Intelligent parking resource scheduling system based on multi-source data analysis

By constructing a multi-source data analysis system for a three-level parking twin, the problems of local optimization and global collaboration being disconnected and insufficient dynamic feature capture in traditional parking scheduling systems have been solved. This has enabled efficient and dynamic resource scheduling optimization, improving parking resource utilization efficiency and traffic flow.

CN120954222APending Publication Date: 2025-11-14BOSINI INFORMATION TECH CO LTD
View PDF 0 Cites 3 Cited by

Patent Information

Application Number
CN202511108010.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-08-08
Publication Date
2025-11-14

AI Technical Summary

Technical Problem

Traditional parking dispatch systems suffer from problems such as a disconnect between local optimization and global collaboration, insufficient dynamic feature capture capabilities, insufficient depth of multi-source data fusion, and a lack of closed-loop simulation evaluation and decision execution, resulting in inaccurate resource allocation and insufficient dynamism.

Method used

A parking resource intelligent scheduling system based on multi-source data analysis is constructed. Through dynamic and static data acquisition modules, micro, meso and macro models are built to form a three-level parking twin, realizing data scale transformation and interaction. Combined with algorithm library and strategy selection unit, periodic simulation and feedback optimization are carried out.

Benefits of technology

It achieves full-scale digital mapping from individual vehicle behavior to regional traffic networks, accurately captures the dynamic characteristics of parking resources, optimizes resource scheduling strategies, improves parking resource utilization efficiency and regional traffic flow, and balances efficiency, balance and user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120954222A_ABST
    Figure CN120954222A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of parking resource scheduling, in particular to an intelligent parking resource scheduling system based on multi-source data analysis, which realizes full-scale digital mapping from vehicle individual behaviors and parking lot flow fields to a regional traffic network by constructing microscopic-mesoscopic-macroscopic three-level parking twin bodies and fusing dynamic and static multi-source data. Through scale conversion and data interaction among the three-level model, dynamic characteristics of parking resources in space and time dimensions are accurately captured, a high-fidelity and fine-grained digital twinborn base is provided for scheduling decision making, splitting of local optimization and global collaboration in traditional scheduling is broken through, the accuracy and foresight of resource allocation are supported, and the scheduling efficiency is improved. Based on dynamic construction of an algorithm strategy library and twinborn simulation evaluation, the scheme takes a multi-dimensional scheduling value as a core to realize closed-loop optimization of algorithm combination-effect rehearsal-strategy optimization-dynamic adjustment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of parking resource scheduling technology, and more specifically, to an intelligent parking resource scheduling system based on multi-source data analysis. Background Technology

[0002] With the acceleration of urbanization, the contradiction between the supply and demand of parking resources is becoming increasingly prominent. Traditional parking scheduling systems face two core challenges: First, the disconnect between local optimization and global coordination. Existing solutions mostly focus on single parking lots or isolated road sections, lacking cross-spatial-scale linkage mechanisms, resulting in an "information silo" effect in resource allocation. Second, insufficient ability to capture dynamic features. Parking demand is affected by multiple dimensions such as time, space, and events, but traditional models struggle to map the complex changes in the physical world in real time, causing scheduling strategies to lag behind actual scenarios. In addition, algorithm adaptability is limited. Most systems adopt static rules or a single optimization objective, making it difficult to balance diverse needs such as efficiency, fairness, and user experience, especially during peak hours or unexpected events, which can easily lead to secondary congestion.

[0003] The rise of digital twin technology has provided a new paradigm for parking resource scheduling, but existing technologies still face key bottlenecks: on the one hand, the depth of multi-source data fusion is insufficient, and the integration of dynamic and static data (such as vehicle trajectories, facility status, and road network topology) often remains at a simple superposition level, lacking a scale transformation mechanism from micro-level individual behavior to macro-level system characteristics; on the other hand, the closed loop between simulation evaluation and decision execution is missing, and most systems can only complete a single strategy generation, unable to achieve dynamic evolution of strategies through periodic simulation feedback. Against this backdrop, this invention proposes an intelligent parking resource scheduling system based on multi-source data analysis. Summary of the Invention

[0004] In view of the shortcomings of existing technologies, the purpose of this invention is to provide an intelligent parking resource scheduling system based on multi-source data analysis.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A smart parking resource scheduling system based on multi-source data analysis, including

[0007] The static and dynamic data acquisition module is used to collect static basic data and dynamic real-time data;

[0008] The micro-model building module is used to build micro-models;

[0009] The meso-level model building module is used to build meso-level models;

[0010] The macro model building module is used to build macro models;

[0011] The three-level parking twin construction module constructs a three-level parking twin for a specified area through data transfer and scale transformation between micro, meso, and macro models.

[0012] The scheduling involves a strategy selection unit. Based on the three-level parking twin, a physical model is periodically loaded, and an initial algorithm library that conforms to the current scenario is constructed simultaneously. The algorithms in the initial algorithm library are randomly combined to generate multiple algorithm switching strategies. The scheduling selection algorithm switching strategy is selected from the algorithm switching strategies, and the scheduling selection algorithm switching strategy is used to schedule parking resources in the specified area.

[0013] Furthermore, static basic data includes micro-level static basic data, meso-level static basic data, and macro-level static basic data, while dynamic real-time data includes micro-level dynamic real-time data, meso-level dynamic real-time data, and macro-level dynamic real-time data.

[0014] Furthermore, the micro-level static basic data includes data from the vehicle attribute database and data from the parking space attribute database;

[0015] The static basic data at the meso level includes parking lot topology data and facility parameters;

[0016] The macro-level static basic data includes regional road network topology data and parking lot node attributes.

[0017] Furthermore, the micro-level dynamic real-time data includes the vehicle's real-time location, speed, steering intention, remaining range, and user response to recommended parking spaces;

[0018] The dynamic real-time data at the meso level includes vehicle trajectories within the parking lot, vehicle density within the passageways, queue length at the entrance, and parking space occupancy status.

[0019] Macro-level dynamic real-time data includes road cross-section traffic density, average vehicle speed, intersection delay time, parking algorithm strategies, and total parking demand in the area.

[0020] Furthermore, the process of constructing the microscopic model:

[0021] Define the state space (S): S = [real-time vehicle location, distance to destination, current vehicle speed, remaining range, distance to the nearest available parking space, system recommended parking space ID and distance, recommended parking space type, surrounding vehicle density, weather impact coefficient, user's historical probability of accepting recommendations, reservation status, remaining lock time];

[0022] Define the action space (A): A = [Accept recommended parking space and proceed, reject recommended parking space and continue searching, park in the currently found empty parking space, abandon the current parking space and continue driving, switch target parking lot];

[0023] Define the reward function (R): R = -α × (time to reach parking space) + β × (parking space matching degree) - γ × (contribution to surrounding congestion) + δ × (user preference matching degree);

[0024] A microscopic model is eventually built through training and iteration.

[0025] Furthermore, the calculation process for the contribution of surrounding congestion is as follows: Contribution of surrounding congestion = Contribution of passage delay * Delay contribution weight + Contribution of interference from adjacent parking spaces * Interference contribution weight + Contribution of area density exceeding threshold * Exceeding threshold contribution weight + Contribution of departure road impact * Road impact contribution weight.

[0026] Furthermore, the contribution of the passageway to traffic delay = ∑(number of subsequently affected vehicles × delay time per vehicle × location sensitivity coefficient);

[0027] The contribution of adjacent parking spaces to interference = ∑(number of affected adjacent parking spaces × interference degree coefficient × ...

[0028] Parking space type coefficient);

[0029] Contribution of regional density exceeding threshold = max(0, actual density - optimal threshold) × region area × density sensitivity coefficient;

[0030] The contribution of departure roads to the overall impact = the increase in departure path delay + the merging delay of surrounding roads.

[0031] Furthermore, the calculation process for user preference matching degree is as follows: User preference matching degree = (parking space type matching degree × parking space type weight) + (location matching degree × location matching weight).

[0032] Furthermore, the scheduling selection algorithm switching strategy is selected from the algorithm switching strategies: the parking resource scheduling values ​​of the three-level parking twin under different algorithm switching strategies are analyzed, and the algorithm switching strategy with the largest parking resource scheduling value is marked as the scheduling selection algorithm switching strategy.

[0033] Furthermore, the parking resource scheduling value of a three-level parking twin under an algorithm switching strategy: Determine an algorithm switching strategy, control the three-level parking twin to execute the algorithm switching strategy for a one-cycle simulation scheduling, after the simulation ends, obtain the micro scheduling index, meso scheduling index and macro scheduling index of the three-level parking twin, sum the micro scheduling index, meso scheduling index and macro scheduling index to calculate the parking resource scheduling value of the three-level parking twin under the algorithm switching strategy;

[0034] The steps to obtain the micro-scheduling index are as follows: obtain the average parking time of all vehicles in the three-level parking twin, the acceptance rate of all users for recommended parking spaces, and the EV charging space matching success rate. The micro-scheduling index is calculated based on the average parking time, the acceptance rate of users for recommended parking spaces, and the EV charging space matching success rate.

[0035] The steps to obtain the meso-level scheduling index are as follows: obtain the average parking space utilization rate of all parking lots and the average queuing time of all parking lot entrances in the three-level parking twin, and calculate the meso-level scheduling index based on the average parking space utilization rate and the average queuing time.

[0036] The steps to obtain the macro-dispatch index are as follows: obtain the average vehicle speed of all roads and the average load difference of all parking lots in the three-level parking twin, and calculate the macro-dispatch index based on the average vehicle speed and the average load difference.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This invention's system constructs a three-tiered parking twin—micro, meso, and macro—and integrates dynamic and static multi-source data to achieve full-scale digital mapping from individual vehicle behavior and parking lot flow fields to regional traffic networks. The scale transformation and data interaction between the three-tiered models accurately capture the dynamic characteristics of parking resources in spatial and temporal dimensions, providing a high-fidelity, fine-grained digital twin foundation for scheduling decisions. This breaks down the disconnect between "local optimization and global coordination" in traditional scheduling, supporting the accuracy and foresight of resource allocation. Relying on the dynamic construction of an algorithm strategy library and twin simulation evaluation, the solution uses multi-dimensional scheduling values ​​as its core, achieving closed-loop optimization of "algorithm combination - effect pre-performance - strategy optimization - dynamic adjustment." The algorithm switching strategy adapts to the characteristics of different time periods and scenarios, balancing multiple objectives such as efficiency, uniformity, and user experience. Continuous periodic simulation and indicator feedback allow the scheduling strategy to dynamically evolve with the physical world, effectively improving parking resource utilization efficiency, regional traffic flow, and user service quality. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the intelligent parking resource scheduling system based on multi-source data analysis.

[0040] Figure 2 A flowchart illustrating the construction process of a microscopic model;

[0041] Figure 3 A flowchart illustrating the construction process of a mesoscopic model;

[0042] Figure 4 This is a flowchart of the macroscopic model construction process. Detailed Implementation

[0043] Reference Figures 1 to 4The intelligent parking resource scheduling system based on multi-source data analysis includes a dynamic and static data acquisition module, a micro-model construction module, a meso-model construction module, a macro-model construction module, a three-level parking twin construction module, and a scheduling strategy selection unit.

[0044] Construct a three-level parking twin for a designated area (for example, if it is necessary to schedule parking resources around a stadium, a three-level parking twin covering 3 square kilometers around the stadium can be constructed).

[0045] The dynamic and static data acquisition module collects static basic data (including micro, meso, and macro static basic data).

[0046] The micro-level static basic data includes data from the vehicle attribute library and data from the parking space attribute library. The vehicle attribute library data includes: vehicle type (small car / SUV / truck), size (length × width × height), minimum turning radius, acceleration / deceleration performance (obtained through the car manufacturer's API or real vehicle testing); the parking space attribute library data includes: parking space size (standard / non-standard), type (ordinary / charging / accessible), location coordinates (three-dimensional coordinates within the parking lot, accuracy ±5cm, generated by LiDAR scanning), and distance between adjacent parking spaces (affecting the difficulty of parking).

[0047] The static basic data at the meso level includes parking lot topology data and facility parameters. Parking lot topology data: CAD drawings + laser point cloud modeling, to obtain channel width (one-way / two-way), turning angle, entrance / exit location, slope (underground parking lot), and obstacle distribution (columns / walls); Facility parameters: charging pile location and power, and gate response time (time to open / close the gate).

[0048] The macro-level static basic data includes regional road network topology data and parking lot node attributes. Regional road network topology data: GIS map + field surveying, to obtain road length, number of lanes, speed limit, intersection signal timing (cycle / green ratio), and traffic restriction rules (such as truck restriction periods). Parking lot node attributes: total capacity, real-time available parking spaces (distinguishing between owned and shared).

[0049] Dynamic real-time data acquisition (including micro, meso, and macro dynamic real-time data).

[0050] The micro-level dynamic real-time data includes the vehicle's real-time location (latitude and longitude / park coordinates), speed, steering intention (turn signal), remaining range (EV), and user response to recommended parking spaces (accept / reject). The data is collected through V2X communication, vehicle GPS, parking lot cameras, and APP interaction logs.

[0051] The meso-level dynamic real-time data includes vehicle trajectories within the parking lot (entrance → parking space → exit) and vehicle density within the passageways (vehicles / m²). 2The data collection methods include entrance queue length (vehicles) and parking space occupancy status (free / occupied / reserved), collected via inductive loop detectors, video analytics (YOLOv8 target tracking), and parking management system.

[0052] Macro-level dynamic real-time data includes road cross-section traffic density (vehicles / km), average vehicle speed (km / h), intersection delay time (s), parking algorithm strategies (current running algorithm, switching threshold), and total parking demand in the area (vehicles / hour). The data is collected from roadside radar (RSU), floating car data (taxi / ride-hailing vehicles), traffic signal controllers, and edge node logs.

[0053] Construct micro, meso, and macro models.

[0054] The micro-model building module is used to build micro-models. The micro-model building process is as follows:

[0055] Define the state space (S): S = [real-time vehicle location (x, y), distance to destination (m), current vehicle speed (km / h), remaining range (%), distance to nearest available parking space (m), system recommended parking space ID and distance (m), recommended parking space type (charging / normal), surrounding vehicle density (vehicles / 100m)] 2 Weather impact coefficient (0-1, rainy day = 0.7), user's historical probability of accepting recommendations (%), reservation status (0 / 1), remaining lock-in time (s);

[0056] Define the action space (A): A = [Accept recommended parking space and proceed, reject recommended parking space and continue searching, park in the currently found empty parking space, abandon the current parking space and continue driving, switch target parking lot];

[0057] Define the reward function (R): R = -α × (time to reach parking space) + β × (parking space matching degree) - γ × (contribution to surrounding congestion) + δ × (user preference matching degree); where: α = 0.4; β = 0.3; γ = 0.2; δ = 0.1 (if EV users choose the charging position, then +0.1);

[0058] The microscopic model is ultimately constructed through training and iteration;

[0059] Definition of arrival time at parking space: The total time (in seconds) from when a vehicle “makes a decision (such as accepting a recommendation or selecting a parking space)” to when it “actually parks in the parking space and turns off the engine”, including travel time (the time taken to travel from the current location to the parking space) and parking operation time (such as the time taken to reverse into the parking space, which is affected by the size of the parking space and surrounding vehicles).

[0060] Definition of parking space matching degree: Quantifying the degree of fit between parking space and vehicle demand (value range 0-1), parking space matching degree = 1 - |parking space size - vehicle size| / standard size. The standard size is the "benchmark reference value" for measuring the degree of fit between parking space and vehicle. Its core function is to transform the size difference between parking space and vehicle into a comparable relative value through "normalization processing". Definition of surrounding congestion contribution: The "incremental impact" (value ≥ 0) on congestion in the surrounding local area (parking lot passage, adjacent parking spaces, and roads near the exit) after a vehicle selects the parking space.

[0061] The calculation process for the contribution of surrounding congestion is as follows: Contribution of surrounding congestion = Contribution of passage delay * Delay contribution weight + Contribution of interference from adjacent parking spaces * Interference contribution weight + Contribution of area density exceeding threshold * Overthreshold contribution weight + Contribution of exit road impact * Road impact contribution weight; The delay contribution weight can be 0.4, the interference contribution weight can be 0.2, the overthreshold contribution weight can be 0.2, and the road impact contribution weight can be 0.2.

[0062] Contribution to traffic delay = ∑(Number of subsequently affected vehicles × Delay time per vehicle × ...

[0063] (Position sensitivity coefficient); Example: Suppose a car chooses a corner parking space in a passageway, and the parking process lasts for 40 seconds. The impact on the passageway during this period is divided into two stages. Stage 1 (0-20 seconds: The vehicle reverses into the parking space, occupying 1 / 2 of the passageway width), Affected vehicles: 2 small cars (coming from the front of the passageway); Delay time per vehicle: Due to the narrowing of the passageway, small cars need to slow down and give way, resulting in a delay of 8 seconds per vehicle; Position sensitivity coefficient: Corner passageway (position sensitivity coefficient = 1.3). Delay contribution in this stage: 2 × 8 × 1.3 = 20.8; Stage 2 (20-40 seconds: After the vehicle stops, the body still protrudes 50cm, completely blocking the passageway), Affected vehicles: 3 large vehicles (coming from the back of the passageway); Delay time per vehicle: Due to the blockage of the passageway, large vehicles need to stop and wait, resulting in a delay of 15 seconds per vehicle; Position sensitivity coefficient: Corner passageway (position sensitivity coefficient = 1.3). Contribution to delays in this phase: 3 × 15 × 1.3 = 58.5; Contribution to passage delays = Phase 1 contribution + Phase 2 contribution = 20.8 + 58.5 = 79.3;

[0064] The contribution of adjacent parking spaces to interference = ∑(number of affected adjacent parking spaces × interference degree coefficient × ...

[0065] Parking space type coefficient); Example: Assume a parking space layout as follows: 1 ordinary parking space (parking space A) on the left; 1 charging parking space (parking space B) on the right; After a vehicle parks, the interference to parking space A is "moderate" (door opening angle reduced to 45°), and the interference to parking space B is "severe" (door cannot be opened at all). S1: Calculate the interference contribution of a single adjacent parking space: Parking space A (ordinary parking space): Number of adjacent parking spaces affected = 1 (parking space A only); Interference degree coefficient = 0.6 (moderate interference); Parking space type coefficient = 1.0 (ordinary parking space); Single parking space contribution = 1 × 0.6 × 1.0 = 0.6. Parking space B (charging parking space): Number of adjacent parking spaces affected = 1 (parking space B only); Interference degree coefficient = 1.0 (severe interference); Parking space type coefficient = 1.2 (charging parking space, special function, greater interference impact); Single parking space contribution = 1 × 1.0 × 1.2 = 1.2. S2: Summing up the total interference contribution of adjacent parking spaces: Interference contribution of adjacent parking spaces = contribution of parking space A + contribution of parking space B = 0.6 + 1.2 = 1.8;

[0066] Contribution of area density exceeding threshold = max(0, actual density - optimal threshold) × area × density sensitivity coefficient; Example: Actual density: current number of vehicles in the area ÷ area (unit: vehicles / 100m²) 2 For example, area A has an area of ​​500m² 2 There were 8 parked cars → actual density = 8 ÷ 5 = 1.6 cars / 100m² 2 ); Optimal threshold: "Highest efficient density in the region" fitted by historical data (open ground parking: optimal threshold = 1.2 vehicles / 100m²) 2 (per 100m) 2 When 1.2 cars are parked, the passageway is unobstructed and parking space turnover is fast; underground parking lot (narrow passageway): optimal threshold = 0.8 cars / 100m 2 (Higher density easily leads to congestion); Density sensitivity coefficient: The density sensitivity coefficient is dynamically adjusted according to the area attributes, time period, and function to ensure that the calculation results closely match the actual scenario: Time period sensitivity: The coefficient for peak hours (such as 10:00-12:00 on weekdays) = 1.5 (excessive density at this time has a greater chain reaction effect on subsequent traffic flow); The coefficient for off-peak hours = 0.8 (the impact is weaker). Area attribute sensitivity: The coefficient for the area surrounding the core passage (such as the necessary passage connecting the entrance / exit) = 1.3 (excessive density easily leads to global congestion); The coefficient for the edge area (such as corner parking spaces) = 0.9 (the impact range is limited). Function sensitivity: The coefficient for the charging pile concentration area and the accessible parking area = 1.2 (congestion in these areas will directly affect users with special needs, and the density needs to be more strictly controlled); Example: Area A of an underground parking lot (area = 600m²) 2 = 6 "100m" 2 The current actual density is 1.5 vehicles / 100m³. 2Optimal threshold = 0.8 vehicles / 100m 2 (Narrow underground parking garage access), currently peak hour (sensitivity coefficient = 1.5). Calculated excess density: max(0, 1.5 - 0.8) = 0.7 vehicles / 100m² 2 (Actual density exceeds threshold 0.7); Area is 6 (unit: 100m²) 2 Contribution of regional density exceeding the threshold: 0.7 × 6 × 1.5 (peak period coefficient) = 6.3;

[0067] The contribution of the exit route to the parking lot's impact is calculated as: Exit route delay increment + Merging delay from surrounding roads. The definition of exit route delay increment is the difference between the time it takes for a vehicle to travel from its selected parking space to the parking lot exit and the "average exit time for this area" (a positive number indicates that the exit route for this parking space takes longer than the average, occupying internal passageways for a longer period and indirectly causing subsequent vehicles to queue). The definition of merging delay from surrounding roads is the "disturbance effect" on the existing traffic flow when a vehicle enters surrounding public roads from the parking lot exit (e.g., causing following vehicles to slow down or brake, resulting in localized congestion). The calculation logic is as follows: based on real-time data from roadside radar / cameras. Traffic flow data is dynamically calculated based on road congestion levels: If the current traffic density is ≤30 vehicles / km (smooth traffic): merging vehicles can merge smoothly without delay → contribution = 0; If 30 vehicles / km < density ≤60 vehicles / km (mild congestion): each merging vehicle causes the subsequent 3 vehicles to slow down by 1 second → delay = 3 × 1 = 3 seconds; If density >60 vehicles / km (near saturation): each merging vehicle causes the subsequent 5 vehicles to slow down by 2 seconds → delay = 5 × 2 = 10 seconds; If density >100 vehicles / km (severe congestion): merging may cause a short-term blockage, delay = 20 seconds (fixed value, as traffic flow is highly vulnerable at this time). Example: Departure scenario for a parking space: Actual departure time = 4 minutes (240 seconds), average departure time in the area = 3 minutes (180 seconds) → Departure path delay increment = 60 seconds; Traffic density on the road outside the exit = 80 vehicles / km (close to saturation) → Merging delay from surrounding roads = 10 seconds. Therefore: Contribution of departure road impact = 60 + 10 = 70 (The parking space selection contributes 70 seconds to the departure road impact).

[0068] The calculation process for user preference matching degree is as follows: User preference matching degree = (parking space type matching degree × parking space type weight) + (location matching degree × location matching weight); the parking space type weight can be 0.6, and the location matching weight can be 0.4; Parking space type matching degree: the user's preference for the "core function" of the parking space (such as the rigid demand of EV users for charging parking spaces), calculate the fit between the current recommended parking space type and the user's historical choices. For example, in 100 parking times, the user chooses a charging parking space 80 times and a regular parking space 20 times → the matching degree of the charging parking space = 80%, and the matching degree of the regular parking space = 20%;

[0069] Location matching degree: Quantifies the degree of fit between recommended parking spaces and the "spatial location characteristics" of user preferences. User historical preference distance: Statistically calculates the average distance between the parking space selected by the user and the specified area (such as a stadium or shopping mall) (e.g., users average less than 100 meters from the parking space). The location matching degree is calculated as follows: If the distance of the recommended parking space = 80 meters (≤ the average user preference distance of 100 meters) → matching degree = 1.0; If the distance = 150 meters (exceeding 50 meters) → matching degree = 1 - (exceeding distance / average user preference distance) = 1 - 50 / 100 = 0.5; If the distance > 200 meters (exceeding 100%) → matching degree = 0.2 (minimum value, to avoid being completely 0).

[0070] The mesoscopic model building module is used to build mesoscopic models. The mesoscopic model building process is as follows:

[0071] Flow field topology mapping: The parking lot is divided into a "node-edge" network (similar to a "grid" in fluid mechanics): Nodes: entrance, exit, parking space groups (every 5 adjacent parking spaces form a node), turning points, and passageway intersections; Edges: passageways connecting nodes (e.g., the passageway from "entrance to parking space group A"), with attributes including length (m), width (m), one-way / two-way, and maximum throughput (vehicles / min). Fluid mechanics equation adaptation: Borrowing from the LWR (Lighthill-Whitham-Richards) traffic flow model, vehicles are treated as "compressible fluid particles," and the flow field is described using density-velocity-flow relationship: Density (k): number of vehicles per unit passage length (vehicles / m); Velocity (v): vehicle speed in the passage, satisfying the "vehicle-density" relationship (k increases → v decreases, free flow velocity = 15km / h, v → 0 during congestion); Flow rate (q): number of vehicles passing through the passage per unit time (q = k × v, limited by passage width, maximum flow rate = 30 vehicles / min / lane).

[0072] Key interaction rules: Car-following rule: The following vehicle maintains a safe distance from the vehicle in front (= current speed × 0.5s + 5m); Lane changing rule: Lane changing is allowed when the lane width is ≥ 4m, and the lane changing time is 2s (affected by the traffic flow of adjacent lanes); Queuing rule: Vehicles queuing at the entrance enter according to the "first come, first served" principle. For every additional vehicle in the queue, the arrival rate of subsequent vehicles decreases by 5% (due to vehicle diversion caused by congestion).

[0073] A mesoscopic model was built through parameter calibration and fitting training.

[0074] The macro model building module is used to build macro models. The macro model building process is as follows:

[0075] Define the graph structure (G = (V, E)): Node (V): Parking lot (number = total number of parking lots N in the area), Node characteristics (dimension = 8): Vi = [Current Utilization Rate (%), Remaining Parking Spaces, Number of Idle Charging Stations, Current Running Algorithm (Code: Greedy = 0 / Particle Swarm = 1), Algorithm Switching Timestamp (s), Average Parking Duration (min), Estimated Number of Vehicles Arriving in 1 Hour, User Satisfaction (%)]; V i This is a multi-dimensional state feature vector describing the "i-th parking lot"; the current utilization rate is defined as: (number of occupied parking spaces / total number of parking spaces) × 100%; the algorithm switching timestamp is defined as: the time when the parking lot last switched algorithms; the average parking time is defined as: the average time vehicles spend in the parking lot (from entry to exit) in the past hour; the expected number of vehicles arriving in the past hour is defined as: based on historical data and real-time reservation information, the predicted number of vehicles entering the parking lot in the next hour; user satisfaction is defined as: the average rating of the parking lot service by users in the past hour (collected through APP feedback, exit QR code evaluation, etc.).

[0076] Edge (E): Roads connecting adjacent parking lots (number = M), edge features (dimension = 5): E i,j = [Road length (m), Design speed (km / h), Current average vehicle speed (km / h), Traffic density (vehicles / km), Signal timing impact coefficient (0-1, Green wave = 0.9)]; E i,j Represents "connecting parking lot node V" i With parking lot node V j The roadside feature set is defined as follows: road length is the physical distance between parking lot i and parking lot j; design speed is the maximum speed set for the road connecting parking lot i and parking lot j; current average speed is the average speed of all vehicles on the road connecting parking lot i and parking lot j; traffic density is the number of vehicles per unit length on the road connecting parking lot i and parking lot j; traffic light timing influence coefficient is the impact of traffic light timing at intersections along the road between parking lot i and parking lot j on traffic efficiency (e.g., "green wave" allows vehicles to pass continuously, resulting in a high coefficient; chaotic traffic light timing leads to frequent stops, resulting in a low coefficient). GraphSAGE is used to achieve the interaction between nodes and edges through "sampling neighbors - aggregating features - updating node embeddings," ultimately building a macroscopic model.

[0077] The three-level parking twin construction module constructs a three-level parking twin by transferring data and scaling between micro, meso, and macro models (micro → meso: summarizing the "parking location" and "driving trajectory" of a single vehicle into the "parking space occupancy status" and "traffic density" of the meso model; meso → macro: using the "utilization rate" and "queue time" of each parking lot as "node features" of the macro model to affect the distribution of road traffic flow; macro → micro: transforming macro strategies (such as "prioritizing guiding vehicles to parking lot B") into the "recommended parking space weight adjustment" of the micro model (increasing the recommendation priority of parking lot B)).

[0078] The scheduling involves a strategy selection unit. Based on a three-level parking twin periodically loading physical model, an initial algorithm library suitable for the current scenario is constructed simultaneously. This initial algorithm library is built based on expert experience and rule extraction, and historical data mining. Expert experience and rule extraction are based on industry experience in parking scheduling, with a pre-defined basic algorithm framework. For example: Off-peak strategy: When vehicle density is <20 vehicles / km², a greedy algorithm (prioritizing response speed) is run; Peak-peak strategy: When vehicle density is >50 vehicles / km², the particle swarm optimization algorithm (prioritizing load balancing) is switched; Special scenario strategy: During heavy rain, a "robust mode" is automatically activated (allowing ±50 meters of recommended error, but ensuring 100% no crashes). Historical data mining: through... Analyze the scheduling logs from the past three months to extract "highly effective algorithms". For example, data shows that "switching to the particle swarm algorithm in the business district between 6:00 PM and 7:00 PM on Friday nights increases parking space utilization by 20%". This time period and threshold are then fixed as the "Friday business district exclusive algorithm" in the algorithm library. Randomly combine the algorithms in the initial algorithm library to generate multiple algorithm switching strategies (e.g., the initial algorithm library contains greedy algorithm, hybrid algorithm, and particle swarm algorithm, and the total duration of one cycle is 2 hours. Algorithm switching strategy 1 could be to use the greedy algorithm for the first hour and the hybrid algorithm for the last hour, while algorithm switching strategy 2 could be to use the greedy algorithm for the first 30 minutes, the hybrid algorithm for the middle 30 minutes, and the particle swarm algorithm for the last hour).

[0079] Analyze the parking resource scheduling values ​​of the three-level parking twin under different algorithm switching strategies, and mark the algorithm switching strategy with the largest parking resource scheduling value as the scheduling selection algorithm switching strategy. Use the scheduling selection algorithm switching strategy to schedule parking resources in a specified area (and distribute the scheduling selection algorithm to the edge node or cloud system).

[0080] Parking resource scheduling value of a three-level parking twin under an algorithm switching strategy: Determine an algorithm switching strategy, control the three-level parking twin to execute the algorithm switching strategy for a one-cycle simulation scheduling, after the simulation ends, obtain the micro-scheduling index, meso-scheduling index and macro-scheduling index of the three-level parking twin, sum the micro-scheduling index, meso-scheduling index and macro-scheduling index to calculate the parking resource scheduling value of the three-level parking twin under the algorithm switching strategy.

[0081] Steps to obtain the micro-scheduling index: Obtain the average search time (Ave) of all vehicles in the three-level parking twin. allcar The acceptance rate of recommended parking spaces among all users. accept And EV charging station matching success rate Suc ev ,pass Calculate the micro scheduling index micro sche (Dimensionless calculation), where "reference time" is the set ideal / standard average parking time, which needs to refer to historical data, p1 is the parking time coefficient, p2 is the recommended parking space acceptance coefficient, and p3 is the charging space matching success coefficient. p1 can be 0.4, p2 can be 0.3, and p3 can be 0.3. p1, p2, and p3 are empirical values ​​and need to be dynamically optimized in combination with the scenario.

[0082] Steps to obtain the meso-level scheduling index: Obtain the average parking space utilization rate of all parking lots in the three-level parking twin. park Average queue time at all parking lot entrances (lineup) time ,pass Calculate the mid-level dispatch index Mid sche (Dimensionless calculation), Tr can be the "reasonable queuing time" recommended by the "Urban Parking Lot Planning Code" (e.g., 5 minutes), p4 is the parking space utilization coefficient, p5 is the queuing time coefficient, p4 can be 0.6, p5 can be 0.4, p4 and p5 are empirical values ​​and need to be dynamically optimized in combination with the scenario.

[0083] Steps to obtain the macro-dispatch index: Obtain the average vehicle speed (speed) of all roads in the Level 3 parking twin. ave Average load difference across all parking lots park (Average load difference: the arithmetic mean of the single-park load differences of all parking lots (reflecting the overall unevenness of parking lot load in the area), single-park load difference: |Single-park load rate - Average load rate| (e.g., if the parking lot load rate is 80%, the average is 70%, and the load difference is 10%), average load rate: "Total number of occupied parking spaces / total number of parking spaces" of all parking lots in the area (e.g., if there are 3 parking lots with a total of 300 parking spaces, and 210 are occupied, the average load rate is 70%)). Calculate the macro scheduling index Macro sche (Dimensionless calculation), p6 is the average vehicle speed coefficient, p7 is the load difference coefficient, speed ref Reference vehicle speed (e.g., road design speed 60km / h); load ref : Reference load difference (e.g., historical maximum load difference of 0.3); p6 can be 0.5, p7 can be 0.5, p6 and p7 are empirical values ​​and need to be dynamically optimized according to the scenario.

[0084] The aforementioned system constructs a three-tiered parking twin—micro, meso, and macro—and integrates dynamic and static multi-source data to achieve full-scale digital mapping from individual vehicle behavior and parking lot flow fields to regional traffic networks. The scale transformation and data interaction between the three-tiered models accurately capture the dynamic characteristics of parking resources in spatial and temporal dimensions, providing a high-fidelity, fine-grained digital twin foundation for scheduling decisions. This breaks down the disconnect between "local optimization and global collaboration" in traditional scheduling, supporting the accuracy and foresight of resource allocation. Based on the dynamic construction of an algorithm strategy library and twin simulation evaluation, the solution uses multi-dimensional scheduling values ​​as its core, achieving closed-loop optimization of "algorithm combination - effect pre-performance - strategy optimization - dynamic adjustment." The algorithm switching strategy adapts to the characteristics of different time periods and scenarios, balancing multiple objectives such as efficiency, uniformity, and user experience. Continuous periodic simulation and indicator feedback allow the scheduling strategy to dynamically evolve with the physical world, effectively improving parking resource utilization efficiency, regional traffic flow, and user service quality.

[0085] The above formulas are all dimensionless calculations, and the preset parameters in the formulas should be set by those skilled in the art according to the actual situation.

[0086] 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, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0087] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0088] Those skilled in the art will recognize that the units 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 design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0089] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0090] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0091] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0092] 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.

Claims

1. A parking resource intelligent scheduling system based on multi-source data analysis, characterized in that, include The static and dynamic data acquisition module is used to collect static basic data and dynamic real-time data; The micro-model building module is used to build micro-models; The meso-level model building module is used to build meso-level models; The macro model building module is used to build macro models; The three-level parking twin construction module constructs a three-level parking twin for a specified area through data transfer and scale transformation between micro, meso, and macro models. The scheduling involves a strategy selection unit. Based on the three-level parking twin, a physical model is periodically loaded, and an initial algorithm library that conforms to the current scenario is constructed simultaneously. The algorithms in the initial algorithm library are randomly combined to generate multiple algorithm switching strategies. The scheduling selection algorithm switching strategy is selected from the algorithm switching strategies, and the scheduling selection algorithm switching strategy is used to schedule parking resources in the specified area.

2. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 1, characterized in that, Static basic data includes micro-level static basic data, meso-level static basic data, and macro-level static basic data, while dynamic real-time data includes micro-level dynamic real-time data, meso-level dynamic real-time data, and macro-level dynamic real-time data.

3. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 2, characterized in that, The micro-level static basic data includes data from the vehicle attribute database and data from the parking space attribute database; The static basic data at the meso level includes parking lot topology data and facility parameters; The macro-level static basic data includes regional road network topology data and parking lot node attributes.

4. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 2, characterized in that, Microscopic dynamic real-time data includes vehicle real-time location, speed, steering intention, remaining range, and user response to recommended parking spaces; The dynamic real-time data at the meso level includes vehicle trajectories within the parking lot, vehicle density within the passageways, queue length at the entrance, and parking space occupancy status. Macro-level dynamic real-time data includes road cross-section traffic density, average vehicle speed, intersection delay time, parking algorithm strategies, and total parking demand in the area.

5. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 1, characterized in that, The process of building a microscopic model: Define the state space (S): S = [real-time vehicle location, distance to destination, current vehicle speed, remaining range, distance to the nearest available parking space, system recommended parking space ID and distance, recommended parking space type, surrounding vehicle density, weather impact coefficient, user's historical probability of accepting recommendations, reservation status, remaining lock time]; Define the action space (A): A = [Accept recommended parking space and proceed, reject recommended parking space and continue searching, park in the currently found empty parking space, abandon the current parking space and continue driving, switch target parking lot]; Define the reward function (R): R = -α × (time to reach parking space) + β × (parking space matching degree) - γ × (contribution to surrounding congestion) + δ × (user preference matching degree); A microscopic model is eventually built through training and iteration.

6. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 5, characterized in that, The calculation process for the contribution of surrounding congestion is as follows: Contribution of surrounding congestion = Contribution of passage delay * Delay contribution weight + Contribution of interference from adjacent parking spaces * Interference contribution weight + Contribution of area density exceeding threshold * Exceeding threshold contribution weight + Contribution of departure road impact * Road impact contribution weight.

7. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 6, characterized in that, Contribution of corridor passage delay = ∑(Number of subsequently affected vehicles × Delay time per vehicle × Location sensitivity coefficient); The contribution of adjacent parking space interference = ∑(number of adjacent parking spaces affected × interference degree coefficient × parking space type coefficient); Contribution of regional density exceeding threshold = max(0, actual density - optimal threshold) × region area × density sensitivity coefficient; The contribution of departure roads to the overall impact = the increase in departure path delay + the merging delay of surrounding roads.

8. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 5, characterized in that, The calculation process for user preference matching degree is as follows: User preference matching degree = (parking space type matching degree × parking space type weight) + (location matching degree × location matching weight).

9. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 1, characterized in that, Select the scheduling algorithm switching strategy from the algorithm switching strategies: Analyze the parking resource scheduling values ​​of the three-level parking twin under different algorithm switching strategies, and mark the algorithm switching strategy with the largest parking resource scheduling value as the scheduling algorithm switching strategy.

10. The intelligent parking resource scheduling system based on multi-source data analysis according to claim 1, characterized in that, Parking resource scheduling value of a three-level parking twin under an algorithm switching strategy: Determine an algorithm switching strategy, control the three-level parking twin to execute the algorithm switching strategy for a period of simulated scheduling, after the simulation ends, obtain the micro-scheduling index, meso-scheduling index and macro-scheduling index of the three-level parking twin, sum the micro-scheduling index, meso-scheduling index and macro-scheduling index to calculate the parking resource scheduling value of the three-level parking twin under the algorithm switching strategy; The steps to obtain the micro-scheduling index are as follows: obtain the average parking time of all vehicles in the three-level parking twin, the acceptance rate of all users for recommended parking spaces, and the EV charging space matching success rate. The micro-scheduling index is calculated based on the average parking time, the acceptance rate of users for recommended parking spaces, and the EV charging space matching success rate. The steps to obtain the meso-level scheduling index are as follows: obtain the average parking space utilization rate of all parking lots and the average queuing time of all parking lot entrances in the three-level parking twin, and calculate the meso-level scheduling index based on the average parking space utilization rate and the average queuing time. The steps to obtain the macro-dispatch index are as follows: obtain the average vehicle speed of all roads and the average load difference of all parking lots in the three-level parking twin, and calculate the macro-dispatch index based on the average vehicle speed and the average load difference.

Citation Information

Cited By

  • City-level parking resource optimization method and system

    CN121811688A

  • A city-level parking resource optimization method and system

    CN121811688B

  • Macro-micro traffic regulation method and device based on optimization solver, electronic equipment and storage medium

    CN122369272A