Urban traffic dynamic deduction and real-time planning method and system based on multi-source position data fusion

Abstract

The invention provides an urban traffic dynamic deduction and real-time planning method and system based on multi-source position data fusion. Time-space references of signaling, MDT and MR data are unified through a dynamic rasterization mechanism, and the problem of data isomerism is solved by adopting anchor point selection of signal weighting and a triple criterion time expansion algorithm; generating a minute-level updated dynamic OD matrix based on the hierarchical road network constraint model; a METANET macroscopic traffic flow model and EKF extended Kalman filtering are fused to realize traffic state deduction, and parameter online calibration is supported; signal timing optimization, path induction and emergency control schemes are generated in combination with real-time deduction results, and a sensing-deduction-decision-evaluation closed loop is formed. The method can improve road network traffic efficiency, and is especially suitable for response of sudden traffic events.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, and is particularly suitable for real-time traffic status deduction and dynamic planning decision-making based on the fusion of multi-source location data. Background Art

[0002] Current urban traffic planning relies primarily on static historical data (e.g., manual surveys and fixed detectors), which suffers from three major bottlenecks: poor timeliness, limited coverage, and high costs. While operator location data (base station signaling, MDT, and MR) offers advantages such as full coverage, strong real-time performance, and multi-dimensional trajectories, practical applications still face the following core challenges: Data heterogeneity: The spatiotemporal accuracy of multi-source data varies significantly. Signaling data has coarse-grained positioning (hundreds of meters), MDT data has high accuracy (tens of meters), and MR data is strongly correlated with network quality. This lack of a unified spatiotemporal benchmark and quality assessment standards creates "data silos." Dynamic modeling bottlenecks: Traditional OD matrix generation methods rely on offline historical data and are unable to integrate heterogeneous multi-source data in real time. This makes it difficult to construct a dynamic OD matrix that updates minute-by-minute, resulting in distorted traffic flow representation. Planning response lags: Existing systems utilize offline analysis, making real-time predictions and dynamic decision-making impossible for sudden traffic events (accidents, large-scale events), resulting in significant delays in emergency response. Model generalization flaws: Classic traffic flow models lack an online parameter calibration mechanism driven by multi-source data. This significantly amplifies prediction errors in the face of sudden changes in road network topology (such as construction closures) or unusual scenarios. Lack of a closed-loop decision-making process: The lack of a real-time feedback loop from "data perception → state deduction → decision generation → effect verification" results in a disconnect between planning solutions and actual traffic conditions. Therefore, this paper proposes a method and system for dynamic deduction and real-time planning of urban traffic using multi-source location data fusion. Summary of the Invention

[0003] To address the three major shortcomings of existing urban traffic planning, which rely on static data and result in poor timeliness, limited coverage, and delayed emergency response, as well as the core issues faced by the fusion of multi-source location data (signaling / MDT / MR), the present invention provides a method and system for dynamic deduction and real-time planning of urban traffic using multi-source location data fusion.

[0004] Core algorithm process:

[0005] Step 1: Spatiotemporal fusion of multi-source data

[0006] The flow chart of the multi-source data spatiotemporal fusion algorithm is in the following appendix. Figure 2 middle.

[0007] 1. Multi-source data fusion preprocessing

[0008] Dynamic gridding mechanism, adaptively adjusting grid size based on regional base station density:

[0009]

[0010] Realize the spatial benchmark unification of signaling data, MDT high-precision positioning data, and MR network quality data.

[0011] Normalize signal quality and establish a multi-source signal strength mapping model:

[0012]

[0013] The weight coefficients (α, β, γ) are dynamically configured through cluster analysis of historical data (for example, the β weight in urban scenarios is increased by 0.3, and the α weight in suburban scenarios is increased by 0.4), solving the problem of inconsistent dimensionality of multi-source signal strength.

[0014] Time window intelligent completion sets time windows based on the characteristics of different data sources:

[0015] (1) Signaling data: Complete the 20-second time window.

[0016] (2) MR data: Complete the 5-second time window.

[0017] (3) MDT data: directly use the original timestamp.

[0018] 2. Signal-weighted anchor point selection mechanism

[0019] Improve the anchor point selection criteria, and optimize the traditional algorithm that simply relies on the number of location point records to a weighted scoring model based on stay duration and signal strength; introduce POI semantic weights (such as subway station POI weight = 1.5, park POI weight = 0.8) to enhance the ability to distinguish the attributes of stay points.

[0020]

[0021] Prioritize high stability areas, and screen out grids with stable signals and long dwell times as anchor points through weighted scoring to enhance the anchoring effect of core dwell points.

[0022] 3. Multidimensional drift determination conditions

[0023] Introducing triple judgment conditions in the time dilation process:

[0024] Dynamic time threshold:

[0025]

[0026] Signal stability constraints:

[0027] The signal strength change rate of adjacent records is required to be less than 20%:

[0028]

[0029] Geofence verification:

[0030] If the location is at a transportation hub such as a subway station or train station, the time dilation mechanism is disabled.

[0031] 4. Intelligent assimilation execution process

[0032] (1) Initialization:

[0033] Input: User multi-source location dataset

[0034] Output: Optimized set of stay points

[0035] Current dataset ←Preprocessing( )

[0036] Result Set ←

[0037] (2) Anchor point selection:

[0038] Calculate weighted scores for each raster

[0039] Select the highest scoring grid As an anchor

[0040] extract Corresponding data subset

[0041] (3) Signal perception expansion:

[0042] Traversal Temporally adjacent records

[0043] When both:

[0044]

[0045] Perform time dilation when:

[0046] Generate merged data subsets

[0047] (4) Multidimensional assimilation:

[0048] Build time coverage interval

[0049] from Delete records that meet the following conditions:

[0050]

[0051]

[0052] (5) Iteration termination:

[0053] like , return to step 2

[0054] Otherwise output

[0055] Step 2: Dynamic OD matrix generation

[0056] Stop point identification:

[0057] Multi-threshold stay determination is adopted: the threshold in the core area (commercial district / station) is φ=5min, and in the suburbs is φ=10min;

[0058] Introducing semantic enhancement: Combined with POI data to determine the attributes of the stop point (such as workplace, residential area).

[0059] Trip chain extraction:

[0060] Constructing a space-time directed graph , node V = stay point, edge E = travel path.

[0061] Hierarchical road network constraint model:

[0062]

[0063] The first layer: road topology connectivity constraints.

[0064] The second layer: real-time traffic flow speed constraints.

[0065] The third layer: POI semantic path preference (e.g., prioritizing main roads during commuting hours).

[0066] Dynamic OD calculation:

[0067]

[0068] Where t is the time slice (5-minute granularity), k and m are traffic zones.

[0069] Step 3: Traffic status simulation

[0070] Input: Dynamic OD matrix + real-time traffic conditions (floating car GPS / geomagnetic detector)

[0071] Deduction model:

[0072] The core step in achieving global traffic state prediction is to integrate a macroscopic traffic flow model with the Extended Kalman Filter (EKF). The core idea is to construct a spatial dynamic correlation of road network states using the METANET macroscopic traffic flow model, and then use the EKF for state estimation and online parameter calibration. Ultimately, the global traffic state is derived based on prediction data from a limited number of detection points. The specific process is as follows:

[0073] 1. METANET Macro Traffic Flow Modeling

[0074] Objective: Discretize the highway into segments and establish the traffic flow dynamic equation.

[0075] Formula description: Assume that the road network is divided into N lengths The time is discretized into a step size T. For the mth segment at time k:

[0076] Traffic density :

[0077]

[0078] in:

[0079] : Upstream inflow

[0080] : Outflow of this section

[0081] , : Ramp inflow / outflow

[0082] : Number of lanes (explicitly models changes in the number of lanes)

[0083] average speed :

[0084]

[0085] in:

[0086] : Equilibrium velocity-density relationship (such as Greenshields linear model)

[0087] : Model parameters (slack time, expected coefficients, etc.)

[0088] Innovation: Explicitly incorporating lane number changes and ramp flow , , accurately depicting the topological impacts of interchange hubs, tunnels, etc.

[0089] 2. State Space Representation and EKF Design

[0090] Objective: Convert METANET into a state-space model and use EKF to fuse detection data.

[0091] State vector definition:

[0092]

[0093] in

[0094] : Density of all segments

[0095] : Speed ​​of all sections

[0096] : Time-varying parameters (such as basic map parameters, undetected ramp flow)

[0097] Dynamic system equations:

[0098] ,

[0099] where f(⋅) is derived from the METANET equation and Q is the process noise covariance.

[0100] Observation equation (detector point data):

[0101] ,

[0102] H is the measurement matrix (only the rows corresponding to the detection points are non-zero), and R is the observation noise covariance.

[0103] 3. EKF recursive deduction process

[0104] step:

[0105] (1) State prediction:

[0106]

[0107]

[0108] where F is the Jacobian matrix of f(⋅).

[0109] (2) Kalman gain calculation:

[0110]

[0111] (3) Status update:

[0112]

[0113]

[0114] Innovation:

[0115] Parameter online calibration: the basic graph parameters Incorporate state vectors to correct model deviations in real time.

[0116] Reconstruction of the state of undetected points: Through the spatial correlation of the H matrix, the global density and velocity are inferred from the detected point data.

[0117] 4. Global State Prediction Driving Mechanism

[0118] Input: Machine learning prediction results of detection points (such as future traffic volume / speed output by LSTM).

[0119] Deduction process:

[0120] The predicted value of the detection point t+1 to t+T As a virtual observation sequence.

[0121] From the current state Start by iterating the EKF prediction step:

[0122]

[0123] At each prediction step t+i, the virtual observation value is used Update status:

[0124]

[0125] Output: Global state in the future period , .

[0126] Step 4: Dynamic programming decision

[0127] Solution Generation:

[0128] 1. Signal optimization: Adjust the traffic light cycle based on the simulation results, with the objective function being to minimize the total delay;

[0129] 2. Route Guidance: Push detour routes to navigation apps via V2X to balance road network load;

[0130] 3. Emergency response: Generate control plans (such as closing ramps and giving priority to public transportation) in the event of an emergency.

[0131] Evaluation Mechanism:

[0132] The digital twin platform is used to simulate and verify the effectiveness of the solution. The evaluation indicators include:

[0133] 1. Average road network speed increase rate.

[0134] 2. Intersection queue length reduction rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0135] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0136] Figure 1 Schematic diagram of a method and system for urban traffic dynamic deduction and real-time planning based on multi-source location data fusion in an embodiment of the present invention;

[0137] Figure 2 Schematic diagram of the multi-source data spatiotemporal fusion algorithm flow. DETAILED DESCRIPTION

[0138] The preferred embodiments of the present invention are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0139] Example 1:

[0140] Application scenario: An accident occurred on a main road during the morning rush hour in a city

[0141] 1. Data input:

[0142] Signaling data detected a surge in users on the accident section;

[0143] MDT data locates the accident point (longitude X, latitude Y);

[0144] The floating vehicle data feedback shows that the vehicle speed has dropped to 5km / h.

[0145] 2. Dynamic response:

[0146] The deduction engine predicts the congestion spread range in 30 minutes;

[0147] Decision-making system generates solutions:

[0148] The east-west green light time at the upstream intersection is extended by 15%;

[0149] Induce vehicles to detour via parallel roads A and B.

[0150] 3. Effect verification:

[0151] The traffic capacity of the accident section has been restored to 75% of normal;

[0152] Traffic volume on detour route A increased by 40%, but did not cause new congestion.

Claims

1. A method for urban traffic dynamic deduction and real-time planning based on multi-source location data fusion, characterized by include: A dynamic gridding mechanism unifies the spatial benchmarks of signaling, MDT, and MR data. The grid size is adaptively adjusted based on base station density (50m / 100m / 500m). A signal-weighted anchor point selection model is used to integrate dwell time, signal strength, and POI semantic weights to calculate grid scores. The weight of subway station POI is 1.5, and the weight of park POI is 0.8; A dynamic OD matrix with minute-level updates is generated based on the hierarchical road network constraint model. The path cost function includes real-time speed and POI preference:

2. Construct the METANET-EKF joint deduction model, state vector Number of included lanes and ramp flow , Generate signal timing optimization solutions with response delays less than 10 seconds, and verify the average speed improvement rate and queue length reduction rate of the road network through the digital twin platform.

3. The method according to claim 1, characterized in that The dynamic rasterization includes: When the base station density is greater than 100 / km², a 50m grid is used; When the base station density is 50-100 / km², a 100m grid is used; When the base station density is less than 50 / km², a 500m grid is used.

4. The method according to claim 1, wherein The signal normalization model middle: In urban areas, the β weight is increased by 0.3 and the α weight is reduced by 0.2; In the suburban scene, the α weight is increased by 0.4 and the γ weight is reduced by 0.

3.

5. The method according to claim 1, wherein The time dilation must simultaneously meet three criteria: Dynamic time threshold (5 minutes for transportation hubs, 10 minutes for commercial areas, and 15 minutes for residential areas); signal change rate <20%: ; Non-transportation transfer hub area.

6. The method according to claim 1, characterized in that In the METANET-EKF model, the dynamic equation is defined as:

7. EKF recursive process calibrates parameters through virtual observation sequence , predict the traffic status in the next 15-30 minutes.

8. A system for implementing the method according to claims 1-5, characterized in that include: Multi-source data interface module, which connects to the operator's signaling platform, MDT collection system and MR database in real time through the Kafka stream interface; The spatiotemporal fusion engine performs dynamic rasterization, signal normalization, and intelligent assimilation based on triple criteria. The dynamic OD generation module outputs minute-level OD matrices to the traffic state deduction engine. The planning and decision-making module generates signal timing optimization instructions and sends them to roadside signals. The instruction format is a JSON structure.

9. The system according to claim 6, wherein: The interaction between the deduction engine and the roadside equipment includes: pushing path guidance information to the vehicle terminal through the V2X communication protocol; receiving the real-time speed data of the floating vehicle and updating the METANET model parameters. .

10. The system according to claim 6, wherein: The digital twin verification platform includes: a road network status visualization module that renders congestion diffusion heat maps in real time; a solution effect evaluation module that calculates the average speed improvement rate .

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory, wherein: When the processor executes the program, the steps of the method according to any one of claims 1 to 5 are implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method according to any one of claims 1 to 5 are implemented.

Images