or excellent, and the current control strategy is highly effective, i.e., excellent. In the excellent state, the existing strategy is maintained and lightweight optimization is paused. Emergency Trigger: Emergency triggering conditions are implemented by monitoring abnormal changes in the traffic order coordination index and the comprehensive interference index in parallel. The system is triggered when any one of the conditions is met. Triggered by the traffic order coordination index: The traffic order coordination index of the current statistical period is compared with the traffic order coordination index of the previous statistical period. If the deviation value is greater than the preset threshold, it is determined that the traffic flow order has dropped significantly; for example, a traffic accident, illegal vehicle cutting in, or other events that cause traffic flow blockage and seriously disrupt normal order have occurred. Triggering based on the comprehensive interference index: Based on the comprehensive interference index of the current statistical period, compare it with the comprehensive interference index of the previous statistical period. If the increase is greater than the preset threshold, it is determined that there is a momentary strong source of interference in the lane; for example, an abnormal vehicle newly identified by the edge perception unit or a sudden roadblock event reported by the roadside unit. When a regular or emergency trigger is detected, a cloud-based global policy is activated to optimize lane traffic flow. S302. Cloud-based Global Policy Generation: After receiving a policy generation request and related data from the edge node, the cloud decision center initiates the global policy generation process. The specific implementation process is as follows: The cloud-based decision center parses the data packets uploaded by edge nodes and extracts key parameters, including the comprehensive traffic characteristic index, traffic effectiveness index, traffic order coordination index, comprehensive interference index, cloud-edge collaborative processing efficiency index, and raw data from multi-dimensional traffic parameter snapshots. At the same time, the cloud-based decision center accesses the historical database to retrieve historical data with the same spatiotemporal attributes as the current scenario, including historical comprehensive traffic characteristic index sequences, historical strategy records, and effect evaluation data. Based on multi-source data, the cloud-based decision center takes the comprehensive traffic characteristic index as its core, integrates high-dimensional feature vectors including macro-dynamic characteristics of traffic flow, multi-vehicle type influence coefficients, vehicle interaction behavior influence, and abnormal state interference, and together with the scene vectors of all records in the historical database, it forms a dynamic traffic flow scene set. When a new real-time traffic scene is identified, the cloud-based decision center calculates the similarity between the current scene and historical scenes in the historical database based on the multidimensional Euclidean distance formula; and selects the R historical scenes with the smallest similarity as the candidate strategy set. For each candidate strategy set, extract the strategy type and strategy effect corresponding to historical scenarios. Based on the strategy effect and historical application frequency, use the formula... Calculate the policy weights ;in, This represents the improvement in the historical effectiveness index of the current strategy. This indicates the number of times the current strategy has been applied in the past; Based on the policy weights of the candidate policy set and combined with the current scenario status, the cloud decision center uses a multi-objective optimization algorithm to generate a global policy. The optimization objectives include maximizing the traffic efficiency index, minimizing the comprehensive interference index, and ensuring that the cloud-edge collaborative processing efficiency index is within a preset range. The optimization variables include traffic light timing parameters, lane control parameters, and vehicle guidance parameters. Before the strategy is issued, the cloud-based decision center simulates and verifies the generated strategy based on the digital twin platform, and then normalizes it before inputting it into the formula. The effect evaluation index is calculated; where S represents the predicted scenario state. Indicates the current scene state; If the average effect index and the comprehensive index of traffic characteristics both increase by more than the corresponding preset threshold, the strategy is deemed effective. If either parameter is less than the preset threshold, the optimization process is retried or the strategy is rolled back to the best historical strategy. Once the strategy is validated, the cloud-based decision center will encapsulate the optimization parameters into a standardized instruction package, including: Control commands, such as traffic light timing schemes and the status of variable lane signs; Guidance instructions, such as sending recommended speeds or route suggestions to vehicles via roadside communication units; Monitoring instructions, such as adjusting the sampling frequency or focusing area of ​​the edge sensing unit; The instruction packet is sent to the edge computing node through a secure communication protocol, and the policy details are recorded in the historical database at the same time; the policy details include the generation time, optimization parameters and prediction effect; The cloud-based decision center monitors the delay in issuing instructions, ensuring that transmission is completed within the constraints of the cloud-edge collaborative processing efficiency index.

[0022] S4. Optimize command execution and verify results: S401, Command Parsing and Device Collaboration: The edge nodes perform security verification and parsing on the received command packets, extract control parameters, and convert them into control commands that the device can recognize. Signal control instructions: Based on the optimized timing scheme, the signal cycle is dynamically adjusted through the traffic signal controller; for example, during peak hours, the green light duration for east-west traffic is adjusted from the base of 30 seconds to 35 seconds, and the remaining time is updated in real time through the countdown display. Lane function commands: Adjust the driving direction of the lane based on the variable lane indicator. For example, change the original left turn lane to a straight and left turn mixed lane during off-peak hours. The switching interval of the indicator sign is controlled within 5 seconds. Speed ​​guidance instructions: Recommended vehicle speeds are sent to connected vehicles based on roadside communication units. For example, in foggy weather conditions, the speed guidance value is adjusted from 60km / h to 40km / h, and visual prompts are simultaneously released through LED information boards. When the edge sensing unit detects a sudden incident, it immediately activates the emergency control mode: Accident warnings are activated by sound and light alarm devices, signal timing is automatically adjusted, passage windows are reserved for emergency rescue, detour suggestions are pushed to vehicles within a 200-meter radius through roadside units, and on-site video streams are simultaneously uploaded to the cloud decision-making center. S402, Effect Verification: At the end of each statistical period, the edge nodes collect the following metrics for real-time performance evaluation: Traffic efficiency indicators: actual traffic flow, average delay time, queue length; Traffic order indicators: coefficient of variation of headway and standard deviation of acceleration; Safety indicators: number of emergency braking incidents, number of trajectory conflict points; After normalizing the data of each indicator, a weighted fusion is performed based on the weighted fusion formula to obtain the comprehensive effect score; When the overall effect score is greater than the preset threshold range, the current strategy is deemed effective. The parameters, application scenario characteristics and effect data of the current strategy are packaged and uploaded to the cloud historical database as a successful case. When the overall effect score is within the preset threshold range, it is determined that the current strategy is basically effective, but there is room for optimization. This triggers the self-adjustment of lightweight parameters on the edge side, such as adjusting the green light duration by ±3 seconds in the next signal cycle, or fluctuating the recommended vehicle speed by ±5km / h, and continuously observing the effect in the following 1 to 3 cycles. When the overall performance score is lower than the preset threshold range, the current strategy is determined to be inapplicable. The edge node immediately sends a strategy failure alarm to the cloud decision center, along with a snapshot of the multi-dimensional access parameters, requesting the activation of the emergency triggering mechanism to regenerate the global strategy.

[0023] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A cloud-edge collaborative optimization method for unattended lane passage, characterized in that, include: S1. Traffic Parameter Perception: Based on the edge perception unit, the basic parameters of the lane traffic scene and the individual vehicle characteristic parameters are obtained. The basic parameters include traffic flow density, average vehicle speed, vehicle type distribution, environmental visibility and the original traffic efficiency benchmark value of the lane. The individual vehicle characteristic parameters include vehicle identity characteristics, vehicle behavior characteristics and vehicle status characteristics. Preprocessing of multi-source heterogeneous data based on edge computing nodes generates multi-dimensional access parameter snapshots; S2. Traffic Characteristics Analysis: Based on the multi-dimensional traffic parameter snapshots preprocessed by edge computing nodes, traffic flow is analyzed in a multi-dimensional hierarchical manner, including basic characteristic decomposition analysis and comprehensive traffic characteristics analysis. S3, Collaborative Strategy Generation: Dynamic thresholds generated by the cloud combining historical and real-time data are used for strategy triggering and condition judgment. Global strategies are generated based on the comprehensive index of traffic characteristics, combined with historical data from the cloud and real-time status of the edge. S4. Optimize instruction execution and effect verification: Edge computing nodes parse instruction packets and convert them into device-recognizable instructions; after each statistical period, multi-dimensional indicators are collected to verify the effect of the strategy.

2. The unattended lane access optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The specific implementation process of the basic characteristic decomposition analysis includes: Traffic flow density, average vehicle speed, and timestamp information are extracted from multidimensional traffic parameter snapshots to construct a three-dimensional correlation matrix. The update interval of the three-dimensional correlation matrix is ​​used as the statistical period. The average vehicle flow is calculated based on the arithmetic mean of the average traffic flow density within the statistical period and the weighted average of the ratio of instantaneous speed to effective travel time of each vehicle. Based on the average vehicle flow and the lane's preset saturation flow benchmark, the current cycle is divided into time periods; based on multi-dimensional traffic parameter snapshots, the rate of change of flow between adjacent time periods is obtained; Based on the vehicle type distribution data, the number of vehicle types and the proportion of each type of vehicle in the total number of vehicles are combined with the equivalent conversion coefficient of each type of vehicle to obtain the comprehensive vehicle type influence coefficient; the equivalent conversion coefficient refers to the initial value of the vehicle type preset based on road type and lane function classification.

3. The unattended lane access optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The specific implementation process of the basic characteristic decomposition analysis also includes: Based on vehicle behavior characteristics, the instantaneous speed of each following vehicle is extracted. Through formula Get the first The theoretical safe distance for a vehicle, among which, Indicates the first The instantaneous speed of the vehicle, Indicates the vehicle's maximum deceleration. Indicates the driver's reaction time; The safety distance compliance rate is obtained by comparing the actual vehicle headway with the theoretical safe distance. The percentage of smooth driving is obtained by comparing the vehicle's instantaneous acceleration with a preset stable acceleration range. A comprehensive interactive impact index was obtained based on the safe distance compliance rate and the proportion of smooth driving. Based on vehicle status characteristics, vehicles with surface temperatures exceeding a threshold are identified through infrared thermal imaging data, or vehicles with mechanical faults are identified through acoustic sensor feature matching and marked as abnormal vehicles. Through formula The abnormal interference index is obtained, where, This represents the total number of vehicles marked as abnormal within the statistical period. This represents the total number of all vehicles passing through within the statistical period. This indicates the average duration of a single interference.

4. The unattended lane access optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The specific operational steps for the comprehensive analysis of traffic characteristics are as follows: The real-time average traffic flow is compared with the lane's preset saturation flow benchmark. At the same time, a comprehensive vehicle type influence coefficient is introduced for correction to obtain the effective traffic index. Based on the comprehensive interaction impact index, the traffic order coordination index is obtained by introducing the rate of change of traffic flow in adjacent time periods. A comprehensive interference index is obtained based on the abnormal interference index and the actual visibility of the lane environment. By obtaining the preprocessing latency of edge computing nodes and the historical data retrieval and decision response time of the cloud through edge sensing network units, and combining the effective index, the cloud-edge collaborative processing efficiency index is obtained. Based on the traffic effectiveness index, traffic order coordination index, comprehensive interference index, and cloud-edge collaborative processing efficiency index, a comprehensive traffic characteristic index is obtained after normalization and weighted formula.

5. The unattended lane access optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The specific process for generating the dynamic threshold is as follows: Based on a cloud-based historical database, obtain a comprehensive index sequence of historical traffic characteristics with the same spatiotemporal attributes as the current time; spatiotemporal attributes include the same workday type, the same time period, and the same lane or area. A historical dataset was constructed by selecting comprehensive traffic characteristic indices with the same spatiotemporal attributes within the past three months; and a real-time data window was constructed by extracting comprehensive traffic characteristic indices from several consecutive statistical cycles completed within the current time period. Sort the historical dataset from smallest to largest, and calculate the upper quartile and lower quartile as the initial upper threshold and initial lower threshold, respectively; The trend slope is calculated based on the comprehensive index of traffic characteristics of real-time window data and the total number of statistical periods. By introducing a correction coefficient to correct the trend slope, an adjustment factor is obtained. Combined with the standard deviation of the data in the real-time data window, the initial threshold is adjusted to obtain the dynamic upper limit threshold and the dynamic lower limit threshold.

6. The unattended lane access optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The specific steps for triggering the strategy and determining the conditions are as follows: Regular triggering: Based on a comparison of the comprehensive index of real-time passage characteristics with dynamic upper and lower thresholds; Determine the lane's traffic status, including: optimized status, maintained status, and excellent status; Emergency Trigger: Emergency trigger conditions are monitored in parallel for abnormal changes in the traffic order coordination index and the comprehensive interference index. The system is triggered when any one of the conditions is met. Based on the traffic order coordination index and comprehensive interference index of the current statistical period, they are compared with the corresponding index of the previous statistical period. If the deviation value is greater than the preset threshold, the emergency triggering mechanism is triggered. When a regular or emergency trigger is detected, a cloud-based global policy is activated to optimize lane traffic flow.

7. The unattended lane access optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The specific generation process of the global strategy is as follows: The cloud-based decision center parses the data packets uploaded by edge nodes and extracts key parameters, including the comprehensive index of traffic characteristics and multi-dimensional traffic parameter snapshots; at the same time, it retrieves historical data with the same spatiotemporal attributes. Taking the comprehensive traffic characteristic index as the core, a high-dimensional feature vector is constructed by integrating macro-dynamic characteristics of traffic flow, multi-vehicle type influence coefficient, vehicle interaction behavior influence and abnormal state interference, and then forming a dynamic traffic flow scenario set with all scenario vectors in the historical database. When identifying new traffic scenarios, the similarity between the current scenario and historical scenarios is calculated based on the multidimensional Euclidean distance formula; the R historical scenarios with the smallest similarity are selected as the candidate strategy set. For each candidate strategy set, the strategy type and strategy effect corresponding to the historical scenario are extracted. The strategy weight is calculated based on the strategy effect and the historical application frequency. Combined with the current scenario state, a multi-objective optimization algorithm is used to generate a global strategy, and simulation verification is performed. Once the verification is successful, the cloud-based decision center will package the optimization parameters into a standardized instruction package and distribute it to the edge computing nodes.

8. The unattended lane access optimization method based on cloud-edge collaboration according to claim 1, characterized in that, The specific implementation process of step S4 is as follows: Edge nodes perform security verification and parsing of command packets, extract control parameters, and convert them into device-recognizable commands, including signal control, lane function, and speed guidance commands. When a sudden accident is detected, the emergency control mode is activated, including sound and light alarms, adjusting timing to reserve emergency windows, pushing detour suggestions to surrounding vehicles, and uploading on-site video streams to the cloud; after each statistical cycle, traffic efficiency, traffic order, and safety indicators are collected and weighted to obtain a comprehensive performance score. The applicability of the strategy is determined by comparing the comprehensive effect score with the preset threshold range.

Images