A parking management method and system based on a full-life-cycle five-dimensional digital twin

Abstract

The application discloses a parking management method and system based on a full-life-cycle five-dimensional digital twin. The method constructs a five-dimensional digital twin model integrating geometric dimension, physical dimension, behavior dimension, rule dimension and data dimension, and realizes full-life-cycle management based on a time axis closed-loop architecture of past state-present state-future state. The past state generates a standardized knowledge template library through historical data mining and intelligently matches to generate a project proposal; the present state realizes millisecond-level synchronization of physical change information through mobile terminal and cloud cooperation, and constructs accurate mapping from design state to built state; the future state uses a PPO reinforcement learning algorithm to perform super-real-time simulation deduction in a virtual space, and outputs a multi-objective optimization scheduling strategy. The application realizes cross-stage knowledge sedimentation and strategy self-evolution, and significantly improves the scientificity and operation efficiency of parking lot planning.

Description

Technical Field

[0001] This application relates to the fields of digital twin technology and intelligent transportation technology, and in particular to a parking management method and system based on a five-dimensional digital twin throughout the entire life cycle. Background Technology

[0002] With the continuous growth of urban vehicle ownership, parking difficulties have become a significant issue affecting urban traffic efficiency and residents' quality of life. Statistics show that drivers in city centers spend an average of more than fifteen minutes searching for a parking space. This process not only wastes time but also leads to additional fuel consumption and exhaust emissions, exacerbating urban traffic congestion and environmental pollution. Smart parking systems, by integrating IoT sensing, cloud computing, and mobile internet technologies, can improve the utilization efficiency of parking resources to a certain extent, enabling functions such as real-time parking space status monitoring, parking guidance, and electronic payment.

[0003] However, existing smart parking technologies still have the following shortcomings:

[0004] First, there is a disconnect in lifecycle management. Current smart parking systems typically use Building Information Modeling (BIM) tools to generate 3D models of parking lots during the design phase for spatial planning and equipment layout. However, this model is difficult to maintain and update continuously during the construction and operation phases. When adjustments to equipment positions or parameters are needed due to site constraints during construction, this change information is often communicated via paper documents or verbally, failing to be automatically reflected in the digital model. This leads to a gradual discrepancy between the model created in the design phase and the actual physical entity, creating a data gap between the design and built states. When analysis or optimization based on the digital model is required during the operation phase, outdated model data will lead to inaccurate decision-making.

[0005] Second, there is a lack of forward-looking predictive capabilities. Most existing smart parking systems adopt a passive, reactive management model, triggering appropriate procedures only after detecting full parking spaces, queues at entrances, or equipment malfunctions. Although some systems have introduced simple rule engines or historical statistical analysis methods, limitations in model accuracy and computing power make it difficult to accurately predict changes in the scenario over the next few minutes or even hours. For example, the system cannot anticipate upcoming peak traffic and proactively adjust its scheduling strategy; it can only react passively after congestion actually occurs, missing the optimal intervention opportunity.

[0006] Third, the value of historical data has not been fully realized. Smart parking systems accumulate a wealth of vehicle entry and exit records, equipment operation logs, user behavior data, and revenue reports during operation. This data contains rich operational patterns and optimization experience. However, in existing technological solutions, this data is typically only used for generating post-event statistical reports and evaluating operational performance, lacking effective mechanisms for knowledge extraction and experience reuse. When planning and designing new parking projects, designers still need to rely on personal experience and industry practices, unable to systematically learn from design solutions and operational strategies validated in historical projects. This leads to similar design flaws and operational problems recurring in different projects.

[0007] Fourth, economic efficiency is difficult to quantify and predict. As commercial infrastructure, the return on investment and operating revenue of parking lots are important considerations in project decisions. However, existing planning and design methods mainly focus on technical feasibility, i.e., ensuring that vehicles can park and exit, while lacking the ability to quantitatively analyze the impact of different design schemes on operating revenue. For example, how design parameters such as parking space layout, lane width, and gate location affect vehicle turnover rate and parking lot revenue lacks calculable evaluation models, leading to design decisions relying on subjective judgment.

[0008] Therefore, there is an urgent need for an intelligent parking management technology solution that can connect the entire lifecycle of planning, construction, and operation, and realize the accumulation of historical experience, real-time status mirroring, and future trend prediction, so as to systematically solve the above-mentioned technical problems. Summary of the Invention

[0009] Based on this, the embodiments of this application provide a parking management method and system based on a five-dimensional digital twin throughout the entire life cycle, in order to solve the technical problems in the prior art such as the data gap between the design state and the operation state, the lack of forward-looking prediction capabilities, the difficulty in reusing historical experience, and the difficulty in quantifying and predicting economic efficiency.

[0010] Firstly, a parking management method based on a five-dimensional digital twin throughout the entire lifecycle is provided, which includes:

[0011] S1. Construct a five-dimensional digital twin model; wherein, the five-dimensional digital twin model includes a geometric model, a physical model, a behavioral model, a rule model, and a real-time data model. The geometric model is used to construct a three-dimensional spatial mapping of the parking lot's physical environment. The physical model is used to define the physical attributes and kinematic characteristics of the equipment. The behavioral model is used to predict the dynamic behavioral trajectories of vehicles and pedestrians. The rule model is used to encode business logic and safety constraints. The real-time data model is used to map the state data flow of the physical world.

[0012] S2. Extract historical data of design parameters, operation data and performance indicators from at least one operational parking lot, generate a standardized knowledge template library through data mining, and use a similarity matching algorithm to match the most similar template in the knowledge template library based on the input parameters of the target project to generate a target project proposal that includes a bill of materials, schedule plan and performance prediction.

[0013] S3. Load the target project proposal into a five-dimensional digital twin model, perform multi-dimensional simulation in virtual space at a speed higher than the physical time flow rate, and input a simulation verification report; wherein, the simulation includes at least spatial layout verification, performance load assessment and fault scenario drills;

[0014] S4. During the project deployment phase, change information of the physical site relative to the original design is collected through mobile terminals, and the change information is synchronized to the five-dimensional digital twin model in real time to realize the mapping update from the design state to the completed state; wherein, the change information includes at least equipment location adjustment, model replacement and parameter modification;

[0015] S5. During the operation phase, continuously collect physical terminal operation data, input the operation data into the five-dimensional digital twin model for strategy optimization training, distribute the optimized scheduling strategy to the physical execution terminal, and send the execution feedback data of the physical terminal back to the knowledge template library to achieve closed-loop iteration throughout the entire life cycle.

[0016] Optionally, in constructing the five-dimensional digital twin model in S1, the construction of the behavioral model specifically includes:

[0017] Historical operational data, including vehicle entry time, dwell time, driving trajectory, and pedestrian movement path, were collected as training samples.

[0018] A time-series prediction model based on a Long Short-Term Memory (LSTM) network is constructed. The input layer of the LSTM network receives timestamps, entry rates, and area occupancy features. The hidden layer contains at least two LSTM units, and the output layer generates predicted values ​​for vehicle arrival rate and pedestrian density distribution within a preset future time window.

[0019] The trained time-series prediction model is embedded into the behavioral model layer of the five-dimensional digital twin model.

[0020] Optionally, in step S2, historical data on design parameters, operational data, and performance indicators are extracted from at least one operational parking lot, and a standardized knowledge template library is generated through data mining, specifically including:

[0021] Analyze the design documents of deployed projects to extract parking lot size, number of parking spaces, equipment configuration, and spatial layout parameters;

[0022] Collect operational logs from deployed projects and extract indicators such as equipment failure rate, maintenance cycle, peak period distribution, and average turnover rate.

[0023] The design parameters and operating indicators are normalized to generate multidimensional feature vectors;

[0024] Clustering algorithms are used to divide multiple historical projects into several categories based on feature similarity. A standardized knowledge template is generated for each category, which includes typical configuration schemes, performance benchmarks, and optimization suggestions.

[0025] Optionally, the S2 step generates a target project proposal that includes a bill of materials, a schedule, and performance forecasts, and also includes economic performance forecasts, specifically including:

[0026] Based on the behavior model in the five-dimensional digital twin model, the vehicle passage trajectory and dwell time under different parking space layout schemes are simulated;

[0027] Based on the preset operational strategy parameters, the time-space turnover rate index corresponding to each plan is calculated. The time-space turnover rate is defined as the number of effective services per parking space per unit time.

[0028] Based on the spatiotemporal turnover rate index, the system outputs parking space layout optimization suggestions and dynamic pricing strategies that take into account both traffic efficiency and economic returns.

[0029] Optionally, the simulation in S3 includes spatial layout verification, performance load assessment, and fault scenario drills, wherein:

[0030] The specific fault scenario drills include: injecting preset fault events into a five-dimensional digital twin model, wherein the fault events include at least one of the following: gate failure, parking space detector offline, network communication interruption, and sudden large passenger flow; simulating the system response process after the fault event occurs, recording vehicle queue length, evacuation time, and the activation status of backup strategies; evaluating the effectiveness of existing emergency plans based on simulation results, and outputting a fault response capability assessment report and plan optimization suggestions;

[0031] The spatial layout verification specifically includes: generating a three-dimensional spatial model in the virtual environment of a five-dimensional digital twin model according to the parking space layout scheme in the target project proposal; simulating the entry and exit processes of different vehicle types to check whether the turning radius of the vehicles meets the spatial constraints and whether the distance between adjacent parking spaces meets the safety standards; simulating the passage of fire lanes and pedestrian walkways to verify the unobstructedness of emergency evacuation routes; outputting a spatial layout verification report, marking areas with conflicts or non-compliance with specifications, and generating optimization and adjustment suggestions;

[0032] The performance load assessment specifically includes: setting different levels of traffic flow input conditions in a five-dimensional digital twin model to simulate four scenarios: low peak, off-peak, peak, and extreme peak; recording entrance queue length, average waiting time, parking space utilization rate, and exit capacity indicators under each scenario; identifying system bottlenecks and assessing whether the number of gate barriers, channel width, and guidance signs meet the design capacity requirements; and outputting a performance load assessment report, including quantitative indicators and expansion suggestions for each scenario.

[0033] Optionally, step S4 involves collecting information on changes in the physical site relative to the original design via a mobile terminal, specifically including:

[0034] The mobile terminal runs a change collection application, which supports obtaining device change information through photo recognition, manual input, or QR code scanning.

[0035] The change information is processed in a structured manner to generate a change record, which includes the change object identifier, change type, parameter value before change, parameter value after change, and change timestamp.

[0036] The change record is uploaded to the cloud-based digital twin data center via a network interface, triggering attribute updates for the corresponding geometric and physical models in the five-dimensional digital twin model;

[0037] After the update is completed, the view coverage, traffic path planning, and performance parameters of the affected area are recalculated to ensure the consistency between the digital twin model and the physical entity.

[0038] Optionally, the policy optimization training in S5 employs the Proximal Policy Optimization (PPO) algorithm, specifically including:

[0039] Multiple future scenarios are instantiated in parallel within the virtual environment of a five-dimensional digital twin model, each scenario corresponding to different initial conditions and random perturbations;

[0040] Define a multi-objective reward function R, wherein the reward function R = α·R_efficiency + β·R_energy + γ·R_revenue, where R_efficiency represents the negative value of the average travel time of vehicles, R_energy represents the negative value of equipment energy consumption, R_revenue represents the spatiotemporal turnover rate, and α, β, and γ are preset weight coefficients;

[0041] The PPO algorithm is used to iteratively optimize the policy network parameters in multiple parallel scenarios until the cumulative reward converges.

[0042] The converged policy network output is used as the optimal scheduling policy, which is then sent to the physical execution end after passing virtual verification.

[0043] Optionally, the method further includes a dual-mode parking space detection step, specifically including:

[0044] Each parking space is equipped with a video detection unit and a geomagnetic detection unit to independently detect the parking space occupancy status.

[0045] The detection results of the two detection units are input into the fusion judgment module, and the final parking space status is determined by a weighted voting algorithm. The weight of the video detection unit is dynamically adjusted based on the lighting conditions, and the weight of the geomagnetic detection unit is dynamically adjusted based on the degree of electromagnetic interference.

[0046] When the detection results of two detection units are inconsistent, an abnormal alarm is triggered and recorded in the operation and maintenance log;

[0047] The final parking space status is synchronized in real time to the real-time data model of the five-dimensional digital twin model.

[0048] Optionally, the method further includes a dynamic switching step for tidal parking spaces, specifically including:

[0049] Based on the aforementioned behavioral model, predict the inbound and outbound vehicle traffic within a future preset time window;

[0050] When the ratio of the predicted inbound traffic flow to the outbound traffic flow exceeds a preset threshold, the system is determined to enter tidal mode.

[0051] In tidal mode, a preset number of bidirectional lanes will be switched to one-way entry or one-way exit lanes, and vehicles will be guided through the lanes by variable message signs and ground projection indicators.

[0052] The tidal switching records are synchronized to the five-dimensional digital twin model for subsequent strategy optimization training.

[0053] Secondly, a parking management system based on a five-dimensional digital twin throughout the entire lifecycle is provided, the system comprising:

[0054] A construction module is used to build a five-dimensional digital twin model. The five-dimensional digital twin model includes a geometric model, a physical model, a behavioral model, a rule model, and a real-time data model. The geometric model is used to construct a three-dimensional spatial mapping of the parking lot's physical environment. The physical model is used to define the physical attributes and kinematic characteristics of the equipment. The behavioral model is used to predict the dynamic behavioral trajectories of vehicles and pedestrians. The rule model is used to encode business logic and safety constraints. The real-time data model is used to map the state data flow of the physical world.

[0055] The past knowledge accumulation module is used to extract historical data of design parameters, operation data and performance indicators from at least one operational parking lot. It generates a standardized knowledge template library through data mining, and uses a similarity matching algorithm to match the most similar template in the knowledge template library based on the input parameters of the target project, generating a target project proposal that includes a bill of materials, schedule plan and performance prediction.

[0056] The future-oriented ultra-real-time simulation module is used to load the target project proposal into a five-dimensional digital twin model, perform multi-dimensional simulation and deduction in virtual space at a speed higher than the physical time flow rate, and input a simulation verification report; wherein, the simulation and deduction includes at least spatial layout verification, performance load assessment and fault scenario drills;

[0057] The current virtual-real synchronization module is used during the project deployment phase to collect change information of the physical site relative to the original design via a mobile terminal, and synchronize the change information to the five-dimensional digital twin model in real time to realize the mapping update from the design state to the completed state; wherein, the change information includes at least equipment location adjustment, model replacement and parameter modification;

[0058] The virtual-real fusion continuous evolution module is used to continuously collect physical terminal operation data during the operation phase, input the operation data into the five-dimensional digital twin model for strategy optimization training, distribute the optimized scheduling strategy to the physical execution terminal, and send the execution feedback data of the physical terminal back to the knowledge template library to realize closed-loop iteration throughout the entire life cycle.

[0059] The beneficial effects of the technical solutions provided in this application include at least the following:

[0060] (1) By constructing a three-state integrated closed-loop architecture of past, present, and future states, the data connection of the entire life cycle of parking lots from planning and design to construction and deployment and operation optimization is realized, which solves the problem of the disconnect between the design state model and the actual operation data in the traditional system. The data generated at each stage can be continuously accumulated and transferred in a unified digital twin model to form an iteratively evolving knowledge asset.

[0061] (2) By constructing a digital twin model that integrates five dimensions—geometry, physics, behavior, rules, and data—the system is equipped with the capabilities of computation, simulation, and reasoning, breaking through the limitations of traditional BIM models that only have visualization functions. The five-dimensional model can not only present the appearance of the parking lot, but also simulate the operating characteristics of equipment, predict vehicle behavior patterns, and execute business rule logic, providing a complete virtual simulation environment for intelligent decision-making.

[0062] (3) By conducting real-time simulations in virtual space, the system can predict congestion trends and potential failure risks minutes to hours in advance, thus realizing a shift from a passive response to an active prevention management model. The simulations are executed faster than in real time in the virtual environment, enabling the system to evaluate and select multiple response strategies before physical events actually occur.

[0063] (4) By introducing the PPO reinforcement learning algorithm for multi-objective policy optimization, the system can automatically output a scheduling strategy that takes into account efficiency, energy consumption, and benefits, avoiding the subjectivity and limitations of relying on human experience. The reinforcement learning algorithm autonomously explores the optimal strategy through a large number of simulation experiments in a virtual environment, and can discover optimization space that is difficult for human experts to detect.

[0064] (5) By automatically collecting physical site change information through mobile terminals and synchronizing it to the digital twin model in real time, data silos between the design state and the built state are eliminated, ensuring that the digital model always remains consistent with the physical entity. This virtual-physical synchronization capability enables subsequent analysis, optimization, and decision-making to be based on data that accurately reflects the actual situation.

[0065] (6) By basing economic efficiency forecasting on quantifiable physical indicators, the system achieves a technical assessment of the commercial value of design schemes. The system does not rely on subjective experience to estimate revenue, but rather derives economic efficiency by calculating physical indicators such as time and space turnover rate and traffic capacity, thus providing an objective basis for comparing different design schemes. Attached Figure Description

[0066] To more clearly illustrate the embodiments of this application or the technical solutions in 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 merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0067] Figure 1 This is a schematic diagram of the overall system architecture of this application;

[0068] Figure 2 This is a schematic diagram of the hierarchical structure of the five-dimensional digital twin model of this application;

[0069] Figure 3 This is a timeline closed-loop flowchart of the three-state fusion of this application;

[0070] Figure 4 This is a flowchart of the training and deployment process for reinforcement learning in this PPO application;

[0071] Figure 5 This is a flowchart of the mobile terminal change collection and synchronization process for this application;

[0072] Figure 6 This is an example diagram of a scenario from the real-time simulation of this application;

[0073] Figure 7 This is a flowchart of the knowledge template generation and matching process for this application;

[0074] Figure 8 This is the flowchart of the dual-mode parking space detection fusion judgment process in this application. Detailed Implementation

[0075] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0076] In the description of this application, the terms include, have, and any variations are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those steps or units that are expressly listed, but may also include other steps or units that are not expressly listed but are inherent to these processes, methods, products, or devices, or steps or units added based on further optimizations conceived in this application.

[0077] The purpose of this application is to provide a method and system for intelligent parking lifecycle management based on five-dimensional digital twin and three-state fusion, so as to solve the technical problems in the prior art, such as the data gap between the design state and the operation state, lack of forward-looking prediction ability, difficulty in reusing historical experience, and difficulty in quantifying and predicting economic efficiency.

[0078] To achieve the above objectives, this application adopts the following technical solution:

[0079] This application provides a method for intelligent parking lifecycle management based on five-dimensional digital twin and three-state fusion, including the following steps:

[0080] S1. Construct a five-dimensional digital twin model: The five-dimensional digital twin model includes a geometric model, a physical model, a behavioral model, a rule model, and a real-time data model. The geometric model is used to construct a three-dimensional spatial mapping of the parking lot's physical environment. The physical model is used to define the physical attributes and kinematic characteristics of the equipment. The behavioral model is used to predict the dynamic behavioral trajectories of vehicles and pedestrians. The rule model is used to encode business logic and safety constraints. The real-time data model is used to map the state data flow of the physical world.

[0081] S2. Construct a past knowledge accumulation module: Extract historical data containing design parameters, operational data, and performance indicators from at least one deployed smart parking solution, and generate a standardized knowledge template library through data mining; Based on the input parameters of the target project, use a similarity matching algorithm to retrieve historical success cases from the knowledge template library, and generate a target project proposal containing a bill of materials, schedule, and performance prediction.

[0082] S3. Construct a future-oriented ultra-real-time simulation module: Import the target project proposal into a five-dimensional digital twin model, and perform multi-dimensional simulation in the virtual space at a speed higher than the physical time flow rate. The simulation includes spatial layout verification, performance load assessment and fault scenario drills, and outputs a simulation verification report.

[0083] S4. Construct a current virtual-real synchronization module: During the project deployment phase, collect change information of the physical site relative to the original design through mobile terminals. The change information includes at least one of equipment location adjustment, model replacement and parameter modification. The change information is synchronized to the five-dimensional digital twin model in real time to realize the mapping update from the design state to the built state.

[0084] S5. Continuous evolution of virtual-real integration: During the operation phase, continuously collect physical terminal operation data, input the operation data into the five-dimensional digital twin model for strategy optimization training, distribute the optimized scheduling strategy to the physical execution terminal, and send the execution feedback data of the physical terminal back to the knowledge template library to achieve closed-loop iteration throughout the entire life cycle.

[0085] The performance forecasts in this application include performance performance forecasts based on physical simulation (such as spatiotemporal turnover rate and equipment load) and economic performance forecasts based on performance indicators and operational strategies (such as revenue and return on investment).

[0086] This application provides an intelligent parking lifecycle management system based on five-dimensional digital twin and three-state fusion, including:

[0087] The 5D digital twin model building module is used to build a 5D digital twin model that includes a geometric model, a physical model, a behavioral model, a rule model, and a real-time data model, and to maintain the model's full lifecycle data.

[0088] Past knowledge template generation module: used to extract design parameters and operational indicators from historical project data, generate a standardized knowledge template library, and generate target project proposals based on new project parameters using a similarity matching algorithm.

[0089] Future-oriented ultra-real-time simulation module: used to perform ultra-real-time simulation in the virtual environment of a five-dimensional digital twin model, to verify spatial layout, evaluate performance load, and conduct fault scenario drills.

[0090] The current virtual-real synchronization module is used to receive physical site change information collected by the mobile terminal and synchronize the change information to the five-dimensional digital twin model to achieve real-time consistency between the physical entity and the digital image.

[0091] Hybrid Intelligent Decision Engine: Used to integrate PPO reinforcement learning algorithms and heuristic algorithms to perform dynamic scheduling decisions and predictive maintenance strategy generation based on bidirectional virtual and real data streams for multi-objective optimization.

[0092] Data communication module: Used to enable data interaction between modules and with physical devices.

[0093] like Figure 1 The diagram illustrates the overall architecture of the intelligent parking lifecycle management system proposed in this application, employing a four-layer structure. The top layer is the cloud platform layer, which includes a five-dimensional digital twin model construction module, a past-state knowledge template generation module, a future-state ultra-real-time simulation module, and a hybrid intelligent decision engine, deployed in a public or private cloud data center. The second layer is the edge computing layer, containing some functions of the present-state virtual-real synchronization module and a data communication module, deployed in the parking lot's local computer room. The third layer is the physical device layer, containing physical devices such as parking space detectors, barrier gate controllers, guidance screens, cameras, environmental sensors, and payment terminals. The fourth layer is the mobile terminal layer, containing mobile devices used by construction and maintenance personnel and the change collection application running on them. Interconnection between layers is achieved through data communication modules, with arrows indicating the data flow.

[0094] Specifically, the intelligent parking lifecycle management system of this application includes four layers: cloud platform, edge computing nodes, physical devices and mobile terminals.

[0095] The cloud platform is deployed in a public or private cloud data center, housing a five-dimensional digital twin model building module, a past-state knowledge template generation module, a future-state ultra-real-time simulation module, and a hybrid intelligent decision engine. The cloud platform possesses elastic computing capabilities, dynamically expanding resources based on the computational load of the simulation task.

[0096] Edge computing nodes are deployed in the parking lot's local data center, carrying some functions of the current virtual-physical synchronization module and the data communication module. Edge nodes are responsible for receiving real-time data from physical devices, performing local preprocessing and caching, and uploading the processed data to the cloud. Edge nodes are also responsible for receiving scheduling policies from the cloud and converting them into device control commands to be sent to the physical devices.

[0097] The physical devices include parking space detectors, barrier gate controllers, guidance screens, cameras, environmental sensors, and payment terminals. These devices connect to edge computing nodes via wired or wireless networks, reporting status data in real time and receiving control commands.

[0098] Mobile terminals are smartphones or tablets equipped for construction and maintenance personnel, running change collection applications to collect change information from the physical site during the construction and operation phases.

[0099] Interoperability between different layers is achieved through a data communication module. The data communication module supports multiple communication protocols such as MQTT, HTTP, and WebSocket to ensure data transmission requirements in different scenarios.

[0100] Example 2: Construction of a Five-Dimensional Digital Twin Model

[0101] Figure 2 This demonstrates the hierarchical structure of the five-dimensional digital twin model. At the center is a unified object identification layer, where each physical entity (parking space, equipment, facility) corresponds to a unique identifier. Extending outwards from this center are five dimensional layers: a geometric model layer (containing 3D spatial coordinates, shape, size, appearance, and material), a physical model layer (containing equipment parameters, physical properties, and motion characteristics), a behavior model layer (containing vehicle behavior prediction, pedestrian behavior prediction, and trajectory patterns), a rule model layer (containing toll rules, passage rules, safety rules, and emergency rules), and a real-time data model layer (containing sensor data streams, state time-series data, and event records). The layers are connected by dashed lines to indicate data relationships.

[0102] Specifically, the five-dimensional digital twin model constructed in this application surpasses the traditional three-dimensional visualization model, and includes five interrelated layers: geometric dimension, physical dimension, behavioral dimension, rule dimension and data dimension.

[0103] The geometric model layer constructs a three-dimensional spatial mapping of the parking lot's physical environment. Data for the geometric model comes from two sources: first, importing BIM model files created during the design phase, supporting IFC and RVT formats; second, acquiring three-dimensional point cloud data from existing parking lots through laser point cloud scanning, which is then processed through noise reduction, registration, and meshing to generate the three-dimensional model. The geometric model employs Hierarchical Level of Detail (LOD) technology, supporting multiple levels of precision from LOD100 to LOD400, controlling the model's data volume while ensuring visualization quality. Objects in the geometric model include building structures, parking space boundaries, passageways, column obstacles, and equipment shapes. Each object has a unique identifier, spatial coordinates, dimensional parameters, and the region it belongs to.

[0104] The physical model layer adds physical properties and kinematic characteristics to the equipment objects in the geometric model. Taking a barrier gate as an example, the physical model defines the following parameters: barrier length (meters), lifting and lowering time (seconds), motor rated power (watts), operating voltage (volts), protection rating (IP code), and operating temperature range (degrees Celsius). For a camera device, the physical model defines the following parameters: installation height (meters), horizontal field of view (degrees), vertical field of view (degrees), resolution (pixels), and night vision capability (infrared distance, meters). For lighting equipment, the physical model defines the following parameters: luminaire type, rated power (watts), luminous flux (lumens), and color temperature (Kelvin). Physical attribute data can be extracted from the specifications provided by the equipment manufacturer or obtained through on-site measurements.

[0105] The behavioral model layer is used to predict the dynamic behavior of vehicles and pedestrians within the parking lot. The behavioral model is a time-series prediction model built on a Long Short-Term Memory (LSTM) network. Training data comes from historical operational data, including fields such as vehicle entry timestamps, license plate numbers, dwell time, parking location, and driving trajectories. Input features of the LSTM network include: timestamps (converted to hourly and weekday periodic codes), current entry rate (number of vehicles entering in the past five minutes), parking space occupancy rates in different areas, and weather conditions (class codes such as sunny, rainy, and snowy). The hidden layers of the LSTM network contain two layers of LSTM units, each with 128 neurons, and use Dropout to prevent overfitting. The network output layer generates predicted vehicle arrival rates for three time windows: 15 minutes, 30 minutes, and 60 minutes. The behavioral model also includes a vehicle trajectory prediction sub-model, trained based on historical trajectory data, capable of predicting the probability distribution of vehicle path selection within the parking lot.

[0106] The rule model layer encodes the parking lot's business logic and operational rules. Implemented using rule engine technology, the rule model supports defining rule conditions and execution actions in a structured language. Rule categories include: charging rules (time-based billing, per-use billing, tiered pricing, discounts, etc.), access rules (one-way traffic, height and width restrictions, no-parking zones, priority lanes, etc.), safety rules (keeping fire lanes clear, designated parking spaces for disabled persons, emergency evacuation routes, etc.), and emergency rules (emergency opening of gates in case of malfunction, local mode in case of network interruption, etc.). The rule model supports dynamic configuration and hot updates of rules, allowing operators to modify rule parameters through the management interface without restarting the system.

[0107] The real-time data model layer establishes a data access pipeline with physical sensors, mapping the real-time state of the physical world. The types of data received by the real-time data model include: parking space status data (occupied or idle, update timestamp), gate status data (open, closed, faulty), environmental data (temperature, humidity, light intensity, air quality), video stream data (structured event data preprocessed by edge nodes), and payment data (transaction amount, payment method, transaction time). Real-time data is stored using a time-series database, supporting high-concurrency writes and fast time-range queries. Historical data is archived and cleaned according to a preset retention strategy.

[0108] The five model layers are linked through a unified object identifier. For example, a parking space object numbered PL-A-001 corresponds to a rectangular area in the geometric model, the parameters of a geomagnetic sensor and video detector in the physical model, the historical occupancy pattern of the parking space in the behavioral model, the charging category (regular or VIP) of the parking space in the rule model, and the current state and last update time of the parking space in the real-time data model. This multi-dimensional association allows the system to obtain a complete information view of the object from any dimension.

[0109] Example 3: Past Tense Knowledge Accumulation and Template Generation

[0110] Figure 3 This demonstrates a closed-loop process integrating the past, present, and future states. The left side shows the past state module, illustrating the process of extracting knowledge from historical project data, generating standardized templates, matching them with new project requirements, and outputting project proposals. The middle side shows the present state module, illustrating the process of data generated by physical devices, changes collected by mobile devices, real-time synchronization to the digital twin model, and maintaining consistency between the virtual and physical worlds. The right side shows the future state module, illustrating the process of importing project proposals or the current state, executing ultra-real-time simulations, evaluating multiple strategies, and outputting the optimal decision. The three modules are connected by arrows: project proposals output from the past state are validated in the future state; strategies output from the future state are executed in the present state; and operational data generated in the present state flows back to the past state to accumulate knowledge, forming a closed loop.

[0111] like Figure 7 As shown, the past knowledge accumulation module extracts experiential knowledge from deployed projects and generates reusable standardized templates.

[0112] During the historical data collection phase, the system extracts data from various management systems of the deployed projects. It extracts design documents, CAD drawings, equipment lists, and budget sheets from the design management system; construction progress records, change orders, and completion reports from the project management system; operation logs, fault records, maintenance work orders, and energy consumption data from the operation and maintenance management system; and revenue reports, customer flow statistics, and user feedback from the business management system. Data collection supports both automatic interface integration and manual import.

[0113] In the feature extraction phase, the system extracts structured feature indicators from the collected raw data. Design-related features include: parking lot type (above ground, underground, multi-level), total area (square meters), number of parking spaces, parking space type distribution (ratio of standard spaces, small spaces, accessible spaces, and charging pile spaces), number of entrances, number of exits, passage width, floor height, and equipment configuration list. Operational features include: average daily traffic flow, peak hour coefficient, average dwell time, turnover rate, parking space utilization rate, equipment failure rate, average fault recovery time, energy consumption per unit area, and average daily revenue per parking space. The feature extraction process is implemented using automated scripts. For unstructured data (such as documents and reports), natural language processing technology is used to extract key information.

[0114] In the normalization stage, the system standardizes the extracted feature indicators to eliminate the influence of dimensional differences. Numerical features are normalized using Z-score normalization or Min-Max normalization; categorical features are encoded using one-hot encoding or label encoding. The normalized feature vectors are used for subsequent similarity calculations and cluster analysis.

[0115] In the template clustering stage, the system uses clustering algorithms to divide historical projects into several categories. This embodiment uses the K-means clustering algorithm and determines the optimal number of clusters using the elbow rule. Taking fifty historical projects accumulated in the system as an example, cluster analysis divides them into the following five categories:

[0116] Category 1: High-turnover commercial complexes, characterized by high daily traffic volume, short average dwell time, high turnover rate, and the configuration of more gates and guidance equipment.

[0117] Category 2: Low-frequency residential communities, characterized by low daily traffic volume, long average dwell time (mainly overnight parking), low turnover rate, and fewer gate barriers.

[0118] Category 3: Tidal type transportation hubs, characterized by a clear bimodal distribution of traffic flow during morning and evening peak hours, a bimodal distribution of dwell time (short-term pick-up and drop-off and long-term travel), and the need to configure tidal lanes.

[0119] Category 4: Hospital visitor type, characterized by moderate traffic volume, dispersed stay duration (short stop at outpatient clinics, long stop at inpatient clinics), and high requirements for accessibility facilities.

[0120] Category 5: Industrial park commuter type, characterized by a single-peak distribution of traffic flow (entering early and exiting late), a stay duration concentrated in more than eight hours during weekdays, and a high proportion of monthly pass users.

[0121] For each category, the system generates a standardized knowledge template. The template includes: typical design parameters (parking space density, lane ratio, equipment configuration standards), performance benchmark values ​​(expected turnover rate, utilization rate, failure rate), optimization suggestions (design points and operation strategies summarized based on historical lessons), and economic indicator reference values ​​(construction cost per parking space, annual revenue per parking space, investment payback period).

[0122] In the similarity matching phase, when a user inputs parameters for a new project, the system calculates its similarity score with each knowledge template and returns the most matching template. Input parameters include: project type, scale, budget, schedule requirements, and special needs. Similarity calculation uses weighted Euclidean distance, and the weights of different parameters can be adjusted according to user preferences. The system returns the top three most similar templates for user reference and automatically generates a target project proposal based on the most matching template.

[0123] The target project proposal includes: equipment list and budget breakdown, construction schedule Gantt chart, expected performance indicators (turnover rate, utilization rate, failure rate, etc.), economic efficiency forecast (expected annual revenue, investment payback period, internal rate of return), and risk warnings (based on common problems and coping suggestions from historical projects).

[0124] Example 4: Technological Prediction of Economic Efficiency

[0125] This application bases the prediction of parking lot economic efficiency on quantifiable physical indicators, avoiding estimation methods that rely on subjective experience. The core idea is that parking lot revenue essentially depends on the spatiotemporal turnover efficiency of vehicles, which is determined by physical traffic capacity and space utilization.

[0126] Definition and calculation of space-time turnover rate: Space-time turnover rate is defined as the number of effective services provided per parking space per unit of time. The calculation formula is as follows:

[0127] Space-time turnover rate = Daily service trips / (Total number of parking spaces × Operating hours)

[0128] Daily service frequency can be predicted using behavioral models, estimated based on historical traffic flow patterns of similar projects and the location characteristics of the target project. The total number of parking spaces is determined by the design plan. Operating hours are determined by the operational strategy (e.g., 24-hour operation or daytime operation).

[0129] Calculation of the impact of barrier gate location on traffic capacity: The system simulates vehicle queuing processes under different barrier gate location schemes in a five-dimensional digital twin model. Simulation parameters include: barrier gate opening and closing time, vehicle arrival rate distribution, payment time distribution, and vehicle start delay. Simulation outputs include: average queue length, average waiting time, and number of vehicles passing through per unit time. By comparing the simulation results of different schemes, the system can quantify the impact of barrier gate location adjustments on traffic capacity. For example, moving the barrier gate forward by five meters can reduce the average queuing time by ten seconds, which translates to an increase of fifty vehicles per day, and further into an estimated increase in annual revenue.

[0130] Equipment duty cycle and energy cost calculation: The system calculates the duty cycle of equipment in each area based on traffic flow distribution and parking space occupancy patterns predicted by a behavioral model. Duty cycle is defined as the proportion of time the equipment is in operation out of the total operating time. Taking lighting equipment as an example, if area C is idle for more than 80% of the nighttime, the duty cycle of the lighting equipment in that area can be optimized to 20%. Based on this, the system calculates the energy-saving potential of the dimming strategy: Energy saving = Lighting power × (Original duty cycle - Optimized duty cycle) × Operating time × Electricity price.

[0131] Comparison of different parking space layout schemes: The system compares the overall efficiency of different parking space layout schemes in a five-dimensional digital twin model. Evaluation indicators include: total number of parking spaces, traffic efficiency (average search time), utilization rate (average daily occupancy time / daily operating time), and space-time turnover rate. The system automatically generates radar charts comparing each scheme and recommends the scheme with the highest overall score.

[0132] Through the aforementioned technical calculations, economic efficiency forecasting is no longer a subjective estimate based on guesswork, but rather a verifiable result of physical simulation, providing a scientific basis for investment decisions.

[0133] Example 5: Future-State Ultra-Real-Time Simulation and Deduction

[0134] like Figure 6 As shown, the future-oriented ultra-real-time simulation module performs simulations faster than real time in the virtual environment of the five-dimensional digital twin model, which is used to predict future trends and evaluate response strategies.

[0135] Implementation of the simulation execution engine: The simulation execution engine is implemented based on Discrete Event Simulation (DIS) technology. The simulation timeline is independent of the physical timeline and can proceed according to a set speedup ratio. For example, when the speedup ratio is set to ten times, the simulation engine can complete the scenario simulation for the next hour within six minutes. The simulation engine supports running multiple scenario instances in parallel, each instance corresponding to a different combination of initial conditions or strategy parameters.

[0136] Spatial Layout Verification: During the planning and design phase, the system imports the parking space layout scheme from the target project proposal into the geometric model, generating a 3D spatial scene. The simulation engine simulates the entry and exit processes of different vehicle types (small cars, SUVs, and commercial vehicles), checking the following indicators: whether the turning radius meets spatial constraints (whether it will scrape against pillars or adjacent vehicles), whether the distance between adjacent parking spaces meets the door opening safety distance, and whether the passage width meets the requirements for two-way passing. The simulation engine also simulates the passage conditions of fire lanes and pedestrian walkways, verifying whether emergency evacuation routes are unobstructed and whether evacuation time meets fire safety regulations. The simulation results are presented in the form of 3D animation and statistical reports, marking areas with conflicts or non-compliance with regulations, and providing optimization and adjustment suggestions.

[0137] Performance Load Assessment: The system simulates four traffic flow levels: low peak (30% of daily average traffic flow), off-peak (50% of daily average traffic flow), peak (1.2 times daily average traffic flow), and extreme peak (1.5 times daily average traffic flow). The simulation engine simulates the complete process of vehicle arrival, queuing, entry, parking, payment, and departure in each scenario. Collected metrics include: maximum queue length at the entrance (vehicles), average waiting time at the entrance (seconds), average parking space search time (seconds), peak parking space utilization, exit throughput capacity (vehicles / hour), and peak equipment load. By comparing the metrics across scenarios, the system identifies bottlenecks (such as insufficient number of gates or excessively narrow passageways in certain areas) and assesses whether the designed capacity meets expected demands. The performance load assessment report includes quantitative metrics, bottleneck analysis, and expansion recommendations.

[0138] Fault scenario drills: The system is injected with preset fault events in a simulation environment to simulate the system's emergency response process. Fault event types include:

[0139] Single gate malfunction: A gate at one entrance cannot be raised. Simulate how vehicles are guided to other entrances, how the queue grows, and how emergency manual gate opening is initiated.

[0140] Parking space detector offline: A batch of parking space detectors in a certain area go offline, simulating the adjustment of guidance strategies when the parking space status in that area cannot be updated.

[0141] Network communication interruption: Communication between edge nodes and the cloud is interrupted, simulating the switching process and functional degradation of local offline mode.

[0142] Sudden surge in passenger flow: Due to the dispersal of a large-scale event in the surrounding area, a large number of vehicles will flood in within a short period of time, simulating the process of dealing with entrance congestion and full parking lots.

[0143] Power failure: Power outage in part of the parking lot, simulating the activation of emergency lighting, manual opening of the gate, and vehicle evacuation.

[0144] The fault simulation records the following data: fault discovery time, emergency plan activation time, vehicle queue length variation curve, evacuation completion time, and customer complaint risk level. Based on the simulation results, the system evaluates the effectiveness of the existing emergency plan and outputs a fault response capability assessment report and plan optimization suggestions.

[0145] Operational Scenario Simulation: During the operational phase, the system initiates future scenario simulations at fixed intervals (e.g., every five minutes) based on current real-time data. Using the current parking space status, queuing situation, and traffic flow trends as initial conditions, the system simulates scenario changes over the next 30 minutes to two hours. If entrance congestion or full capacity in a certain area is predicted, the system evaluates the effectiveness of various response strategies in a virtual environment.

[0146] Strategy 1: Maintain the status quo and do not take any intervention measures.

[0147] Strategy 2: Enable tidal parking spaces and switch some two-way lanes to one-way entry lanes.

[0148] Strategy 3: Dynamic pricing, lowering rates in areas with high vacancy rates to divert vehicles.

[0149] Strategy 4: Issue a full parking warning and guide subsequent vehicles to nearby parking lots through navigation applications.

[0150] The simulation engine evaluates the expected effects of each strategy, including: entrance queue length, average waiting time, customer satisfaction rating, and revenue impact. The system selects the strategy with the highest overall score and recommends it to the decision engine.

[0151] Example 6: Dynamic Scheduling Optimization Driven by PPO Reinforcement Learning

[0152] like Figure 4 As shown, this application employs the Proximal Policy Optimization (PPO) algorithm to train a scheduling policy network in a virtual environment of a five-dimensional digital twin model.

[0153] Construction of the Reinforcement Learning Environment: A reinforcement learning environment is constructed based on a five-dimensional digital twin model. The environment state space includes: parking space occupancy rate in each area (normalized to a continuous value between zero and one), entrance queue length (normalized value), current time period (coded as weekday / weekend, morning / noon / evening, etc.), weather conditions (coded as sunny / rainy / snow, etc.), and predicted inbound traffic flow for the next 30 minutes. The action space includes: guidance weight for each area (probability of guiding vehicles to that area), whether to activate tidal flow mode, and dynamic pricing coefficient (the percentage increase or decrease based on the base rate).

[0154] Design of a multi-objective reward function: The reward function comprehensively considers three objectives: efficiency, energy consumption, and revenue, and is defined as follows:

[0155] R = α × R_efficiency + β × R_energy + γ × R_revenue

[0156] in:

[0157] R_efficiency = -average_waiting_time / baseline_waiting_time, representing the efficiency metric, with a negative average waiting time as the reward; shorter waiting times result in higher rewards. R_energy = -total_energy_consumption / baseline_energy_consumption, representing the energy consumption metric, with a negative total energy consumption as the reward; lower energy consumption results in higher rewards. R_revenue = total_revenue / baseline_revenue, representing the revenue metric, with total revenue as the reward; higher revenue results in higher rewards. α, β, and γ are weighting coefficients that can be adjusted according to operational priorities. For example, during peak holiday periods, the weight of α can be increased to prioritize traffic efficiency, while during off-peak nighttime periods, the weight of β can be increased to prioritize reducing energy consumption.

[0158] The weighting coefficients α, β, and γ are configured through the management interface, and operators can dynamically adjust them according to different management objectives. For example, during peak holiday periods, α=0.6, β=0.2, and γ=0.2 are set to prioritize traffic efficiency; during nighttime operation periods, α=0.2, β=0.6, and γ=0.2 are set to prioritize reducing energy consumption.

[0159] Training process of the PPO algorithm: The training process is executed on a cloud server, utilizing GPUs to accelerate the update computation of the policy network. The training steps are as follows:

[0160] Step 1: Initialize the parameters of the policy network and the value network. The policy network uses a two-layer fully connected neural network with 256 neurons in each hidden layer and ReLU activation function. The value network has the same structure as the policy network.

[0161] Step 2: Run multiple scenario instances in parallel in the simulation environment (32 parallel environments in this example) and collect state-action-reward sequences.

[0162] Step 3: Calculate the generalized advantage estimate (GAE) and the target value.

[0163] Step 4: Update the policy network parameters using the truncation objective function of PPO, and update the value network parameters using mean squared error loss.

[0164] Step 5: Repeat steps 2 to 4 until the cumulative reward converges or the preset training rounds are reached.

[0165] After training, the policy network is deployed to the hybrid intelligent decision engine. In actual operation, the policy network receives the current environmental state, outputs scheduling actions, and issues them for execution after heuristic rule verification.

[0166] Virtual Validation and Physical Deployment of Policies: To ensure policy reliability, the system employs Shadow Mode for policy validation. In Shadow Mode, the newly trained policy network runs in parallel with the current production policy, but only the output of the production policy is actually executed. The system records the output differences between the two policies and compares the simulated effects. When the new policy consistently outperforms the production policy in Shadow Mode for a certain period, the system prompts operations personnel to approve the policy switch. After approval, the new policy is upgraded to the production policy. This mechanism ensures the safety of policy optimization and avoids the risks associated with unvalidated policies directly affecting the physical system.

[0167] Example 7: Current Virtual-Real Synchronization and Change Management

[0168] like Figure 5 As shown, the current virtual-real synchronization module collects change information of the physical site through a mobile terminal to achieve accurate mapping from the design state to the completed state.

[0169] Example of a change in data collection scenario: During the construction of an underground parking lot, construction workers discovered that the planned location for a camera was obstructed by a fire hydrant, requiring the camera to be moved 1.5 meters to the right. In traditional processes, this change might only be communicated verbally or through written records, leading to a discrepancy between the digital model and the actual situation. In the process described in this application:

[0170] Construction workers open the mobile application for data collection and scan the QR code on the camera device. The system automatically identifies the device number as CAM-B2-015.

[0171] Construction workers select "location adjustment" as the change type, mark the new location on the plan, or enter the coordinate offset (X direction +1.5 meters, Y direction +0 meters, Z direction +0 meters).

[0172] Construction workers took photos of the site as evidence for the changes and filled in the reason for the changes (the original location was blocked by fire pipes).

[0173] Submitting a change record will automatically generate a change order, which includes: the object number, change type, location before change, location after change, reason for change, person making the change, time of change, and on-site photos.

[0174] The change log is uploaded to the cloud-based digital twin data center. The system automatically updates the spatial coordinates of the CAM-B2-015 object in the five-dimensional digital twin model.

[0175] The system triggers a recalculation of the field of view coverage. Because the camera's position has changed, its monitoring range also changes. The system recalculates the field of view coverage area in the geometric model and detects any new blind spots.

[0176] If a new blind spot is detected, the system will issue an alarm and provide adjustment suggestions (such as adjusting the camera angle or adding a blind spot replacement camera).

[0177] Structured storage of change data: Change records are stored in the database in a structured format, with key fields including:

[0178] Change order number (unique identifier)

[0179] Change object number (identifier for equipment or facility)

[0180] Change type (enumerated values: position adjustment, model replacement, parameter modification, addition, deletion)

[0181] State before change (JSON format, including attribute values ​​before the change)

[0182] Status after change (JSON format, including the changed attribute values)

[0183] Reason for change (text)

[0184] Change of Person (Personnel Identifier)

[0185] Change time (timestamp)

[0186] Attachment list (storage path of on-site photos or documents)

[0187] Approval status (enumerated values: pending approval, approved, rejected)

[0188] Synchronization status (enumeration values: pending synchronization, synchronized, synchronization failed)

[0189] Automated Model Synchronization: The system automatically updates the corresponding attributes in the five-dimensional digital twin model based on the change log. For location adjustment changes, the spatial coordinates of objects in the geometric model are updated; for model replacement changes, the equipment parameters of objects in the physical model are updated (e.g., if a different model of barrier gate is used, parameters such as lifting and lowering time and motor power need to be updated); for parameter modification changes, the corresponding parameter values ​​of objects in the physical model or rule model are updated. After the update is completed, the system triggers correlation calculations for the affected area, such as view coverage, channel capacity, and lighting coverage. If the correlation calculation results show abnormalities (e.g., monitoring blind spots or channel bottlenecks), the system issues an alarm to notify relevant personnel for handling.

[0190] Consistency verification mechanism: The system periodically performs consistency verification, comparing the degree of conformity between the digital model and the physical environment. Verification methods include:

[0191] Automatic verification: The system verifies the accuracy of model parameters based on real-time data from physical sensors. For example, if the actual opening and closing time of the barrier gate deviates from the parameters recorded in the model by more than a threshold, the system will indicate a verification error.

[0192] Manual verification: Maintenance personnel conduct regular on-site inspections and use a mobile application to check the equipment status against the model records, marking any discrepancies found.

[0193] Discrepancies discovered during verification are recorded as changes pending confirmation and, after manual review, are synchronized to the model. This mechanism ensures long-term consistency between the digital twin model and the physical entity.

[0194] Example 8: Dual-mode parking space detection and fusion judgment

[0195] like Figure 8 As shown, this application adopts a dual-mode parking space detection scheme that combines video detection and geomagnetic detection, and improves the detection accuracy through a fusion judgment algorithm.

[0196] Video Detection Unit: One or more high-definition cameras are deployed in each detection area, covering several parking spaces. The video streams captured by the cameras are analyzed in real time at edge computing nodes, using deep learning-based object detection algorithms (such as YOLO) to identify the presence of vehicles within the parking space area. The advantage of video detection is that it eliminates the need for independent sensors at each parking space, allowing a single camera to cover multiple spaces, and it can identify additional information such as vehicle type and color. The disadvantage of video detection is that it is greatly affected by lighting conditions; detection accuracy may decrease at night, in backlight, or in strong shadow environments.

[0197] Geomagnetic Detection Unit: A geomagnetic sensor is buried beneath each parking space to detect changes in the geomagnetic field of the parking area. When a vehicle drives in and parks above the space, the vehicle's metal body causes a significant change in the geomagnetic field, which the sensor uses to determine that the parking space is occupied. The advantage of geomagnetic detection is that it is unaffected by lighting conditions and can operate stably in various environments. The disadvantage of geomagnetic detection is that it may be affected by surrounding electromagnetic interference and cannot provide additional information such as vehicle type.

[0198] Fusion Judgment Algorithm: The system inputs the video detection results and geomagnetic detection results into the fusion judgment module, and uses a weighted voting algorithm to determine the final parking space status. The algorithm logic is as follows: Let the video detection result be V (0 indicates vacant, 1 indicates occupied), and the geomagnetic detection result be M (0 indicates vacant, 1 indicates occupied). Let the video detection weight be Wv, and the geomagnetic detection weight be Wm, satisfying Wv + Wm = 1. Calculate the weighted score S = Wv × V + Wm × M. If S ≥ 0.5, the parking space status is determined to be occupied; if S < 0.5, the parking space status is determined to be vacant.

[0199] The weighting mechanism dynamically adjusts Wv based on current lighting conditions. The system obtains the current light intensity through an ambient light sensor. When the light intensity is below a threshold (e.g., at night or in a dimly lit basement area), the value of Wv is reduced; when the light intensity is normal, the standard value of Wv is restored. Wm is dynamically adjusted based on the degree of electromagnetic interference. The system monitors signal fluctuations from a geomagnetic sensor. When abnormal electromagnetic interference is detected (e.g., the operation of high-power electrical equipment nearby), the value of Wm is reduced.

[0200] By default, Wv = 0.5 and Wm = 0.5, meaning the two detection methods have equal weight. When the two detection results are consistent, the final judgment result is the same. When the two detection results are inconsistent, the system determines the final result based on the weights, triggers an anomaly alarm, and records the event in the operations log for manual review.

[0201] Real-time synchronization of detection results: The parking space status after fusion judgment is synchronized to the real-time data model of the five-dimensional digital twin model in real time. Status updates adopt a message queue mechanism to ensure the reliability of data transmission under high concurrency. Each status update message includes: parking space number, current status (occupied / idle), status update timestamp, and detection source (video / geomagnetic / fusion).

[0202] Example 9: Predictive Maintenance

[0203] This application uses time-series analysis of equipment operation data to achieve early warning of equipment failures and automatic generation of maintenance work orders.

[0204] Definition of Equipment Health Indicators: The system defines health assessment indicators for each type of equipment. Taking a barrier gate as an example, the health indicators include:

[0205] Cumulative opening and closing times: Record the cumulative number of times the barrier gate is opened and closed, and compare it with the design life (e.g., 500,000 times) to calculate the consumption ratio.

[0206] Opening and closing time deviation: Record the actual time taken for each opening and closing, compare it with the nominal value (such as 1.5 seconds) to calculate the deviation. A continuous increase in deviation may indicate mechanical wear.

[0207] Abnormal motor current: Monitor the motor's operating current. An abnormally high current may indicate an increased load or motor aging.

[0208] Failure frequency: Statistics on the number of recent failures; frequent failures indicate a deterioration in equipment condition.

[0209] Construction of the Health Prediction Model: The system employs the Gradient Boosting Decision Tree algorithm to construct the equipment health prediction model. Training data is derived from historical fault records; positive samples are equipment operation data from a period prior to the fault (e.g., seven days), and negative samples are equipment operation data during normal operation. Model input features include the time series of the aforementioned health indicators and their statistics (mean, variance, trend slope, etc.). Model output is the predicted fault probability or remaining service life.

[0210] Automatic generation of maintenance work orders: The system performs a health assessment on all equipment daily. When the predicted remaining useful life of a device is lower than a preset threshold (e.g., 30 days) or the probability of failure exceeds a preset threshold (e.g., 80%), the system automatically generates a preventative maintenance work order. The work order includes: device number, device location, device type, health assessment result, recommended maintenance content, and recommended completion time. The work order is pushed to the mobile terminals of maintenance personnel and synchronized to the maintenance management system.

[0211] Closed-loop feedback for maintenance effectiveness: After completing maintenance, maintenance personnel fill out a maintenance record via mobile terminal, including: maintenance content, replaced parts, maintenance time, and post-maintenance test results. The maintenance record is then sent back to the knowledge template library for optimization of the following parameters:

[0212] Equipment lifespan correction based on experience: If the actual replacement cycle of a certain type of barrier gate is generally shorter than the design lifespan, the system adjusts the lifespan prediction parameters for that type of equipment.

[0213] Maintenance interval optimization: If preventive maintenance effectively avoids failures, the system optimizes maintenance interval recommendations accordingly.

[0214] Iterative training of the fault prediction model: New fault cases are added to the training set, and the prediction model is retrained periodically to improve accuracy.

[0215] Example 10: Collaborative Scheduling of Multiple Parking Lots

[0216] For scenarios where there are multiple parking lots in the same area, this application supports regional-level collaborative scheduling to achieve balanced traffic flow and resource sharing across parking lots.

[0217] Construction of a regional-level digital twin model: The system establishes relationships between multiple parking lots within the same region, constructing a regional-level digital twin model. The regional model includes five-dimensional digital twin model instances for each parking lot, as well as spatial relationships (distance, travel time) and business relationships (whether they belong to the same operating entity) between parking lots. The regional model summarizes key indicators for each parking lot in real time: total number of parking spaces, number of available parking spaces, parking space occupancy rate, entrance queue length, and current fee rate.

[0218] Triggering the coordinated scheduling strategy: When the occupancy rate of a parking lot exceeds a preset threshold (e.g., 90%) or the queue length at the entrance exceeds a preset threshold (e.g., 20 vehicles), the system triggers the coordinated scheduling strategy. The decision logic for coordinated scheduling includes:

[0219] Identify nearby vacant parking lots: Retrieve parking lots within a five-minute drive of the target parking lot in the regional model that have an vacancy rate of more than 30%.

[0220] Assess the feasibility of guidance: Consider factors such as the walking distance between the destination of the guided vehicle and the candidate parking lot, and the difference in rates, to assess the driver's willingness to accept guidance.

[0221] Generate guidance strategies: Generate guidance suggestions for subsequently arriving vehicles and publish them through navigation applications and variable information signs.

[0222] Execution method of collaborative scheduling: Collaborative scheduling is executed through the following channels:

[0223] Navigation application integration: The system interfaces with mainstream navigation applications. When the driver sets the destination to the target parking lot, if the parking lot is in high occupancy, the navigation application will display a prompt message (the parking lot is in short supply, we recommend going to the nearby XX parking lot, which is about a five-minute walk from the destination), and provide alternative routes.

[0224] Variable information signs: Install variable information signs on roads around the target parking lot to display the number of available parking spaces and guidance suggestions in real time.

[0225] Entrance guidance screen: The guidance screen at the entrance of the target parking lot displays a full parking notice and directions to nearby parking lots.

[0226] Evaluation and optimization of collaborative scheduling effectiveness: The system records the execution status and effectiveness data of collaborative scheduling, including: the number of times guidance information is issued, the proportion of drivers accepting guidance (estimated by comparing navigation route change data), changes in the overall parking space utilization rate of the area, and driver satisfaction feedback. This data is fed back to the regional-level digital twin model to optimize the parameters of the collaborative scheduling strategy, such as occupancy thresholds, the ranking algorithm for recommended parking lots, and the wording of guidance information.

[0227] Example 11: Security Protection Mechanism

[0228] The system in this application includes a robust security protection mechanism to ensure data security and operational traceability.

[0229] Data Encryption: The system employs the national standard SM4 algorithm to encrypt and protect sensitive data. Encryption covers: user personal information (license plate number, mobile phone number), payment information, device control commands, and policy parameters. Data is encrypted using the TLS protocol during transmission and at the field level during storage. Key management utilizes a hardware security module (HSM), with keys rotated periodically.

[0230] Access Control: The system adopts a role-based access control (RBAC) policy, defining the following typical roles and permissions:

[0231] System administrator: Has permissions for user management, role configuration, and system parameter settings.

[0232] Operations Administrator: Has permissions to adjust rates, configure strategies, and view reports.

[0233] Maintenance personnel: Have permissions to view equipment status, process maintenance work orders, and submit change records.

[0234] Data Analyst: Has permissions to query historical data and export reports.

[0235] Visitors have access to parking space and fee inquiry permissions.

[0236] Each user's access permissions are strictly limited to their assigned role, and sensitive operations require approval from a superior.

[0237] Operation Auditing: The system records operation logs for all users. Log content includes: operator, operation time, operation type, operation object, pre-operation status, post-operation status, and client IP address. Operation logs are retained for at least three years and support retrieval by time, operator, operation type, and other criteria. Abnormal operations (such as sensitive operations outside of working hours, batch data export, etc.) trigger real-time alarms.

[0238] Blockchain Evidence Storage: The system stores key operation logs and important data change records on the blockchain to ensure data immutability and traceability. The stored content includes: a change record summary (hash value), operation timestamp, operator identifier, and operation type. The blockchain adopts a consortium blockchain architecture, with nodes jointly maintained by the operating entity and regulatory agencies. The authenticity and integrity of historical records can be verified through the blockchain at any time.

[0239] In one embodiment of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described parking management method based on a five-dimensional digital twin throughout the entire lifecycle.

[0240] In one embodiment of this application, a computer program product is provided, including a computer program / instructions, which, when executed by a processor, implements the steps of the parking management method based on a five-dimensional digital twin throughout the entire lifecycle.

[0241] The computer-readable storage medium and computer program product provided in this embodiment are similar in implementation principle and technical effect to the above method embodiments, and will not be repeated here.

[0242] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.

[0243] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0244] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A parking management method based on a five-dimensional digital twin throughout the entire lifecycle, characterized in that, The method includes: S1. Construct a five-dimensional digital twin model; wherein, the five-dimensional digital twin model includes a geometric model, a physical model, a behavioral model, a rule model, and a real-time data model. The geometric model is used to construct a three-dimensional spatial mapping of the parking lot's physical environment. The physical model is used to define the physical attributes and kinematic characteristics of the equipment. The behavioral model is used to predict the dynamic behavioral trajectories of vehicles and pedestrians. The rule model is used to encode business logic and safety constraints. The real-time data model is used to map the state data flow of the physical world. S2. Extract historical data of design parameters, operation data and performance indicators from at least one operational parking lot, generate a standardized knowledge template library through data mining, and use a similarity matching algorithm to match the most similar template in the standardized knowledge template library based on the input parameters of the target project, and generate a target project proposal that includes a bill of materials, schedule plan and performance prediction. S3. Load the target project proposal into a five-dimensional digital twin model, perform multi-dimensional simulation in virtual space at a speed higher than the physical time flow rate, and output a simulation verification report; wherein, the simulation includes at least spatial layout verification, performance load assessment and fault scenario drills; S4. During the project deployment phase, change information of the physical site relative to the original design is collected through mobile terminals, and the change information is synchronized to the five-dimensional digital twin model in real time to realize the mapping update from the design state to the completed state; wherein, the change information includes at least equipment location adjustment, model replacement and parameter modification; Specifically, S4 includes: The mobile terminal runs a change collection application, which supports obtaining device change information through photo recognition, manual input, or QR code scanning. The change information is processed in a structured manner to generate a change record, which includes the change object identifier, change type, parameter value before change, parameter value after change, and change timestamp. The change record is uploaded to the cloud-based digital twin data center via a network interface, triggering attribute updates for the corresponding geometric and physical models in the five-dimensional digital twin model; After the update is completed, the view coverage, path planning, and performance parameters of the affected area are recalculated to ensure the consistency between the five-dimensional digital twin model and the physical entity. Specifically, the recalculation of the view coverage of the affected area involves recalculating the view coverage area in the geometric model and detecting whether new blind spots have appeared due to the change in camera position. If a new blind spot is detected, an alarm is issued and adjustment suggestions are given, such as adjusting the camera angle or adding a supplementary camera. S5. During the operation phase, continuously collect physical terminal operation data, input the operation data into the five-dimensional digital twin model for strategy optimization training, distribute the optimized scheduling strategy to the physical execution terminal, and send the execution feedback data of the physical terminal back to the standardized knowledge template library to achieve closed-loop iteration throughout the entire life cycle.

2. The parking management method according to claim 1, characterized in that, In constructing a five-dimensional digital twin model in S1, the construction of the behavioral model specifically includes: Historical operational data, including vehicle entry time, dwell time, driving trajectory, and pedestrian movement path, were collected as training samples. A time-series prediction model based on Long Short-Term Memory (LSTM) network is constructed. The input layer of the LSTM network receives timestamps, entry rates, and area occupancy features. The hidden layer contains at least two LSTM units. The output layer generates predicted values ​​for vehicle arrival rate and pedestrian density distribution within a future preset time window. The trained time-series prediction model is embedded into the behavioral model layer of the five-dimensional digital twin model.

3. The parking management method according to claim 1, characterized in that, S2 generates a target project proposal that includes a bill of materials, a schedule, and performance forecasts, and also includes economic performance forecasts, specifically: Based on the behavior model in the five-dimensional digital twin model, the vehicle passage trajectory and dwell time under different parking space layout schemes are simulated; Based on the preset operational strategy parameters, the time-space turnover rate index corresponding to each plan is calculated. The time-space turnover rate is defined as the number of effective services per parking space per unit time. Based on the spatiotemporal turnover rate index, the system outputs parking space layout optimization suggestions and dynamic pricing strategies that take into account both traffic efficiency and economic returns.

4. The parking management method according to claim 1, characterized in that, The simulation in S3 includes spatial layout verification, performance load assessment, and fault scenario drills, among which: The specific fault scenario drills include: injecting preset fault events into a five-dimensional digital twin model, wherein the fault events include at least one of the following: gate failure, parking space detector offline, network communication interruption, and sudden large passenger flow; simulating the system response process after the fault event occurs, recording vehicle queue length, evacuation time, and the activation status of backup strategies; evaluating the effectiveness of existing emergency plans based on simulation results, and outputting a fault response capability assessment report and plan optimization suggestions; The spatial layout verification specifically includes: generating a three-dimensional spatial model in the virtual environment of a five-dimensional digital twin model according to the parking space layout scheme in the target project proposal; simulating the entry and exit processes of different vehicle types to check whether the turning radius of the vehicles meets the spatial constraints and whether the distance between adjacent parking spaces meets the safety standards; simulating the passage of fire lanes and pedestrian walkways to verify the unobstructedness of emergency evacuation routes; outputting a spatial layout verification report, marking areas with conflicts or non-compliance with specifications, and generating optimization and adjustment suggestions; The performance load assessment specifically includes: setting different levels of traffic flow input conditions in a five-dimensional digital twin model to simulate four scenarios: low peak, off-peak, peak, and extreme peak; recording entrance queue length, average waiting time, parking space utilization rate, and exit capacity indicators under each scenario; identifying system bottlenecks and assessing whether the number of gate barriers, channel width, and guidance signs meet the design capacity requirements; and outputting a performance load assessment report, including quantitative indicators and expansion suggestions for each scenario.

5. The parking management method according to claim 1, characterized in that, The method also includes a dynamic switching step for tidal parking spaces, specifically including: Based on the aforementioned behavioral model, predict the inbound and outbound vehicle traffic within a future preset time window; When the ratio of the predicted inbound traffic flow to the outbound traffic flow exceeds a preset threshold, the system is determined to enter tidal mode. In tidal mode, a preset number of bidirectional lanes will be switched to one-way entry or one-way exit lanes, and vehicles will be guided through the lanes by variable message signs and ground projection indicators. The tidal switching records are synchronized to the five-dimensional digital twin model for subsequent strategy optimization training.

Images