A Smart City Traffic Optimization Method and System Based on 5G Network

Abstract

This invention discloses a smart city traffic optimization method and system based on 5G networks, belonging to the field of smart city traffic management technology. It addresses the issues of information consistency and conflict resolution in urban traffic flow control and shared parking space guidance. The method includes uploading traffic flow and signal phase status, combining parking space occupancy status and vehicle owner location information for path guidance and parking space recommendation, and identifying and managing parking space status conflicts across multiple heterogeneous platforms. It constructs parking space status behavior sequences, heterogeneous contention indices, state disturbance propagation graphs, and conflict score clusters, performing state masking and path correction for high-conflict parking spaces. Simultaneously, it collects vehicle owner and vehicle behavior data to complete parking space usage records and settlements, identify and process illegal parking, and constructs a shared parking space reservation permission management mechanism. This method is applicable to optimization and guidance under dynamic urban traffic control and multi-platform access to parking space resources.

Description

Technical Field

[0001] This invention relates to the field of smart city traffic management technology, and more specifically, to a smart city traffic optimization method and system based on a 5G network. Background Technology

[0002] In the context of current smart city development, urban traffic optimization has gradually evolved from static strategies to real-time dynamic scheduling. Existing systems generally collect traffic flow and traffic signal status data by deploying sensors and combine this with navigation platforms for route recommendations, but they mainly suffer from three problems:

[0003] First, discrepancies exist in the collection cycles, transmission delays, and data consistency of traffic flow and signal status, causing a mismatch between the prediction model and the real-time status, affecting the accuracy of signal timing strategies. Second, the multi-source heterogeneous access of shared parking platforms has not yet established a unified information index structure. The lack of unified standards among platforms regarding status marking, reporting intervals, and permission adjustment rules leads to issues such as permission mismatch and status drift during parking status guidance pushes. Third, the subsequent responsibility for parking conflict guidance is unclear, making it impossible to conduct data analysis and strategy tracing of the root causes of guidance failures, and lacking a mechanism for reconstructing multi-platform behavioral models.

[0004] Current solutions attempt to establish congestion prediction mechanisms using historical data and predictive models, or introduce berth popularity indices to optimize recommendation results. However, they fail to deeply analyze the formation mechanism of induced conflicts, especially lacking systematic analysis and behavioral pattern extraction of the lag in switching status labels across multiple platforms. They are unable to perform fine-grained status masking and path correction for high-contested areas, thus affecting the scheduling efficiency of berth resources and user experience.

[0005] To address the above problems, this invention proposes a solution. Summary of the Invention

[0006] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a smart city traffic optimization method and system based on 5G network to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] In a preferred embodiment, it includes:

[0009] Collect traffic flow and signal phase status data, upload and cache them via 5G network, and create a dataset with device ID, timestamp and geographic location index;

[0010] The uploaded traffic data is cleaned, features are extracted and modeled to generate traffic flow prediction results and traffic status labels, and congested intersections are identified and signal timing adjustments are performed.

[0011] Collect parking space occupancy status and establish a parking space status pool. Combine the driver's location with the parking space attributes to perform route calculation, generate parking space recommendations and push them to the terminal, and at the same time release guidance information.

[0012] Retrieve historical events related to induced failures from the cloud berth status pool, extract permission status mismatch samples by comparing the induced generation time with the platform permission response time, classify and analyze the behavioral characteristics of each platform in terms of status label switching, reporting interval and authorization adjustment cycle, construct a drift coupling curve reflecting the response lag law, and derive the minimum observation time window based on the overlapping interval of multi-platform status fluctuations.

[0013] Within the minimum observation time window, the state upload behavior sequence of each berth is constructed, multi-dimensional indicators are extracted to generate heterogeneous contention index, and then the temporal coupling relationship between berths is analyzed. A state disturbance propagation map and an induced conflict base sub-map are generated. Through spatial mapping and multi-source indicator superposition, conflict-induced potential partitions are identified. Finally, berths are clustered according to conflict scores to form multiple state contention sub-clusters.

[0014] Based on the berth conflict score and the division of state contention sub-clusters, a threshold for heterogeneous contention index is set. Berths with contention indices exceeding the threshold and belonging to high conflict sub-clusters are subject to state masking or path penalty score correction, and their participation weight in navigation path planning is adjusted.

[0015] Within each conflict-inducing potential partition, a cluster of induced task deviations, a set of platform rewriting hypotheses and hypothesis drift paths are constructed. Parallel simulation analysis is conducted. By comparing the induced deviations under each hypothesis path, the path model with the smallest deviation is selected. Based on the state change information and induced response records of the optimal drift path, the responsibility ratio of the main control platform is backfitted to generate suggestions for adjusting the state upload and permission synchronization strategies of the main control platform.

[0016] For high-conflict parking spaces, implement status masking and path correction, conduct induced deviation simulation and platform responsibility fitting, collect vehicle owner entry and exit records to complete settlement and illegal parking processing, and establish a shared parking space reservation and access management mechanism.

[0017] In a preferred embodiment, traffic flow sensors and traffic signal monitoring equipment are deployed to collect the number of vehicles and the phase status of the signals. The data on the number of vehicles and the phase status of the signals are then uploaded at high speed, forwarded at the edge, and classified and cached via a 5G network. At the same time, a structured dataset indexed by device number, timestamp, and geographic location is established for format verification and anomaly isolation.

[0018] In a preferred embodiment, during the cloud data processing stage, traffic perception data is cleaned, features are extracted, and classification modeling is performed to generate traffic flow prediction results and traffic status labels for each path. Based on the prediction values, intersections that are about to become congested are identified and included in the signal optimization objects. Traffic efficiency is improved by adjusting the green light duration.

[0019] In a preferred embodiment, the occupancy status of each parking space is collected and reported, a parking space status pool is constructed and a multi-dimensional index is established, and path calculation and accessibility judgment are performed based on the driver's current location and parking space attributes to generate parking space recommendation results and push them to the user terminal. At the same time, the remaining parking space information in the area is published through the guidance screen.

[0020] In a preferred embodiment, vehicle owner and vehicle information are collected and linked, vehicle entry and exit times are recorded to generate parking space usage records and complete settlement, illegal parking behavior is identified and reminders are pushed, and if there is no response, an illegal parking record is generated and the system is sent to law enforcement dispatch, thus building a shared parking space platform that supports reservation and usage permission management.

[0021] In a preferred embodiment, it includes: a traffic flow acquisition and data structuring module, a cloud analysis and signal dispatching module, a berth status recommendation and guidance release module, and a berth conflict identification and comprehensive management module, with signal connections between the modules;

[0022] The traffic flow acquisition and data structuring module is mainly used to collect traffic flow and signal phase status data, upload and cache them through the 5G network, and build a dataset with device number, timestamp and geographic location index;

[0023] The cloud-based analytics and signal scheduling module is mainly used to clean, extract features, and model the uploaded traffic data, generate traffic flow prediction results and traffic status labels, identify congested intersections, and perform signal timing adjustments.

[0024] The parking space status recommendation and guidance release module is mainly used to collect parking space occupancy status and establish a parking space status pool. It combines the driver's location and parking space attributes to perform route calculation, generate parking space recommendations and push them to the terminal, and release guidance information at the same time.

[0025] The berth conflict identification and comprehensive management module is mainly used to retrieve historical events related to induced failures in the cloud berth status pool. By comparing the induced generation time with the platform permission response time, it extracts permission status mismatch samples, classifies and analyzes the behavioral characteristics of each platform in terms of status label switching, reporting interval and authorization adjustment cycle, constructs a drift coupling curve that reflects the response lag law, and derives the minimum observation time window based on the overlapping interval of multi-platform status fluctuations.

[0026] Within the minimum observation time window, the state upload behavior sequence of each berth is constructed, multi-dimensional indicators are extracted to generate heterogeneous contention index, and then the temporal coupling relationship between berths is analyzed. A state disturbance propagation map and an induced conflict base sub-map are generated. Through spatial mapping and multi-source indicator superposition, conflict-induced potential partitions are identified. Finally, berths are clustered according to conflict scores to form multiple state contention sub-clusters.

[0027] Based on the berth conflict score and the division of state contention sub-clusters, a threshold for heterogeneous contention index is set. Berths with contention indices exceeding the threshold and belonging to high conflict sub-clusters are subject to state masking or path penalty score correction, and their participation weight in navigation path planning is adjusted.

[0028] Within each conflict-induced potential partition, a cluster of induced task deviations, a set of platform rewriting hypotheses and hypothesis drift paths are constructed. Parallel simulation analysis is conducted. By comparing the induced deviations under each hypothesis path, the path model with the smallest deviation is selected. Based on the state change information and induced response records of the optimal drift path, the responsibility ratio of the main control platform is backfitted to generate suggestions for adjusting the state upload and permission synchronization strategies of the main control platform.

[0029] For high-conflict parking spaces, implement status masking and path correction, conduct induced deviation simulation and platform responsibility fitting, collect vehicle owner entry and exit records to complete settlement and illegal parking processing, and establish a shared parking space reservation and access management mechanism.

[0030] The technical effects and advantages of the smart city traffic optimization method and system based on 5G network of this invention are as follows:

[0031] This invention achieves unified caching and format verification of high-frequency sensing data by constructing a structured index system for traffic flow and signal status data. Combined with a cloud-based predictive modeling mechanism, it performs advance identification and timing strategy scheduling for intersections about to become congested. For parking guidance, a parking space status pool and path reachability judgment mechanism are established, and accurate recommendations and guidance information dissemination are achieved by combining user location. Based on this, a mechanism for identifying guidance failure behavior samples is proposed, and a drift coupling curve and minimum observation time window are constructed based on state switching lag characteristics to effectively restore the platform response delay mode. A heterogeneous contention index is generated based on the state upload behavior sequence to analyze the conflict structure and guidance propagation path between parking spaces. Parking spaces are clustered based on conflict scores, and state masking and path weight adjustment are implemented in the conflict partitions to achieve dynamic guidance control of parking resources. Finally, parking space usage record generation and settlement are linked by combining driver behavior data, and a shared parking space reservation and permission management mechanism is constructed to complete full-cycle optimization control. Attached Figure Description

[0032] Figure 1 This is a timing diagram of a smart city traffic optimization method and system based on a 5G network according to the present invention.

[0033] Figure 2 This is a schematic diagram of a smart city traffic optimization method and system module based on a 5G network according to the present invention. Detailed Implementation

[0034] 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.

[0035] Example: This invention discloses a smart city traffic optimization method based on a 5G network, such as... Figure 1 As shown, it includes:

[0036] Collect traffic flow and signal phase status data, upload and cache them via 5G network, and create a dataset with device ID, timestamp and geographic location index;

[0037] The uploaded traffic data is cleaned, features are extracted and modeled to generate traffic flow prediction results and traffic status labels, and congested intersections are identified and signal timing adjustments are performed.

[0038] Collect parking space occupancy status and establish a parking space status pool. Combine the driver's location with the parking space attributes to perform route calculation, generate parking space recommendations and push them to the terminal, and at the same time release guidance information.

[0039] Construct a parking space status behavior sequence and heterogeneous contention index, identify potential conflict areas and classify status contention sub-clusters, perform status masking and path correction for high-conflict parking spaces, conduct induced deviation simulation and platform responsibility fitting, collect vehicle owner entry and exit records to complete settlement and illegal parking processing, and establish a shared parking space reservation and access control mechanism.

[0040] In dynamic traffic operation management, the primary goal is to achieve continuous perception of road traffic conditions, relying on next-generation low-latency, high-speed 5G communication technologies to complete the task of monitoring traffic flow across the entire area. During this process, by deploying network-enabled sensing devices and constructing data channels with edge access capabilities, high-frequency perception and real-time reporting of the operational status of key traffic nodes are achieved.

[0041] Specifically, traffic flow sensors and traffic signal monitoring equipment are deployed first at transportation hubs, road intersections, and main road sections, with the principle of covering major traffic flow paths. The traffic flow sensors employ a combination of microwave sensing devices with dedicated traffic detection protocols and lidar to collect the number of vehicles passing through a cross-section in real time. The traffic signal monitoring equipment obtains the current signal phase status by connecting to the traffic controller's output port, along with a time sequence identifier within the traffic light cycle. All network-enabled sensing devices have their internal firmware control program setting the collection frequency and timestamp recording rules. The collection frequency is preset and matched according to road grade and traffic pressure level, and the timestamp information is generated by a local high-precision clock chip and written to each data record, ensuring a consistent and traceable time reference.

[0042] Furthermore, to achieve high-speed transmission and low-latency uploading, each network-enabled sensing device wirelessly accesses the nearest 5G base station node via its integrated 5G communication terminal. The access process follows the low-latency uplink scheduling mechanism defined in the 5G NR protocol suite, and device identification binding and data structure encapsulation are completed at the physical channel layer. The data collected by the device is automatically encoded into timestamped data frames, retaining its original structure while including the source device ID and location identifier, and uploaded in real time via the 5G channel.

[0043] After the upload is complete, the 5G base station pushes the data to a high-performance edge server deployed at the edge location through preset edge forwarding rules.

[0044] In this processing node, automatic classification is first performed based on the data type field carried in the data frame, and traffic flow quantity data, signal phase data and time series data are stored in different logical buffer queues respectively.

[0045] Subsequently, the data partitioning structure initialization program is called with data type as the index dimension to build sub-datasets for different purposes, and a multi-field index table structure is generated in each sub-dataset. The index fields include: device number, collection timestamp, geographic location information and device category tag.

[0046] It should be noted that the above dataset construction process employs hash mapping combined with a B+ tree index structure to optimize high-frequency access paths, ensuring that subsequent modeling algorithms can retrieve the required historical time-series records within milliseconds. Simultaneously, to ensure data structure consistency, format validation logic is executed during each data write process, performing real-time checks on field completeness and timestamp continuity. If missing fields or time misalignments are detected, an anomaly flag is triggered, and abnormal data is stripped and written to an asynchronous isolation buffer to prevent it from affecting the integrity of the official dataset.

[0047] During the cloud-based data processing phase, the data cleaning program is first invoked to preprocess the sensed data. This program takes single-path traffic flow data as input, sets a unique timestamp index as the discrimination criterion, and sequentially removes duplicate records, strips out abnormal samples, and fills in missing value gaps. The logic for filling in missing data is based on a combination of time series moving average and nearest neighbor interpolation. After completion, an integrity check is immediately performed to ensure the continuity of the basic data for subsequent statistical analysis.

[0048] The cleaned data stream is grouped according to the path number, and a sliding window of the past n sampling points is extracted in each group to calculate the average traffic flow, peak traffic flow and minimum traffic flow in the corresponding time period. The above three statistical indicators are used as the basic feature quantities of each path.

[0049] Next, the above statistical features are combined with the signal phase state corresponding to the current period to construct a multidimensional input vector.

[0050] In this combined structure, the signal phase state is represented in encoded form as the current state and duration of the traffic lights (red, green, yellow) at each intersection. All vectors are arranged in time step order to form a training sample sequence. Furthermore, through standardization, each feature is mapped to a uniform numerical range, ensuring a balanced numerical distribution of the model's input features.

[0051] Subsequently, the LSTM modeling channel is invoked to construct a training set using the input sequence, and the model is iteratively trained based on the error gradient propagation over time. The Adam optimizer is used during the training process to control the convergence rate. After the model training is complete, the predicted traffic flow value for the next time step is directly output and recorded in the prediction result buffer with the intersection number and timestamp.

[0052] To further identify typical traffic flow patterns during different road operating periods, the clustering channel is used to classify the feature results output from the modeling. Specifically, the K-means clustering algorithm is used, with the average traffic volume and signal duration of each path within a certain time period as two standardized indicators to construct the feature vector input for clustering.

[0053] Simultaneously, during the clustering operation, cluster centers are initialized and multiple iterations are performed to classify samples with the objective function of minimizing Euclidean distance until the cluster centers stabilize. Then, the final output cluster labels are defined as three state types: traffic peak, off-peak, and normal. These labels are then mapped back to the original path-time index to form a state-labeled dataset, providing conditional branching basis for matching traffic control strategies in different regions and time periods.

[0054] Furthermore, after completing the operation status identification, the predicted traffic flow is used as the input source, and a preset first threshold for traffic flow is called for comparison and judgment. If the predicted value exceeds the threshold and the corresponding path is in the marked interval of the cluster label as traffic peak state, the intersection is automatically marked as "about to be congested" and its path number is written into the optimization target queue as the strategy application point for subsequent signal adjustment.

[0055] For road sections with identified congestion trends, the green light duration allocation ratio is dynamically adjusted based on the premise that the total duration of red, green and yellow lights in the current cycle remains unchanged. Specifically, more green light time is allocated to the direction of travel with higher LSTM prediction values, thereby increasing the effective traffic flow per unit time.

[0056] The adjustment process introduces an upper limit constraint on the green light duration to prevent over-allocation of green signals from causing congestion to spread to adjacent intersections, while ensuring that the safe interval between signal phases meets the standard requirements for minimum delay between adjacent phases in the traffic signal control specifications.

[0057] In the allocation of static traffic resources, geomagnetic sensors, infrared beam sensors, and video stream analysis terminals are first deployed at the entrances and exits of each parking lot and at roadside public parking spaces.

[0058] Among them, the geomagnetic sensor determines whether a vehicle has entered a parking space by detecting changes in the geomagnetic field; the infrared beam detector determines the vehicle's entry and exit status by detecting whether the infrared beam is blocked; and the video stream analysis terminal uses an embedded image recognition algorithm to generate a continuous video stream input at the deployment location, extracting the vehicle's outline and edge features to determine whether the parking space is occupied. All sensing devices are pre-integrated with 5G communication terminals and automatically connect to the corresponding 5G access point in the area through set wireless configuration parameters, enabling millisecond-level uploading of vehicle entry and exit events.

[0059] The raw data uploaded by the sensing device includes event type, timestamp, berth number, and physical location coordinates. After being transmitted via a 5G channel, it is written into the cloud berth status pool in real time.

[0060] In the berth status pool, a multidimensional data index is constructed based on the berth number and geographical location information. The data index structure establishes a composite primary key through three fields: berth ID, administrative region code, and GPS location. Status update time and data source marker fields are added to each berth status record to ensure timeliness and traceability.

[0061] Subsequently, to implement the parking space navigation recommendation function, the city navigation engine is invoked and the vehicle owner's current location data is acquired simultaneously. The vehicle owner's location data is obtained through mobile device authorization, relying on GPS coordinates output by the user's terminal positioning function. The navigation engine uses the current vehicle location as the path starting point, selects all parking spaces marked "idle" in the parking space status pool as candidate destinations, and calls the path planning model to calculate the path time from the current location to each candidate parking space. Simultaneously, the path planning model, based on the A* path algorithm, integrates real-time traffic data and traffic restriction rules to output the shortest estimated travel time for each path, and sorts them according to this time value.

[0062] Furthermore, after the candidate paths are sorted, the berth attribute parsing logic is invoked to perform reachability checks on each candidate berth. The reachability check logic includes determining whether the berth is in a restricted area, whether specific access permissions are required, and whether the current user has the necessary access rights.

[0063] After the verification is passed, the berth charging strategy parsing function is called to parse the billing rules of the berth in the current time period and combine the path time, accessibility results and expected cost to form a joint score. Finally, the berth with the highest comprehensive score is selected as the recommended target.

[0064] Specific access permissions include parking spaces within organizations and enclosed parking spaces within residential communities; billing rules include time-based billing, time-limited pricing, and free access for the first hour.

[0065] Subsequently, after the target parking space is calculated and confirmed, the parking space number, route guidance data, and parking space status details are pushed to the user terminal via the navigation API interface. If the user terminal is an in-vehicle navigation system, it receives the data via the CAN bus interface and displays the navigation route in a graphical interface; if it is a mobile device, it uses a mobile app to call the map SDK to display the route and mark the target parking space information, allowing users to view and confirm in real time.

[0066] Meanwhile, to assist vehicle users who have not yet entered navigation mode in obtaining information on available parking spaces within the area, electronic parking guidance screens are installed along major urban traffic arteries. The data required by the guidance screens comes from the remaining parking space data in the parking space status pool. The system divides areas according to administrative divisions or functional blocks, summarizes the number of parking spaces marked as vacant in each area, and forms a data structure of "area number - number of remaining parking spaces", which is then refreshed to the guidance screens periodically through an information publishing program.

[0067] It should be noted that in scenarios where multiple entities access shared parking space resources, there is heterogeneous reporting of parking space status information from various sources, such as government agencies, commercial properties, and community management platforms. There are no unified standards among parking space managers regarding data collection frequency, permission status marking, reservation strategies, and status feedback rules. Significant differences exist, particularly in data upload intervals, status synchronization triggering mechanisms, and authorization policy adjustment cycles. Current technologies collect parking space status data from various sensors and upload it to a cloud-based status pool via 5G networks. A multi-dimensional index is then created based on parking space number and geographic location information for navigation route planning. However, there is no unified resolution and integration system for data consistency, status drift history, and differences in management permissions. This leads to situations where the same parking space appears available in the navigation engine during guidance screen pushes or navigation recommendations, but becomes unusable before the user arrives due to the platform's failure to promptly report permission changes. Navigation route planning guides users to areas with frequent status anomalies, causing vehicles to frequently detour and compete for parking spaces. Furthermore, some high-permission parking spaces are consistently absent from the guidance system, resulting in reduced resource utilization. Therefore, in this embodiment, the system first retrieves all historical events in the cloud-based berth status pool corresponding to the 5G network that show a conflict between the actual landing of the induced path and the final availability of the berth within the past T consecutive sampling periods.

[0068] The set of conflict events is generated based on the matching logic of the berth status update record and the guidance recommendation record. It is also used to form a three-element index structure with berth number, timestamp and platform identifier to locate the data source and response process corresponding to each guidance task.

[0069] After obtaining the complete set of conflict events, the platform status tag, induction publishing platform identifier, permission verification result, and berth platform definition associated with each event at the time of induction generation were extracted. The induction publishing platform identifier and platform status tag are derived from the induction task generation log, with log fields including platform ID, induction publishing time, target berth number, and platform record status before publishing. The permission verification result is obtained by extracting fields from the platform response result log structure, specifically including the permission request action response code, berth availability flag, and call return status value. All log data has complete time-series identification and is synchronously recorded to a unified time base for subsequent error measurement.

[0070] Furthermore, after obtaining all conflict sample data, by comparing the timestamps of the induced task generation time and the permission status response records, the set of all samples where the induced task failed due to the main control platform's failure to upload permission changes in a timely manner is identified, forming a permission status mismatch sample set. For each event in this permission status mismatch sample set, the time offset between the induced task generation time and the latest update time of the corresponding permission status is further located, and a permission adjustment delay mapping table is constructed to reflect the specific offset behavior of the platform's response lag.

[0071] Subsequently, the aforementioned time offset samples were categorized and grouped according to their respective platforms, and the following three indicators were extracted for each platform:

[0072] The platform status label switching amplitude per unit time is defined as the ratio of the number of consecutive status label change events to the duration of the time.

[0073] The dispersion of the status reporting interval sequence is defined as the standard deviation of the upload interval time within the last T periods;

[0074] The authorization policy adjustment cycle is defined as the minimum cycle length for policy changes within the platform's permission control logic.

[0075] Next, stability, responsiveness and structural weights are assigned to these three indicators respectively, and a weighted integral expression is constructed. With time as the horizontal axis and the probability of induced response mismatch as the vertical axis, the platform drift coupling curve is modeled under the exponential weighted moving average rule.

[0076] In this embodiment, the calculation expression for the platform drift coupling curve is: Dp(t)=Σ{k=0}^{T}λ×(1-λ)^k×δp(tk);

[0077] Where Dp(t) represents the state response drift intensity of platform p at time t;

[0078] δp(tk) is a binary index (1 for mismatch, 0 for consistency) indicating whether a mismatch of permission status occurs at time tk.

[0079] λ is the exponential decay factor, which is either the system default configuration or calibrated using historical samples.

[0080] Each platform drift coupling curve represents the probabilistic evolution trajectory of the platform's induced response mismatch due to delayed permission response within any historical time period. After modeling, the starting time point of the induced task corresponding to the first induced failure event in the platform's drift coupling curve is extracted, and the platform's state record sequence is traced back to mark the continuous interval with the most concentrated state fluctuations before that time point. Simultaneously, across all platforms, this most concentrated state fluctuation segment is cross-compared, and the minimum time segment length with overlap between multiple platforms is extracted, defined as the minimum observation time window.

[0081] The derivation formula for the minimum observation time window Tmin is as follows:

[0082] Tmin=max{Δti|i∈[1,N]}, where Δti represents the maximum change interval of traffic flow in the same lane per unit time within the i-th observation period.

[0083] By sampling all Δti sequences in historical data and selecting the maximum value as the lower limit of the window, we can ensure that the model input can cover at least one complete cycle of change, thus avoiding the problems of data sparsity or pattern truncation during model training.

[0084] It should be noted that the generation of the minimum observation time window depends on the dynamic linkage behavior of multiple platforms with inconsistent access control boundaries, different state label granularities, different upload protocol response rules, and synchronous participation in the induction triggering mechanism. Therefore, it is only meaningful under the structural condition that multiple heterogeneous platforms jointly participate in berth guidance and navigation but without unified state constraint rules. Once it is removed from this type of guidance system architecture, it will be impossible to generate the multi-platform drift coupling curve and extract the offset closure interval, thus the time scale itself loses its definability.

[0085] Next, the minimum observation time window length is defined as N1, and a sliding time window of length N1 is set as a unified time-domain benchmark for the historical state behavior analysis of each berth. Based on this, taking each berth as a unit, all state upload records within the past N1 time length are retrieved, and a behavior sequence matrix is ​​constructed according to the upload behavior in chronological order to characterize the dynamic response trajectory and heterogeneous platform behavior characteristics of the berth within this window.

[0086] In this behavior sequence matrix, each upload behavior consists of a quintuple, including the source platform identifier, upload frequency, platform status label at the time of generation, permission change marker, and call response result.

[0087] Specifically, the source platform identifier is used to indicate which berth management platform reported the data, the upload frequency is obtained by counting the number of uploads per unit time, the status label is extracted from the status field at the time of platform call in the induction task log, the permission change marker is calculated by comparing the change trend of the permission status field in continuous records, and the call response result is read from the response field when the request is initiated by the induction publishing platform. All five fields have a timestamp index structure and are filled into the matrix in order.

[0088] After constructing the behavior sequence matrix, the heterogeneous contention index for each berth is further calculated to characterize the degree of competition and conflict potential among the uploaded behaviors of each platform within the N1 window. Specifically, the heterogeneous contention index is calculated by fusing the following three quantitative factors:

[0089] The state stability factor is defined as the reciprocal of the number of state marker switching times within the N1 time window, and is used to reflect the stability of the berth's state during that period.

[0090] The lag delay coefficient of the main control platform is defined as the matching rate between the platform's permission status and the user's reservation behavior. That is, between the initiation of the induction call and the formation of the reservation response, whether the status provided by the platform is consistent with the user's operable permissions, serves as a lag indicator of response credibility.

[0091] The upload behavior drift tensor is defined as the shortest path distance between state records submitted by different source platforms in the state transition graph structure, and is used to measure the degree of temporal differences in behavior between platforms.

[0092] The heterogeneous contention index is calculated as follows: Ci = α × (1 / (nsi+1)) + β × (1−mri) + γ × dij;

[0093] nsi represents the number of times the berth's status marker changes within window N1, i.e., the state stability factor; miri represents the berth's permission matching rate, defined as the ratio of the number of successful matches to the total number of inducements, i.e., the master control platform's lag delay coefficient; dij represents the shortest drift path distance of the platform's upload behavior in the state transition graph structure, i.e., the upload behavior drift tensor; α represents the state stability factor weight coefficient, β represents the master control platform lag delay coefficient weight coefficient, and γ represents the upload behavior drift tensor weight coefficient, which are set by system default configuration or based on empirical data;

[0094] Subsequently, using the N1 sliding time window as the anchor point, the status label change trajectory, upload behavior trigger time sequence, access control markers and induction call paths of all berths within this time period are uniformly retrieved and synchronized on the time axis to build a foundation for response coordination analysis across berths.

[0095] Based on this, the coupling relationship between each berth and its surrounding berths is calculated in the following three dimensions:

[0096] Consistency in the direction of status label switching;

[0097] Synchronization of frequency switching;

[0098] The overlap of upload response latency;

[0099] Based on the above three types of coupling relationships, a four-dimensional interaction tensor structure is constructed. This four-dimensional interaction tensor field has four dimensions: berth number, upload trigger time, state switching event identifier, and permission change marker, in order to capture the heterogeneous temporal coupling characteristics of induced response behavior between berths within the N1 window.

[0100] Subsequently, using berths as nodes, a state perturbation propagation graph generation algorithm is invoked. Edge weights are generated based on the coupling degree between the dimensions of the aforementioned tensor, and dynamic connections are established between berth nodes to form a directed weighted graph expressing the temporal coupling relationship of heterogeneous behaviors. On this directed weighted graph structure, a minimum weight edge adsorption strategy is executed. For berth sets with the same master control platform and whose response coupling degree within the N1 window is significantly higher than the neighborhood average, node aggregation is performed to generate a set of induced conflict basic subgraphs, serving as the initial structural basis for subsequent spatial analysis and conflict partitioning.

[0101] Next, using these conflict-inducing sub-graphs as references, spatial mapping is performed in the map coordinate domain, generating a spatial distribution model of graph nodes based on the actual geographical location of the berths. On this basis, an upload anomaly frequency distribution map from the 5G network and a navigation-guided path error density heatmap are overlaid. The overlay method is achieved through a unified grid partitioning and berth number-grid mapping relationship. Ultimately, multiple regions exhibiting both high state coupling density and frequent induced errors are identified; these regions are defined as conflict-inducing potential zones.

[0102] Within each conflict-induced potential zone, the number of state anomaly triggers, the number of induced failure samples, and the heterogeneous contention index value of each berth within the N1 time window are retrieved, taking the berth as the basic unit. Among these, the three indicators reflect the activity of state fluctuations, the sensitivity to mission failure, and the platform conflict density, respectively.

[0103] Subsequently, the multi-factor linear fusion function is called to weight and synthesize the three indicators to generate a berth conflict score for each berth in the potential zone. The score is then used as input to perform clustering on all berths, forming multiple state contention sub-clusters.

[0104] The berth conflict score is calculated in the following way in this embodiment:

[0105] Si = w1 × ei + w2 × fi + w3 × ci;

[0106] Where ei is the number of times the berth's status anomaly is triggered within the N1 time window;

[0107] fi is the number of failed induction records, that is, the number of samples where the actual berth is unusable after induction;

[0108] ci is the heterogeneous contention index;

[0109] w1, w2, and w3 are weight coefficients set by the scheduling strategy, used to express the relative importance of each factor in the conflict intensity.

[0110] The clustering process uses the K-means algorithm, with the input features being the conflict score Si of each berth and its spatial density index within the conflict-induced potential zone.

[0111] The initial number of clusters was determined using the Elbow Method, and the value of k corresponding to the inflection point was selected based on the changing trend of the number of clusters k and the total squared error SSE.

[0112] Clustering results serve as the structural basis for navigation path planning and state masking, ensuring that conflicting berths are effectively classified and managed.

[0113] Next, based on the completion of berth conflict scores and the division of state contention sub-clusters, in order to avoid the phenomenon of concentrated landing points in areas with high incidence of induced misguided navigation paths, which would lead to the navigation engine making incorrect guidance in areas with uncertain states, a dedicated state shield for the guidance engine is further constructed. With the heterogeneous contention index as the core trigger condition, state intervention and path correction operations are performed on the sub-clusters of berths with frequent conflict behaviors.

[0114] Specifically, the threshold for the heterogeneous contention index is first set. The threshold for the heterogeneous contention index is derived from the mean and variance of the distribution of the berth contention index of the whole sample within the N1 sliding window of statistical analysis. The threshold range of the heterogeneous contention index is set by setting multiple standard deviation levels.

[0115] When the contention index of a berth exceeds the set threshold for heterogeneous contention index and falls within an identified sub-cluster of state contention, a state masking strategy is triggered. The current state of the berth is marked as masked in the berth state pool and is automatically removed during navigation path planning. If a berth does not meet the complete masking condition but is still in a highly conflict-related area, its path penalty weight is calculated using the berth conflict score as input. This dynamically increases the cost function in the path planning model, enabling the navigation path to proactively avoid such berth nodes when the shortest path is similar, thereby correcting the path penalty score.

[0116] Furthermore, within each induced conflict potential partition, three parallel simulation streams are periodically executed: state response backtracking, induced result deviation simulation, and master control platform drift simulation. Simultaneously, the parallel simulation streams use historical induced task execution records as input to generate multi-path parallel analysis links.

[0117] During the status response backtracking phase, samples from all executed guidance tasks that show discrepancies between the guided landing and the final berth availability are first retrieved to construct a guidance result deviation cluster. This cluster consists of a four-element data structure with berth number, guidance execution time, user path ID, and final status marker, serving as the baseline deviation source.

[0118] Based on this, a set of hypotheses for rewriting the master control platform is constructed. Specifically, by simulating scenarios where the master control over the issuance of induced tasks changes across different platforms, multiple state rewriting models are generated. In each model, the induced task is set to be mastered by different platforms, and their respective state update mechanisms are re-implanted to form alternative state paths. In the state flow, the rewriting paths are embedded into the original induced state chain according to the original timing and state logic, generating a set of hypothesis drift paths.

[0119] Subsequently, in a parallel simulation environment, the simulation engine is invoked to perform induction effect simulations on all the hypothetical drift path sets mentioned above. The simulation process uses the induction execution path as the main index and the final berth availability label as the result variable. By comparing the induction deviation rate under each drift path, the path model with the smallest deviation is statistically analyzed. Then, through a weighted similarity analysis algorithm, a multi-dimensional similarity matrix is ​​constructed between the historical state label sequence, the permission synchronization time offset sequence, and the user reservation execution result. Finally, the drift path with the optimal similarity is selected as the optimal master control logic simulation model for this potential partition under the current state structure.

[0120] Finally, based on three indicators—the platform permission update time point corresponding to the optimal drift path, the degree of lag in status label response, and the efficiency of induced response—a reverse responsibility fitting operation is performed on the current master control platform.

[0121] The specific method involves constructing a platform responsibility function, using the platform response lag as the input variable and the induced deviation probability as the output of the response function. This function fits the contribution ratio of platform state lag to navigation engine failure and generates suggestions for main control platform state update strategies, including: suggestions for adjusting the upload cycle, suggestions for expanding the permission synchronization window, and suggestions for postponing the approval logic of induced tasks. These suggestions are output in text form for the administrator to receive and execute.

[0122] The platform responsibility fitting function is defined as follows in this embodiment:

[0123] Rp=μ1×τp+μ2×εp+μ3×θp;

[0124] Where τp is the average platform status response delay;

[0125] εp is the platform induction deviation rate, which is the proportion of samples that failed to be induced by the platform's main control.

[0126] θp represents the percentage of platform permission synchronization timeouts;

[0127] μ1, μ2, and μ3 are the weighting coefficients in the responsibility sharing model.

[0128] In the parking fee collection process, in order to automate the processing of parking space usage records and financial settlement, car owners are first required to complete the registration process through the front-end interface before using the platform services. Then, the registration process interface is called to collect the car owner's vehicle information and payment account information, and to perform identity authentication and account binding verification. After registration, a unique car owner identifier and vehicle identifier binding comparison table is generated and stored in the user data pool.

[0129] During the process of a vehicle entering or exiting the parking lot, a video recognition device pre-installed at the entrance and exit captures an image stream at a fixed frame rate, executes a license plate recognition algorithm, extracts the license plate number, and calls the vehicle owner lookup table to complete the identity verification.

[0130] Once the entry time is linked to the license plate number, it is recorded as the start time of parking space usage. After the vehicle leaves, the video recognition device performs exit recognition again and records the end time. The time difference between the two records is the parking space usage duration. The parking space usage duration data is then input into the billing calculator, which calculates the fee payable based on the current parking space billing strategy template. The fee is automatically deducted from the linked payment account, and the transaction status information is returned.

[0131] Finally, parking space usage records and payment data are synchronously written to the cloud data archive area according to timestamp index. The structure fields include vehicle owner ID, parking space number, entry and exit time, usage duration, charging strategy version number and settlement result status, realizing unified archiving and retrieval of parking space resource usage information and financial flow information.

[0132] In terms of maintaining order, parking video recognition devices deployed on both sides of the road continuously collect parking image streams along the route and execute illegal parking detection algorithms. Specifically, the illegal parking detection algorithm combines parking area boundary models and parking space authorization maps to determine the degree of spatial overlap between the spatial location of a vehicle appearing in the video and the current legal parking space pool. If the detection result indicates illegal occupation of the area, the illegal parking event is marked in the image recognition result. Simultaneously, the illegal parking event is written to a scheduling buffer queue, and the vehicle owner's registration information is immediately retrieved to push warning reminders to the vehicle owner via SMS API and voice reminder channels.

[0133] Furthermore, the system calls the available parking space database interface to filter legal and available parking spaces within the current area, executes a fast recommendation algorithm based on path distance and parking space rating, and pushes the optimal parking space path and status to the driver through the mobile interface, prompting them to leave the illegal area as soon as possible and go to the recommended parking space.

[0134] The reminder process is set with a default response period of 8 minutes. After the reminder is sent, the countdown monitoring logic is started. If the vehicle is not detected to leave within the specified period, the illegal parking record processing process is automatically initiated, and an illegal parking record is generated and written to the law enforcement pending queue. The backend scheduling logic calls the law enforcement task scheduling table, marks the corresponding road segment as a high-priority patrol segment, and the backend law enforcement personnel handle the illegally parked vehicle offline based on the patrol trajectory.

[0135] For car owners with multiple parking violations or other dishonest behaviors, a credit factor is constructed based on the cumulative number of parking violations, the number of unresponsive parking violations, and the proportion of historical response timeouts. This forms a mechanism for recording and reporting the behavior of dishonest car owners. For users who reach the credit threshold, their access to certain high-level public parking spaces is restricted through platform permission control logic, thus realizing the governance intervention of credit constraints in the static resource allocation process.

[0136] Furthermore, to improve the utilization efficiency of parking space resources, especially the sharing rate of vacant parking spaces at night, a unified shared parking space platform will be built, allowing government agencies and residential property management companies to set sharing time periods, access permissions, and reservation status through a management interface. The platform updates the sharing availability marker based on the parking space number and makes it available to users via a mobile app for querying and reservation.

[0137] After a user completes a reservation, the reservation information is automatically recorded, and a reminder is sent before the shared time slot. The frequency of shared parking space usage and the reservation rate are simultaneously tallied, and points or subsidies are awarded to the property owner based on incentive rules to encourage sharing and create a positive feedback loop of resource openness and benefit return.

[0138] This invention also proposes a smart city traffic optimization system based on a 5G network, such as... Figure 2 As shown, it includes: a traffic flow acquisition and data structuring module, a cloud analysis and signal dispatching module, a berth status recommendation and guidance release module, and a berth conflict identification and comprehensive management module, with signal connections between each module.

[0139] The traffic flow acquisition and data structuring module is mainly used to collect traffic flow and signal phase status data, upload and cache them through the 5G network, and build a dataset with device number, timestamp and geographic location index;

[0140] The cloud-based analytics and signal scheduling module is mainly used to clean, extract features, and model the uploaded traffic data, generate traffic flow prediction results and traffic status labels, identify congested intersections, and perform signal timing adjustments.

[0141] The parking space status recommendation and guidance release module is mainly used to collect parking space occupancy status and establish a parking space status pool. It combines the driver's location and parking space attributes to perform route calculation, generate parking space recommendations and push them to the terminal, and release guidance information at the same time.

[0142] The berth conflict identification and comprehensive management module is mainly used to construct berth status behavior sequences and heterogeneous contention indices, identify potential conflict areas and classify status contention sub-clusters, perform status masking and path correction for high-conflict berths, conduct induced deviation simulation and platform responsibility fitting, collect vehicle owner entry and exit records to complete settlement and illegal parking processing, and establish a shared berth reservation and access control mechanism.

[0143] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0144] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0145] Those skilled in the art will recognize that the modules and 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 inventive 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.

[0146] 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.

[0147] 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 the claims.

[0148] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A smart city traffic optimization method based on a 5G network, characterized in that it includes: Collect traffic flow and signal phase status data, upload and cache them via 5G network, and create a dataset with device ID, timestamp and geographic location index; The uploaded traffic data is cleaned, features are extracted and modeled to generate traffic flow prediction results and traffic status labels, and congested intersections are identified and signal timing adjustments are performed. Collect parking space occupancy status and establish a parking space status pool. Combine the driver's location with the parking space attributes to perform route calculation, generate parking space recommendations and push them to the terminal, and at the same time release guidance information. Retrieve historical events related to induced failures from the cloud berth status pool, extract permission status mismatch samples by comparing the induced generation time with the platform permission response time, classify and analyze the behavioral characteristics of each platform in terms of status label switching, reporting interval and authorization adjustment cycle, construct a drift coupling curve reflecting the response lag law, and derive the minimum observation time window based on the overlapping interval of multi-platform status fluctuations. Within the minimum observation time window, the state upload behavior sequence of each berth is constructed, multi-dimensional indicators are extracted to generate heterogeneous contention index, and then the temporal coupling relationship between berths is analyzed. A state disturbance propagation map and an induced conflict base sub-map are generated. Through spatial mapping and multi-source indicator superposition, conflict-induced potential partitions are identified. Finally, berths are clustered according to conflict scores to form multiple state contention sub-clusters. Based on the berth conflict score and the division of state contention sub-clusters, a threshold for heterogeneous contention index is set. Berths with contention indices exceeding the threshold and belonging to high conflict sub-clusters are subject to state masking or path penalty score correction, and their participation weight in navigation path planning is adjusted. Within each conflict-induced potential partition, a cluster of induced task deviations, a set of platform rewriting hypotheses and hypothesis drift paths are constructed. Parallel simulation analysis is conducted. By comparing the induced deviations under each hypothesis path, the path model with the smallest deviation is selected. Based on the state change information and induced response records of the optimal drift path, the responsibility ratio of the main control platform is backfitted to generate suggestions for adjusting the state upload and permission synchronization strategies of the main control platform. For high-conflict parking spaces, implement status masking and path correction, conduct induced deviation simulation and platform responsibility fitting, collect vehicle owner entry and exit records to complete settlement and illegal parking processing, and establish a shared parking space reservation and access management mechanism.

2. The smart city traffic optimization method based on 5G network according to claim 1, characterized in that: Deploy traffic flow sensors and traffic signal monitoring equipment to collect traffic flow quantity and signal phase status. Then, use the 5G network to upload, forward, and classify the traffic flow quantity and signal phase status data at high speed. At the same time, establish a structured dataset indexed by device number, timestamp, and geographical location, and perform format verification and anomaly isolation processing.

3. The smart city traffic optimization method based on 5G network according to claim 1, characterized in that: During the cloud data processing stage, traffic perception data is cleaned, features are extracted, and classified and modeled to generate traffic flow prediction results and traffic status labels for each path. Based on the prediction values, intersections that are about to become congested are identified and included in the signal optimization objects. Traffic efficiency is improved by adjusting the green light duration.

4. The smart city traffic optimization method based on 5G network according to claim 1, characterized in that; The system collects and reports the occupancy status of each parking space, constructs a parking space status pool and establishes a multi-dimensional index, performs path calculation and accessibility judgment based on the driver's current location and parking space attributes, generates parking space recommendation results and pushes them to the user's terminal, and simultaneously publishes information on the remaining parking spaces in the area through the guidance screen.

5. The smart city traffic optimization method based on 5G network according to claim 1, characterized in that; The system collects and binds vehicle owner and vehicle information, records vehicle entry and exit times to generate parking space usage records and completes settlement, identifies illegal parking behavior and sends reminders, and generates illegal parking records and initiates law enforcement dispatch if there is no response. The system also builds a shared parking space platform that supports reservation and usage permission management.

6. A smart city traffic optimization system based on a 5G network, characterized in that, include: The system includes a traffic flow acquisition and data structuring module, a cloud analysis and signal dispatching module, a berth status recommendation and guidance release module, and a berth conflict identification and comprehensive management module, with signal connections between the modules. The traffic flow acquisition and data structuring module is mainly used to collect traffic flow and signal phase status data, upload and cache it through the 5G network, and build a dataset with device number, timestamp and geographic location index; The cloud-based analytics and signal scheduling module is mainly used to clean, extract features, and model the uploaded traffic data, generate traffic flow prediction results and traffic status labels, identify congested intersections, and perform signal timing adjustments. The parking space status recommendation and guidance release module is mainly used to collect parking space occupancy status and establish a parking space status pool. It combines the driver's location and parking space attributes to perform route calculation, generate parking space recommendations and push them to the terminal, and release guidance information at the same time. The berth conflict identification and comprehensive management module is mainly used to retrieve historical events related to induced failures in the cloud berth status pool. By comparing the induced generation time with the platform permission response time, it extracts permission status mismatch samples, classifies and analyzes the behavioral characteristics of each platform in terms of status label switching, reporting interval and authorization adjustment cycle, constructs a drift coupling curve that reflects the response lag law, and derives the minimum observation time window based on the overlapping interval of multi-platform status fluctuations. Within the minimum observation time window, the state upload behavior sequence of each berth is constructed, multi-dimensional indicators are extracted to generate heterogeneous contention index, and then the temporal coupling relationship between berths is analyzed. A state disturbance propagation map and an induced conflict base sub-map are generated. Through spatial mapping and multi-source indicator superposition, conflict-induced potential partitions are identified. Finally, berths are clustered according to conflict scores to form multiple state contention sub-clusters. Based on the berth conflict score and the division of state contention sub-clusters, a threshold for heterogeneous contention index is set. Berths with contention indices exceeding the threshold and belonging to high conflict sub-clusters are subject to state masking or path penalty score correction, and their participation weight in navigation path planning is adjusted. Within each conflict-induced potential partition, a cluster of induced task deviations, a set of platform rewriting hypotheses and hypothesis drift paths are constructed. Parallel simulation analysis is conducted. By comparing the induced deviations under each hypothesis path, the path model with the smallest deviation is selected. Based on the state change information and induced response records of the optimal drift path, the responsibility ratio of the main control platform is backfitted to generate suggestions for adjusting the state upload and permission synchronization strategies of the main control platform. For high-conflict parking spaces, implement status masking and path correction, conduct induced deviation simulation and platform responsibility fitting, collect vehicle owner entry and exit records to complete settlement and illegal parking processing, and establish a shared parking space reservation and access management mechanism.

Images