A hierarchical processing and scheduling method for intelligent transportation data

Abstract

This invention relates to the field of data processing technology, specifically to a hierarchical processing and scheduling method for intelligent transportation data, comprising the following steps: acquiring multi-source intelligent transportation data streams of a target area, and then constructing a traffic anchoring data unit set that can be reused across layers; performing hierarchical processing on the traffic anchoring data unit set according to business timeliness and spatial scope, so that the available results within each layer and the distortion-controlled results both carry the validity period identifier and the cross-layer reuse index; selecting delivery candidates that meet the budget from the available results within each layer and the distortion-controlled results based on obsolescence budget description, and generating a delivery execution plan. This invention achieves on-demand adaptation and differentiated control of scheduling response content and behavior, and is particularly suitable for accurate response in diverse business request scenarios in intelligent transportation.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for hierarchical processing and scheduling of intelligent transportation data. Background Technology

[0002] As the reliance on multi-source heterogeneous data in scenarios such as intelligent transportation, urban sensing, and disaster monitoring continues to deepen, data-driven business processing systems face a series of challenges, including untimely data scheduling, difficulty in guaranteeing result accuracy, and strong business heterogeneity. Especially in time- and space-sensitive applications, business requests often include dual requirements for data timeliness and result accuracy, making it difficult for existing systems to balance scheduling efficiency and service reliability.

[0003] Traditional data scheduling mechanisms often rely on spatial matching and timestamp sorting for result filtering, neglecting the multi-level generation sources of results and the distortion handling characteristics at different levels. With the widespread deployment of multi-level caching and downgraded results, scheduling directly based on data generation time or spatial coverage can easily lead to results that are new but have uncontrollable distortion, or results with controllable distortion but outdated data, ultimately resulting in increased business failure rates and significant resource waste. Furthermore, the data delivery phase lacks a scheduling mechanism that dynamically adapts to business needs. The current common practice is to prioritize the most recent and complete results, failing to consider fine-grained business reliability requirements. In high-reliability scenarios such as disaster emergencies and traffic congestion, a lack of elastic configuration in the scheduling strategy will directly threaten the stability of task completion. Summary of the Invention

[0004] This invention provides a hierarchical processing and scheduling method for intelligent transportation data, which is designed in terms of hierarchical result management, distortion-controlled delivery, and scheduling instruction adaptation strategies. It is particularly suitable for intelligent business scheduling systems that need to meet the requirements of high timeliness and high accuracy.

[0005] A hierarchical processing and scheduling method for intelligent transportation data includes the following steps:

[0006] S1. Obtain multi-source intelligent transportation data streams of the target area, map the multi-source intelligent transportation data streams to the road reference frame, and generate validity period identifiers and traceable source information related to the spatial scope and business timeliness of each data, thereby constructing a traffic anchoring data unit set that can be reused across layers. S2. Perform layered processing on the traffic anchoring data unit set according to business timeliness and spatial scope, output the layer-specific available results in each layer, and simultaneously generate the degradeable version of the distortion-controlled results when generating the layer-specific available results, so that both the layer-specific available results and the distortion-controlled results carry the validity period identifier and cross-layer reuse index. S3. Receive a service request and parse it to obtain the obsolescence budget description and reliability requirements corresponding to the service request. Based on the obsolescence budget description, select delivery candidates that meet the budget from the available results within each layer and the distortion-controlled results, and generate a delivery execution plan. Complete the delivery according to the delivery execution plan.

[0007] Optionally, the multi-source intelligent transportation data stream is collected in real time through roadside units, vehicle-mounted units, traffic monitoring systems and mobile terminal devices. The multi-source intelligent transportation data stream includes at least fixed sensor data, floating car trajectory data, traffic light status data, traffic incident report data and map update data. Protocol adaptation and format parsing are performed on the multi-source intelligent transportation data stream to extract the timestamp, location coordinates and business payload from each data stream.

[0008] Optionally, the road reference frame is pre-constructed, and the construction includes a road centerline and lane-level topology as the core. The road reference frame includes a bidirectional mapping index of a linear reference system and a spatial grid coding system. By spatially matching and linearly referencing the position coordinates in each parsed data stream with the road reference frame, the multi-source intelligent transportation data streams are uniformly mapped to linear reference positions or spatial grid cells under the road reference frame, generating spatiotemporally aligned anchored data elements. Based on the road entity range, influence radius, or grid set associated with the anchored data element, its spatial scope description is extracted, and based on the preset effective duration of the service carried by the anchored data element, its service expiration time is calculated. The spatial scope description and the service expiration time are encoded as a validity period identifier.

[0009] Optionally, the generation of traceable source information includes: collecting the original device identifier, data link level and processing history of the anchored data element, generating traceable source information including the data source, collection time, upload path and key processing node fingerprint, and encapsulating it together with the validity period identifier into the metadata field of the anchored data element; Each anchored data element carrying the validity period identifier and traceable source information is aggregated and deduplicated according to the spatial index of the road reference frame to form a traffic anchored data unit set with spatial grid units or road segments as the organizational granularity, supporting on-demand retrieval and cross-layer reference at different business levels.

[0010] Optionally, the layered processing includes parsing the validity period identifier of each data unit in the traffic anchoring data unit set, extracting the business time-limit deadline and spatial scope, and pre-setting multiple business processing levels. The business processing levels include at least a millisecond-level decision layer for real-time control, a second-level analysis layer for traffic situation awareness, and an offline batch processing layer for traffic planning and auditing. Based on the differentiated requirements of each layer for data timeliness and spatial granularity, a layered processing strategy corresponding to each layer is generated. According to the layered processing strategy, timeliness filtering and spatial scope pruning are performed on the traffic anchoring data unit set required by each layer. Data units whose business time-limit deadline is later than the current processing time and whose spatial scope covers the area of ​​interest of the layer are retained. Data merging and statistics are performed according to the time window granularity and spatial aggregation granularity required by the layer to form the intra-layer intermediate view of the layer.

[0011] Optionally, the generation of the available results within the layer includes performing feature extraction, pattern recognition, or index calculation on the intermediate view within the layer that is adapted to the business objectives of that layer, generating available results within the layer that can be directly called by upper-layer businesses or services in the same layer. The available results within the layer include at least lane-level event alarms, average travel time of road segments, regional traffic congestion index, or intersection turning flow distribution. While generating the available results within the layer, a distortion-controlled downgrading process is initiated. By reducing the spatiotemporal sampling frequency, merging adjacent spatial units, reducing quantization precision, or pruning secondary attribute fields, lossy compression and information simplification are performed on the available results within the layer, generating downgradable versions with multiple distortion levels. The distortion error of each downgradable version is calibrated to ensure that its distortion amplitude does not exceed a preset business tolerance threshold, thus forming a distortion-controlled result.

[0012] Optionally, the validity period identifier of the original traffic anchoring data unit is inherited to the available results within the layer and the distortion-controlled results. Based on the original data unit from which the available results within the layer and the distortion-controlled results originate, the business level to which they belong, and the spatial range they cover, a cross-layer reuse index with hierarchical association and forward backtracking path is generated.

[0013] Optionally, S2 further includes encapsulating the cross-layer reuse index and the validity period identifier together into the metadata field of the available results within the layer and the distortion-controlled results, so as to realize the location and on-demand reference of the same data derivative between different processing layers.

[0014] Optionally, S3 receives a service request and parses it to obtain the target spatial area, service type, and accompanying obsolescence budget description and reliability requirements of the service request. The obsolescence budget description includes at least the maximum tolerable data age or the highest available distortion level. Based on the target spatial region and service type, candidate result units that have not expired and have matching spatial coverage are recalled from the available results and distortion-controlled results within the corresponding layer; the current obsolescence margin of each candidate result unit is calculated one by one, and they are sorted according to the degree of obsolescence satisfaction.

[0015] Optionally, based on the ranking results, data to be delivered is selected, and a delivery execution plan including delivery address, transmission protocol and retry strategy is generated in combination with the reliability requirements; the communication interface is called to complete the data delivery according to the plan, and the delivery status is recorded.

[0016] The beneficial effects of this invention are: This invention, by introducing a data obsolescence budget description and a distortion level upper limit constraint, achieves precise screening and dynamic optimization sorting of existing data units. Compared with traditional scheduling methods that directly trigger full calculation or data refresh, this method can prioritize the reuse of existing results from the system cache while ensuring acceptable timeliness and data accuracy for the business. This significantly reduces redundant calculations and data transmission, improves the overall system response efficiency, and reduces storage load and network pressure.

[0017] This invention supports attaching personalized budget information such as maximum tolerable data age and highest available distortion level to business requests, and automatically generates an execution plan including delivery address, transmission protocol and retry strategy in combination with reliability requirements. It breaks the traditional one-size-fits-all scheduling mode and realizes on-demand adaptation and differentiated control of scheduling response content and behavior. It is particularly suitable for accurate response in diverse business request scenarios in intelligent transportation.

[0018] This invention constructs a multi-dimensional screening framework for candidate result units, simultaneously incorporating current obsolescence calculation and distortion level assessment into the candidate screening process. This is then jointly compared with the maximum tolerable data age and the "highest available distortion level" in the business request, forming an effective candidate set that meets business requirements. Compared to traditional scheduling methods that rely solely on spatial matching or data availability, this mechanism improves scheduling decision accuracy. By quantifying the fit between obsolescence and distortion level, it avoids misdelivering usable but unreliable data, improving the effectiveness and stability of business responses. In scenarios such as intelligent transportation and multi-source data fusion, some businesses require both high data timeliness and cannot tolerate high distortion. This method can automatically eliminate potential mismatched results under multiple constraints. Attached Figure Description

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

[0020] Figure 1 This is a schematic diagram of the scheduling method according to an embodiment of the present invention; Figure 2 This is a schematic diagram of scheduling and delivery based on obsolescence budget according to an embodiment of the present invention. Detailed Implementation

[0021] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. For some well-known technologies, those skilled in the art may also use other alternative methods to implement the invention. Moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0022] like Figures 1-2 As shown, a hierarchical processing and scheduling method for intelligent transportation data includes the following steps: S1. Data anchoring for validity period perception: Acquire multi-source intelligent transportation data streams in the target area, map the multi-source intelligent transportation data streams to the road reference frame, and generate validity period identifiers and traceable source information related to the spatial scope and business timeliness of each data, thereby constructing a traffic anchoring data unit set that can be reused across layers.

[0023] S11 collects multi-source intelligent transportation data streams in real time within the target area through roadside units, vehicle-mounted units, traffic monitoring systems, and mobile terminal devices. These data streams include fixed sensor data, floating car trajectory data, traffic light status data, traffic incident report data, and map update data. Protocol adaptation and format parsing are performed on the multi-source intelligent transportation data streams, converting data from different communication protocols into a standard communication interface format acceptable to the system, and extracting unified structure fields from each data stream. ;in, Indicates the first The original data records after parsing, Represents data timestamps, Geographic coordinates The service payload is the part that is actually meaningful for traffic, including service content such as speed, flow rate, occupancy rate, vehicle type, traffic light phase, event type, and congestion level. It does not include IP address, transmission header, checksum, etc. The parsed data enters the unified processing pipeline to provide input for subsequent road reference frame mapping.

[0024] S12, based on spatiotemporal alignment of road reference frame mapping: The original data is geographic coordinates, but the traffic dispatch system is a road topology system. That is to say, the equipment provides latitude and longitude, while traffic control needs to know which road it belongs to, its position on the road, which lane it belongs to, and which control area it belongs to. If no mapping is done, all subsequent layered processing, dispatch budget, etc. will be invalid. Therefore, S12 essentially converts geospatial data into road reference frame data.

[0025] S121, a road reference frame is pre-constructed with the road centerline and lane-level topology as its core. The road reference frame includes a linear reference system, a spatial grid coding system, and a bidirectional mapping index relationship. For each parsed data record... Perform spatial matching and linear reference transformation, and the output is: ,and ;in, For road section markings, It is a linear reference position along that road segment, i.e., the distance from the starting point; specifically, The goal of spatial matching is to determine Which road section does it belong to? By using the nearest distance projection method, all road centerlines are pre-stored as a set of line segments. The perpendicular distance from a point to each candidate line segment is calculated, and the road segment with the smallest distance within a reasonable threshold is selected. The selection criteria prioritize the smallest distance, consistent direction, and continuity of historical trajectory. If it is trajectory data, it combines the road segment from the previous time step and searches only within adjacent topological segments to avoid road skipping. The purpose of linear reference transformation is to determine which road the point falls on. Next, calculate the mileage position on this road. Specifically, the point is first projected onto the road centerline, the cumulative distance from the projected point to the starting point of the road is calculated, and then output. .

[0026] S122, if represented using trellis encoding: The purpose is to divide geographic space into regular units, facilitating aggregation and scheduling. Among these, Encoding spatial grid cells, This is a spatial grid encoding function that uses regular raster encoding to divide the region into fixed-size grids. The row and column numbers of the calculation points are represented as follows: ; ; x represents the horizontal coordinate of the target point, such as longitude or the X-coordinate of the projection plane, and y represents the vertical coordinate of the target point, such as latitude or the Y-coordinate of the projection plane. Indicates the lateral starting reference coordinates of the spatial grid division region. The vertical starting reference coordinates for dividing the spatial grid region. The value represents the side length of a single grid cell, floor represents the floor function, row is the row number of the grid cell containing the target point, and col is the column number of the grid cell containing the target point.

[0027] S123, After completing the mapping, the anchor data element is generated: or ;in, Anchored data elements are spatiotemporally aligned data elements. Essentially, an anchored data element represents a business state that occurs at a certain moment, at a certain road reference location, or on a spatial grid unit.

[0028] S13, Quantitative Identification of Spatial Scope and Business Timeliness: It solves two core problems: the scope of influence of the current data and the timeliness of the current data. Because intelligent traffic scheduling does not only look at what the data is, but also needs to know how large the spatial scope of influence is and how long it is still valid. If these two dimensions are not clear, subsequent scheduling cannot determine whether it has expired, whether it needs to be downgraded, or whether it needs to be spread to adjacent road segments. Therefore, S13 gives each anchored data element a dual label of spatial scope and time validity.

[0029] S131, Spatial Scope Description: Based on anchored data elements The associated road segments or radius of influence generate a set of spatial domains: ;in, Indicates the first A set of spatial scopes for anchored data elements. Indicates a road segment or grid cell; Some data points don't fall precisely on a specific section but rather have a diffuse effect, such as sudden accidents, low visibility weather, or large-scale construction closures. The impact of these data points isn't fixed on a particular road but spreads outwards from a central location. Therefore, a radius is used to define the impact, and all road units within this radius are considered affected, thus introducing the concept of an impact radius. If based on the radius of influence ,but: ;in, This indicates the distance between the road unit and the anchorage. The radius of influence varies depending on the data type. Speed ​​sampling has a very small radius, general congestion uses a medium radius, severe accidents use a large radius, and severe weather uses the entire area. The base radius for different business categories can be preset.

[0030] The spatial scope means that, in addition to its location, this data will affect other road units. Typically, a speed data point only affects the current lane or the current cross section, a congestion event may affect 500 meters or even further ahead, and an accident may affect the entire section. Therefore, data cannot be simply assumed to belong to only one point; it has a range of influence.

[0031] S132, Calculation of Business Validity Deadline: Calculate the business validity deadline based on the preset valid duration. ;in, The deadline for business operations. This is the original timestamp. For effective duration; S133, encodes the spatial scope and business timeliness into a validity period identifier: ;in, This serves as a validity period identifier for anchored data elements.

[0032] S14, Establish source traceability link: Collect the original device identifier, data link level, and processing history of the anchored data element to generate traceable source information, represented as: ;in, For traceable source information, For equipment identification, For upload path identification, This is the set of fingerprints for key processing nodes. A processing node fingerprint can be represented as: ; Indicates the first The fingerprint identifier of each processing node; Will With expiration date Encapsulate them together into the anchored data element: ; This represents an enhanced anchored data element that carries complete metadata.

[0033] S15, Construct a traffic anchorage data unit set: S151, aggregate and deduplicate according to the spatial index of the road reference frame.

[0034] If organized according to a spatial grid: Organizing by spatial grid means that all data falling into the same spatial grid cell are grouped together. This method is suitable for regional situation analysis and large-scale congestion judgment. The rules are uniform, making it easy to retrieve data quickly and to perform parallel computing. If organized by road section: Organizing by road segment means grouping all data belonging to the same road segment together. This method is suitable for signal control, ramp control, and lane-level scheduling. Intelligent transportation is a multi-level system that employs the two methods described above: Control layer: Dependent on road segment organization Situational layer: Relies on spatial grid organization S152, then performs deduplication judgment on duplicate data within the same spatial unit: ; This formula indicates that if two anchored data elements are in the same spatial location, that is... Same, time difference less If the data is not identical, it is considered duplicate data.

[0035] in, To ensure the time is close to the threshold, it is set between 0.5 and 1 times the sampling period, ultimately resulting in: ; For traffic anchoring data unit set, This represents a subset of data organized at the spatial unit level. This traffic anchoring data unit set supports retrieval by spatial unit, filtering by validity period identifier, backtracking by traceable source information, and cross-layer index reuse.

[0036] S2. Generation of layered products based on controllable distortion: The traffic anchoring data unit set is processed in layers according to business timeliness and spatial scope. The usable results within each layer are output in each layer. When generating usable results within a layer, a downgradeable version of the distortion-controlled result is generated simultaneously, so that both the usable results within a layer and the distortion-controlled result carry the validity period identifier and the cross-layer reuse index.

[0037] S21, Generating a Layered Processing Strategy: First, define the data requirements of different business layers, then generate corresponding processing rules. S1 constructs a traffic anchoring data unit set and a validity period identifier for each data unit. However, different business layers have completely different data requirements. The signal control layer needs the latest, detailed, and localized data; the situation analysis layer needs slightly smoother, regional-level data; and the offline planning layer needs complete, long-term data. If all layers use the same processing method, either the control layer will have excessively high latency, the situation analysis layer will have excessively high precision leading to wasted computation, or the offline layer will have incomplete data. Therefore, it is necessary to first define the timeliness and spatial granularity requirements for each layer. Therefore: S211, parse the traffic anchoring data unit set Validity period identifier for each anchored data unit Extract the deadline for business operations. and the set of spatial domains ; Preset multiple business processing layers: ;in, This indicates a millisecond-level decision-making layer with extremely low maximum allowable latency, small time window granularity, strict validity period judgment, and the use of only data near the current time, disallowing the use of data close to expiration. The spatial granularity requirement is at the lane or road section level, without large-scale aggregation, and requires precise positioning without allowing cross-road mixing. This indicates a second-level analysis layer. The timeliness requirement is that a certain delay is acceptable, the time window can be extended, and slight smoothing is allowed. The spatial granularity requirement is at the road segment or grid level, and it supports the merging of adjacent spatial units. It is used for regional congestion analysis and operational status identification. This indicates an offline batch processing layer. The timeliness requirement is not to emphasize real-time processing, but to focus more on completeness. The spatial granularity requirement is at the regional or administrative level, which can perform large-scale aggregation for planning, auditing, or trend analysis.

[0038] S212, for each layer Define its timeliness and spatial granularity requirements: ;in, This indicates the granularity of the time window for this layer. This indicates the spatial aggregation particle size of the layer; S213, Generate a layered processing strategy based on the requirements of each layer: ;in, Indicates the first Layered processing strategy The strategy generation function takes into account the timeliness and spatial granularity requirements of the current layer, and outputs a complete hierarchical processing strategy, including data filtering rules, time window rules, spatial aggregation rules, distortion reduction rule trigger thresholds, and result update frequency. The strategy generation function is implemented according to the following logic: The first step is to generate time filtering rules: determine the time window length based on timeliness requirements, determine whether near-expiration data is allowed, and generate a time filter.

[0039] The second step is to generate spatial trimming rules: determine the spatial unit type and aggregation range based on the spatial granularity requirements, and generate a spatial filter.

[0040] The third step is to generate aggregation strategies based on hierarchy: Control layer: No averaging is performed; only the latest value is retained. Analysis layer: Can perform statistical averaging; Offline layer: Can perform historical trend statistics.

[0041] Step 4: Generate update frequency rules Control layer: High-frequency updates; Analysis layer: Intermediate frequency update; Offline layer: Low-frequency batch updates.

[0042] S22, data filtering and aggregation based on business timeliness and spatial scope: mainly filtering out expired data, filtering out data that is not in the current layer's area of ​​interest, and reorganizing the remaining data according to time and space.

[0043] S221, Timeliness Filtering: Determines whether the business timeliness deadline of the data is later than the current processing time for each anchored data unit. judge: ;in, This indicates that only data units that meet the above conditions will be retained during the current processing time.

[0044] S222, Spatial Scope Clipping: Each layer does not focus on the entire area. The control layer focuses on a specific intersection, the situational awareness layer focuses on a specific urban area, and the offline layer focuses on the entire large area. Therefore, it is determined whether the spatial scope of the data intersects with the area of ​​interest of the current layer. If there is no intersection, even if the data has not expired, it does not belong to this layer and is not included in the calculation of this layer. Define the first... The area of ​​interest is Retain the following: The data unit; where, To anchor the spatial scope of data units, For the first The layer focuses on the spatial range.

[0045] S223, Time Window Aggregation: Different layers have different requirements for time continuity. The control layer only cares about data from the most recent short period, the analysis layer needs averages or trends over a certain period, and the offline layer needs long-term statistics. Without time window division, the data is discrete and cannot be used for stable statistics. Therefore, data that meets the conditions is aggregated according to time windows: ; indicates that all data falling within the j-th time period form a set, that is, grouping the data by time period, where, This represents the start time of the j-th time window. For the first Layer A time window data set This refers to the granularity of the time window for this layer. Assuming the current layer's time window granularity is 1 second, and the current time starts from 10:00:00, then the first time window is 10:00:00~10:00:01; the second time window is 10:00:01~10:00:02, and all data within each window is grouped together.

[0046] S224 merges data by spatial aggregation granularity. It performs spatial aggregation, grouping data belonging to the same spatial unit together, such as data from the same road segment or the same grid cell. This facilitates the calculation of average speed, traffic flow statistics, and congestion index. Represented as: ;in, Indicates the first Layer space unit Aggregated data set, For spatial grid units or road sections; Forming the first Layers: ;in, For the first The middle view within each layer.

[0047] S23, Generate usable results within the layer: For intermediate views within the layer Perform feature extraction or metric calculation: ;in, For the first Results available within the layer. Processing functions adapted to hierarchical business objectives. Available results within the layer may include lane-level event alarms, average travel time per road segment, regional traffic congestion index, and intersection turning flow distribution. Processing functions The process is divided into three stages: Phase 1, State Feature Extraction: Extracting core features describing traffic conditions from aggregated data.

[0048] Phase Two, Pattern Recognition or Metric Calculation: Calculate the results based on hierarchical business objectives.

[0049] Phase 3: Structured encapsulation of results: Generate in-layer usable results that can be directly called.

[0050] S231, in the millisecond-level decision-making layer (control layer), aims for speed, low latency, and direct control applications. It generates lane-level event alarms by extracting current lane speed features, speed change rate features, occupancy rate features, and queue length features. The judgment rules are: speed below a preset threshold, abnormally high occupancy rate, and queue length exceeding the tolerance range. The output includes the generated lane ID, alarm type, trigger time, and severity level; this is the lane-level event alarm.

[0051] S232, in the second-level analysis layer (situation layer), aims to identify operational trends and regional congestion. It generates average travel time for road segments and a regional traffic congestion index, where: Average travel time generation for road segments: Input a set of vehicle trajectories for the same road segment within a time window, calculate the average time difference for each vehicle to pass through the segment, and if there is no complete trajectory, estimate it using speed × segment length. Output the segment ID, current travel time, and timestamp.

[0052] The generation of the regional traffic congestion index is as follows: For all spatial units within a certain region, the average speed, speed standard deviation, density, and occupancy rate are extracted. The above features are normalized and combined to obtain the congestion level, including normal, light congestion, and severe congestion. The region ID, congestion index value, and congestion level are output.

[0053] S233, in the offline batch processing layer, aims to provide statistical analysis, trend analysis, and planning support. It generates intersection turning traffic distribution, taking as input vehicle trajectory data entering the intersection within a time window. It identifies straight-ahead, left-turn, and right-turn directions based on trajectory direction changes, counts the number of vehicles in each direction, and outputs the intersection ID, traffic flow for each turn, and its percentage.

[0054] S24, Simultaneous generation of a reducible, distortion-controlled version of the result: In-layer usable result This is typically a refined, complete version with many fields and a high update frequency. It is suitable for use when resources are plentiful. However, in situations such as data surges, network congestion, or insufficient computing power, if we insist on producing full-precision results, it will either be too late or the output will be unstable. Therefore, the goal of the reduced-order function is to reduce the precision of the output. Converted to multiple lighter, faster, and more stable versions. At the same time, it ensures that information loss (distortion) is within the range that the business can tolerate. Therefore, in generating At the same time, perform order reduction processing: ;in, For the first Layer High-distortion version, The distortion level represents the degree of degradation. Based on budget-driven principles, 'd' is defined as the minimum degradation level required to meet latency / link / reliability budgets. This means first checking the budget provided by the business request; if directly using... If the budget cannot be met, increase d incrementally until it is met. First, try d=0 without reducing the order to estimate if it can be submitted within the budget. If not, try d=1. If it still doesn't work, continue increasing d until a d that meets the budget is found, and output the corresponding value. This is the distortion level of the minimum budget. This is a price reduction mechanism. It includes reducing the time sampling frequency, merging spatial units, reducing numerical quantization precision, and pruning minor fields, which can be executed sequentially. The price reduction mechanism is defined as a hierarchical rule mapping: The strength of each of the four types of downgrade methods is determined, and the strength can be described using a ranking table: d=0: No reduction in order. ; d=1: Only time downsampling is performed, which is mild; d=2: Temporal downsampling + spatial merging, moderate; d=3: Temporal downsampling + spatial merging + numerical quantization, which is relatively heavy; d=4: Add additional field clipping.

[0055] The distortion error is defined as: ;in, This is distortion error. Set a business tolerance threshold for the result difference metric function: ;in, For the first The tolerance threshold for business operations at each layer is determined by simulating different degradation levels using historical real data, statistically analyzing the difference distribution, and identifying the error range that does not affect business decisions as the threshold. Generally, the threshold is set at 3% to 5% for the control layer, 5% to 8% for the analysis layer, and 8% to 12% for the offline layer. Forming a set of distorted and controlled results Different levels and different result types use different methods to measure differences. The difference measurement function adopts categorical measurement technology. For numerical results, such as travel time, speed, and congestion index, a combination of relative error and mean squared error is used to calculate the mean absolute error. For graded results, such as congestion level and risk level, the grade offset distance is used to calculate the grade difference to determine whether it has crossed a grade. For example, if it was originally moderate congestion and was downgraded to mild congestion, it is considered acceptable as long as it does not cross more than one grade. For categorical / event-based results, such as lane alarms, the consistency ratio is used to statistically analyze whether key fields (type, location, severity level) are consistent and calculate the consistency ratio. For example, whether the alarm still exists or whether the severity level is maintained. If the key fields are consistent, the difference is low.

[0056] In addition, the specific methods for the four types of order reduction are as follows: A. Reduce the time sampling frequency, i.e., time downsampling: change the result that was originally updated in every time window to be updated once every few windows, or summarize multiple windows and then output.

[0057] Control layer: Instead of significant downsampling, it primarily outputs only critical alarms; Analysis layer: Window merging is available, for example, merging 1 second into 3 seconds; Offline layer: It is inherently batch processing and can use longer windows; Output format: The results of the continuous window are summarized as recent values / median / quantiles / maximum and minimum values, and then output as a sparser time series.

[0058] B. Merging adjacent spatial units, i.e., spatial merging: combining multiple adjacent lanes / segments / grids into a coarser-grained unit.

[0059] C. Grid layer: Combines adjacent grids into larger grid blocks; Road segment layer: Merge adjacent road segments into corridor segments according to road topology; Intersection level: Combines multiple approach lanes into an approach group; Merging is not arbitrary; it must follow the principles of topological adjacency, same direction or same control unit, and same area of ​​interest.

[0060] C. Reduced precision of numerical quantization: Coarsening continuous values ​​reduces transmission volume and sensitivity to fluctuations. For example, speed is retained in the range of 1 km / h → 5 km / h, congestion index is retained to two decimal places, and flow rate is adjusted from precise values ​​to low / medium / high ranges, which reduces both the amount of data and meaningless jitter caused by noise.

[0061] D. Pruning secondary attribute fields, i.e., field pruning: Define the order of field importance, and gradually remove non-critical fields at higher demotion levels. Taking event alerts as an example: Required fields: spatial location, alarm type, time, severity; Selectable fields include: source details, confidence decomposition, candidate target list, and supplementary explanatory text. Field pruning should follow the principle of minimum available field set for business purposes, so that even after pruning, the field can still be consumed by downstream applications and will not cause the interface to become unavailable.

[0062] S25, Dynamic inheritance of validity period identifiers and binding of cross-layer reuse indexes: S251, Validity Inheritance: Inheriting the validity period identifier of the original anchored data unit. Inherited to and .

[0063] S252, define the cross-layer reuse index as follows: ;in, Represents the set of source data units. Indicates the business level identifier. For spatial coverage area.

[0064] Forming a mapping relationship: Supports: forward tracing, from original data to derived results; and reverse backtracking, from results back to original data units, ultimately encapsulated. ;in, This indicates the available results within a layer that carry an expiration date identifier and a cross-layer reuse index.

[0065] S3. Scheduling and delivery based on obsolescence budget: Receive a service request and parse it to obtain the obsolescence budget description and reliability requirements corresponding to the service request. Based on the obsolescence budget description, select delivery candidates that meet the budget from the available results within each layer and the distortion-controlled results, and generate a delivery execution plan. Complete the delivery according to the delivery execution plan.

[0066] S31, Receive a service request initiated by an upper-layer application or downstream service. Semantic parsing of the request message yields: ;in, For the target spatial range, For the target time window, These are business logic types, including signal control, situation assessment, and auditing. Description of the obsolescence budget. For reliability requirements, , This indicates the maximum allowed data age, which is the upper limit of the difference between the current time and the data generation time. This indicates the highest available distortion level; the lower the value, the higher the fidelity.

[0067] S32, Preliminary screening of the candidate set: from the available results generated within each layer. and distortion-controlled results In the middle, read its metadata fields: Expiry date indication: ; Cross-layer index reuse: ; Select the result units that meet the following three conditions to form the initial candidate set. : Condition one, in, This indicates that the valid deadline for the i-th candidate result is later than the deadline for the business request; Indicates the current time Generation time of candidate result unit i The time difference between them is less than or equal to the maximum tolerable data age set in the user request. Only then can the candidate result be considered acceptable in terms of novelty; Condition 2, Data space coverage matching: ; Condition 3: The level of data distortion is acceptable. ;in, The distortion level of the current result unit.

[0068] get: .

[0069] S33, to The result cells are sorted according to their remaining obsolescence margin: Higher scores indicate fresher and more in line with budget.

[0070] Form an ordered candidate sequence sorted by obsolescence margin: .

[0071] The core purpose of S33 is to avoid redundant calculations or data loading. Through a triple filtering mechanism of spatial matching, time constraints, and distortion control, it quickly finds the optimal matching result from existing data. At the same time, it uses obsolescence margin to quantify quality, enabling the system to intelligently select more suitable and fresher data, thereby serving business requests more efficiently.

[0072] S34, Delivery Execution Plan Generation and Delivery: Select the first or more optimal candidate units from the sorting results: ; represents the set of the top N candidate result units selected from the sorting results; This represents the j-th candidate result unit after being sorted by merit.

[0073] Based on the reliability requirements in the business request Generate the corresponding delivery execution plan , This represents the expected data delivery success rate. This indicates the upper limit of the data delivery response time. This indicates the maximum number of retries; if the initial transmission fails, this represents the maximum number of retries allowed. The execution plan is then submitted. The structure is as follows: ;in, Indicates the identifier of the data to be delivered. Indicates the delivery target address or service identifier. Indicates communication protocol parameters, Indicates the maximum number of retries. This indicates the delivery timeout period.

[0074] Call the communication module to send data: Encapsulate response message ; To target address Initiate delivery; Record the sending timestamp and initial delivery status, and enable receipt monitoring.

[0075] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0076] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A hierarchical processing and scheduling method for intelligent transportation data, characterized in that, Includes the following steps: S1. Obtain multi-source intelligent transportation data streams of the target area, map the multi-source intelligent transportation data streams to the road reference frame, and generate validity period identifiers and traceable source information related to the spatial scope and business timeliness of each data, thereby constructing a traffic anchoring data unit set that can be reused across layers. S2. Perform layered processing on the traffic anchoring data unit set according to business timeliness and spatial scope, output the layer-specific available results in each layer, and simultaneously generate the degradeable version of the distortion-controlled results when generating the layer-specific available results, so that both the layer-specific available results and the distortion-controlled results carry the validity period identifier and cross-layer reuse index. S3. Receive a service request and parse it to obtain the obsolescence budget description and reliability requirements corresponding to the service request. Based on the obsolescence budget description, select delivery candidates that meet the budget from the available results within each layer and the distortion-controlled results, and generate a delivery execution plan. Complete the delivery according to the delivery execution plan.

2. The method for hierarchical processing and scheduling of intelligent transportation data according to claim 1, characterized in that, The multi-source intelligent transportation data stream is collected in real time through roadside units, vehicle-mounted units, traffic monitoring systems and mobile terminal devices. The multi-source intelligent transportation data stream includes at least fixed sensor data, floating car trajectory data, traffic light status data, traffic incident report data and map update data. Protocol adaptation and format parsing are performed on the multi-source intelligent transportation data stream to extract the timestamp, location coordinates and business payload from each data stream.

3. The hierarchical processing and scheduling method for intelligent transportation data according to claim 2, characterized in that, The road reference frame is pre-constructed, with the road centerline and lane-level topology as its core. The road reference frame includes a bidirectional mapping index of a linear reference system and a spatial grid coding system. By spatially matching and linearly referencing the position coordinates in each parsed data stream with the road reference frame, the multi-source intelligent transportation data streams are uniformly mapped to linear reference positions or spatial grid cells under the road reference frame, generating spatiotemporally aligned anchored data elements. Based on the road entity range, influence radius, or grid set associated with the anchored data element, its spatial domain description is extracted, and based on the preset effective duration of the service carried by the anchored data element, its service expiration time is calculated. The spatial domain description and the service expiration time are encoded as an expiration identifier.

4. The hierarchical processing and scheduling method for intelligent transportation data according to claim 3, characterized in that, The generation of traceable source information includes: collecting the original device identifier, data link level and processing history of the anchored data element, generating traceable source information including the data source, collection time, upload path and key processing node fingerprint, and encapsulating it together with the validity period identifier into the metadata field of the anchored data element; Each anchored data element carrying the validity period identifier and traceable source information is aggregated and deduplicated according to the spatial index of the road reference frame to form a traffic anchored data unit set with spatial grid units or road segments as the organizational granularity, supporting on-demand retrieval and cross-layer reference at different business levels.

5. The hierarchical processing and scheduling method for intelligent transportation data according to claim 1, characterized in that, The hierarchical processing includes parsing the validity period identifier of each data unit in the traffic anchoring data unit set, extracting the business time expiration time and spatial scope, and pre-setting multiple business processing levels. The business processing levels include at least a millisecond-level decision layer for real-time control, a second-level analysis layer for traffic situation awareness, and an offline batch processing layer for traffic planning and auditing. Based on the differentiated requirements of each layer for data timeliness and spatial granularity, a hierarchical processing strategy corresponding to each layer is generated. Based on the layered processing strategy, timeliness filtering and spatial range pruning are performed on the traffic anchoring data unit sets required for each layer. Data units whose business timeliness deadline is later than the current processing time and whose spatial scope covers the area of ​​interest of the layer are retained. Data merging and statistics are performed according to the time window granularity and spatial aggregation granularity required by the layer to form the intra-layer intermediate view of the layer.

6. The method for hierarchical processing and scheduling of intelligent transportation data according to claim 5, characterized in that, The generation of usable results within the layer includes performing feature extraction, pattern recognition, or index calculation on the intermediate view within the layer, adapted to the business objectives of that layer, to generate usable results that can be directly called by upper-layer businesses or services in the same layer. The usable results within the layer include at least lane-level event alarms, average travel time of road segments, regional traffic congestion index, or intersection turning flow distribution. Simultaneously with the generation of the usable results within the layer, a distortion-controlled downgrading process is initiated. By reducing the spatiotemporal sampling frequency, merging adjacent spatial units, reducing quantization precision, or pruning secondary attribute fields, lossy compression and information simplification are performed on the usable results within the layer, generating multiple downgradable versions with different distortion levels. The distortion error of each downgradable version is calibrated to ensure that its distortion amplitude does not exceed a preset business tolerance threshold, thus forming a distortion-controlled result.

7. The method for hierarchical processing and scheduling of intelligent transportation data according to claim 6, characterized in that, The validity period identifier of the original traffic anchoring data unit is inherited to the available results within the layer and the distortion-controlled results. Based on the original data unit from which the available results within the layer and the distortion-controlled results originate, the business level to which they belong, and the spatial range they cover, a cross-layer reuse index with hierarchical association and forward backtracking path is generated.

8. The method for hierarchical processing and scheduling of intelligent transportation data according to claim 7, characterized in that, The S2 further includes encapsulating the cross-layer reuse index and the validity period identifier together into the metadata field of the available results within the layer and the distortion-controlled results, so as to realize the location and on-demand reference of the same data derivative between different processing layers.

9. The method for hierarchical processing and scheduling of intelligent transportation data according to claim 1, characterized in that, The S3 receives a service request, parses it to obtain the target spatial area, service type, and associated obsolescence budget description and reliability requirements of the service request. The obsolescence budget description includes at least the maximum tolerable data age or the highest available distortion level. Based on the target spatial region and service type, candidate result units that have not expired and have matching spatial coverage are recalled from the available results and distortion-controlled results within the corresponding layer; the current obsolescence margin of each candidate result unit is calculated one by one, and they are sorted according to the degree of obsolescence satisfaction.

10. The method for hierarchical processing and scheduling of intelligent transportation data according to claim 9, characterized in that, Based on the ranking results, data to be delivered is selected, and a delivery execution plan including delivery address, transmission protocol and retry strategy is generated in combination with the reliability requirements; the communication interface is called to complete the data delivery according to the plan, and the delivery status is recorded.

Images