Digital Twin-enabled Digital Parking Management System

Abstract

This invention relates to the field of digital parking management technology and discloses a digital twin-enabled digital full-domain parking management system. The system includes generating a full-domain digital twin model of the parking lot, integrating multi-source sensing data such as parking space status, vehicle identity, and environmental parameters; combining user parking requests and real-time parking space status to generate a parking space allocation matrix with priority and time windows; outputting a set of timestamped vehicle guidance paths based on parking space QR code hardware parameters, the digital twin model, and the parking space allocation matrix; constructing a real-time obstacle avoidance and dynamic adjustment system to generate a set of safe driving paths for vehicle dispatch; combining energy storage unit power, power distribution network data, and the set of safe driving paths to output a power distribution control command set and update the digital twin model; and finally, achieving system knowledge evolution and end-to-end closed-loop optimization through dispatch execution logs. This system achieves efficient full-domain parking management, improving parking space utilization, vehicle dispatching, and energy utilization.

Description

Technical Field

[0001] This invention relates to the field of digital parking management technology, specifically a digital twin-enabled digital full-domain parking management system. Background Technology

[0002] With the acceleration of urbanization and the continuous increase in car ownership, parking lots, as a crucial infrastructure in the urban transportation system, directly impact the smoothness of urban traffic and the travel experience of users through their management efficiency and service quality. Currently, most parking lots still adopt relatively traditional management models, exhibiting significant shortcomings in multiple aspects such as parking space management, vehicle scheduling, and energy utilization.

[0003] In terms of parking space management, traditional parking lots often rely on single sensors to detect parking space status. These methods typically suffer from limited sensing range and insufficient data accuracy. Some parking lots can only achieve simple identification of whether a parking space is occupied, failing to obtain crucial information such as vehicle identity and model, resulting in a lack of targeted parking space allocation. Furthermore, data collected by different sensing devices is independent and difficult to integrate effectively. Managers struggle to comprehensively grasp the parking space usage across the entire parking lot, often resulting in some areas having idle spaces while others have queues, leading to inefficient use of parking resources.

[0004] Vehicle dispatching and guidance also suffer from numerous drawbacks. Existing parking lot route planning is largely based on fixed, static maps, failing to consider real-time vehicle flow, temporary obstacles, and other dynamic factors. When unexpected situations arise within the parking lot, such as vehicle convergence, pedestrian crossings, or equipment malfunctions, pre-planned routes often become inapplicable, easily leading to traffic congestion. Furthermore, guidance information is not precise enough; some parking lots only use simple indicator lights or signs to guide vehicles, failing to provide drivers with real-time, personalized guidance routes. This causes drivers to wander around and struggle to find parking spaces, wasting time and exacerbating traffic chaos within the parking lot.

[0005] In terms of energy management, parking lot energy storage units and power distribution networks typically operate independently, without effective coordination with vehicle dispatching. The charging and discharging strategies of energy storage units are mostly based on preset time or energy thresholds, without dynamic adjustments based on actual conditions such as vehicle charging needs and driving status within the parking lot. This can lead to excessive load on the power distribution network during peak charging periods, resulting in unstable power supply; while during off-peak periods, the energy stored in the units cannot be fully utilized, resulting in energy waste. Furthermore, energy management data is fragmented from other parking lot management data, making it difficult for managers to plan energy use holistically, further reducing energy efficiency. Summary of the Invention

[0006] The purpose of this invention is to provide a digital twin-enabled digital full-domain parking management system to solve the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention provides a digital twin-enabled digital full-domain parking management system, the system comprising:

[0008] Generate a digital twin model of the entire parking lot area, and perform multi-source perception and dynamic data fusion on the parking lot area. The multi-source perception includes parking space status recognition, vehicle identity recognition and environmental parameter collection.

[0009] The system acquires user parking requests and real-time parking space status, performs request parsing and dynamic allocation based on the parking lot's full-domain digital twin model, and generates a parking space allocation matrix with priority and time windows.

[0010] Collect the device parameters of the QR code hardware installed in the parking space, and perform multi-objective hierarchical path planning based on the parking lot full-domain digital twin model and the parking space allocation matrix, and output a set of vehicle guidance paths with timestamps.

[0011] A real-time obstacle avoidance and dynamic adjustment system is constructed based on the vehicle guidance path set, a safe driving path set is generated, and vehicle scheduling and control are performed based on the safe driving path set.

[0012] The system acquires the power consumption of the parking lot's energy storage unit and the power distribution network data, combines them with the set of safe driving paths, performs energy collaborative optimization and dynamic scheduling, outputs a set of power distribution control commands, and triggers the update of the parking lot's full-domain digital twin model.

[0013] Obtain vehicle dispatch execution logs, perform system knowledge evolution, extract features of successful dispatch cases, and complete end-to-end closed-loop optimization.

[0014] Preferably, the process of generating a full-domain digital twin model of the parking lot includes:

[0015] Real-time images are captured by parking space cameras, and deep learning models are used to identify the parking space occupancy status and vehicle identity information.

[0016] Vibration data of parking spaces are obtained by using geomagnetic sensors to determine vehicle parking behavior and abnormal conditions.

[0017] Temperature, humidity, light intensity, and road surface condition parameters are collected via IoT terminals.

[0018] Based on the parking space occupancy status, the vehicle identity information, the vehicle parking behavior, and the environmental parameters, a multi-dimensional parking lot status map is constructed.

[0019] By integrating the multi-dimensional state map of the parking lot with high-precision map data, a full-domain digital twin model of the parking lot is generated.

[0020] Preferably, the process of generating the parking space allocation matrix with priority and time window includes:

[0021] Parse the vehicle type, reservation time, and user level information in the user's parking request;

[0022] Query the historical parking database to obtain user credit scores and preference settings;

[0023] Based on the real-time parking space status in the full-domain digital twin model of the parking lot, the matching degree of each parking space is calculated;

[0024] Dynamically prioritize the vehicle type, user level, credit score, and matching degree.

[0025] Generate the parking space allocation matrix, which includes allocation time windows and priorities.

[0026] Preferably, the process of outputting the time-stamped vehicle guidance path set includes:

[0027] Obtain the communication protocol and positioning accuracy parameters of the QR code hardware;

[0028] Based on the parking lot's global digital twin model and the parking space allocation matrix, global path planning is performed;

[0029] An improved ant colony algorithm is used to initialize the path population, and historical best paths and manually preset paths are injected.

[0030] Adjust the path intersection strategy based on parking space priority;

[0031] The path variation rate is dynamically adjusted based on real-time environmental data.

[0032] The output is the set of vehicle guidance paths containing the coordinate sequence, driving speed, and timestamp.

[0033] Preferably, the process of generating a set of safe driving routes includes:

[0034] The surrounding obstacles are detected by fusing lidar and ultrasonic sensors;

[0035] Establish a multi-level obstacle avoidance mechanism to dynamically adjust the path offset based on the distance to obstacles;

[0036] A segmented optimization algorithm is used to generate a smooth obstacle avoidance path;

[0037] Route adjustment information is exchanged in real time via the vehicle-side communication module;

[0038] The set of safe driving paths is generated and synchronized to each vehicle control unit.

[0039] Preferably, the process of outputting the power distribution control command set includes:

[0040] Monitor the remaining power of the parking lot energy storage unit and the working status of the charging piles; estimate the energy demand for each time period based on the safe driving path set; perform dynamic load allocation in combination with the power distribution network topology; generate the power distribution control instruction set containing charging scheduling instructions and power distribution switching instructions;

[0041] The energy state layer of the parking lot global digital twin model is updated according to the power distribution control instruction set.

[0042] Preferably, the QR code hardware installed in the parking space includes a multi-color light source and a communication module;

[0043] The light source distinguishes user level and operation status through color changes and flashing patterns;

[0044] The communication module supports vehicle identity authentication and wireless charging control command transmission;

[0045] Scanning the QR code triggers a parking space locking, charging activation, or shared vehicle dispatch process.

[0046] Preferably, the QR code hardware supports nationwide networked shared vehicle dispatching; scanning the QR code activates the cross-regional vehicle reservation function; coordinates off-site parking space resources based on the parking lot's full-domain digital twin model; and generates cross-regional route planning instructions and automatic parking control sequences.

[0047] Preferably, the process of completing end-to-end closed-loop optimization includes:

[0048] Collect data on actual vehicle driving trajectories and scheduling results; extract feature vectors of successful obstacle avoidance paths and efficient scheduling sequences; train and optimize path planning parameters using a neural network model; and deploy the updated parameters to the overall management system.

[0049] Preferably, the system further includes a wireless charging control module mounted in the parking space;

[0050] Scanning the QR code triggers hardware authentication and power transmission for the charging pile.

[0051] The charging power is dynamically adjusted based on the energy state layer of the parking lot's full-domain digital twin model.

[0052] The color change of the light source indicates the charging status and provides anomaly warnings.

[0053] Compared with the prior art, the beneficial effects of the present invention are:

[0054] In the parking space allocation process, the system uses a full-domain digital twin model to analyze and dynamically allocate parking spaces based on user parking requests and real-time parking space status, generating a parking space allocation matrix with priority and time windows. This allocation method no longer relies on single parking space availability information, but comprehensively considers factors such as the urgency of the user's parking needs and vehicle type suitability, enabling precise allocation of parking spaces to the most suitable users. Whether it's a short-term temporary parking user or a long-term parking user, they can obtain a parking space that meets their needs, significantly reducing the inconvenience caused by unsuitable parking spaces. It also allows for a more rational allocation of parking resources, avoiding the coexistence of idle parking spaces and overcrowded demand.

[0055] To address the challenge of vehicle guidance, the system collects device parameters from parking space QR code hardware and combines this with a digital twin model and parking space allocation matrix to perform multi-objective hierarchical path planning, outputting a set of timestamped vehicle guidance paths. Hierarchical path planning considers multiple objectives, including shortest path, highest traffic efficiency, and avoidance of congested areas. The addition of timestamps allows the guidance paths to be synchronized with the real-time driving status of vehicles. Drivers can use the timestamped guidance paths to accurately control their driving pace, avoiding sudden acceleration or braking in parking lots, thus improving driving safety and reducing traffic conflicts. Even during peak hours, it effectively alleviates traffic congestion, allowing vehicles to reach their target parking spaces quickly and smoothly.

[0056] The construction of a real-time obstacle avoidance and dynamic adjustment system further ensures vehicle safety within the parking lot. The system controls vehicle dispatch based on a generated set of safe driving paths, and can monitor emergencies in the parking lot in real time, such as temporary obstacles or other vehicles violating traffic rules, and quickly adjust vehicle routes. Compared to traditional parking lots that rely on manual detection and handling of emergencies, this system responds faster and adjusts more promptly, minimizing the occurrence of traffic accidents and providing a safer environment for drivers and pedestrians within the parking lot.

[0057] In terms of energy management, the system acquires data on the power supply of parking lot energy storage units and the power distribution network, and combines this with a set of safe driving paths for coordinated energy optimization and dynamic scheduling, outputting a set of power distribution control commands. This process achieves deep integration of energy storage units, the power distribution network, and vehicle scheduling, enabling flexible adjustments to the charging and discharging strategies of energy storage units and the power allocation of the power distribution network based on the real-time driving status and charging needs of vehicles in the parking lot. When vehicles need to charge in a concentrated manner, the system can rationally allocate the power supply of energy storage units to avoid excessive load on the power distribution network; when vehicles are traveling more dispersedly and charging demand is lower, the system can charge energy storage units in a timely manner, making full use of the power grid's off-peak hours. Through this dynamic scheduling, not only is the stability of the parking lot's power supply ensured, but energy utilization efficiency is also improved, and energy waste is reduced. Attached Figure Description

[0058] Figure 1 A sequence diagram of the digital twin-enabled digital full-domain parking management system described in this invention;

[0059] Figure 2 A flowchart for generating a full-domain digital twin model of a parking lot;

[0060] Figure 3 A flowchart for generating a parking space allocation matrix with priority and time window. Detailed Implementation

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] Please see Figure 1This invention provides a digital twin-enabled digital full-domain parking management system. The system integrates multi-source sensing, dynamic data fusion, and intelligent decision-making to achieve comprehensive digital management of parking lots. First, the system generates a full-domain digital twin model of the parking lot, built upon multi-source sensing data, including parking space status recognition, vehicle identification, and environmental parameter collection. Multi-source sensing is achieved through a sensor network deployed in the parking area, such as parking space cameras for capturing real-time images, geomagnetic sensors for monitoring vibration data, and IoT terminals for collecting temperature, humidity, light intensity, and road surface condition parameters. This data is dynamically fused to form a consistent environmental status representation. User parking requests are submitted via mobile applications or network interfaces. The system parses the request content and combines it with real-time parking space status, using the digital twin model for dynamic allocation, generating a parking space allocation matrix with priority and time windows. The matrix generation process considers vehicle type, user level, and historical credit score to ensure fair and efficient allocation. Subsequently, the system collects device parameters of the QR code hardware installed in the parking spaces, such as communication protocols and positioning accuracy, and performs multi-objective hierarchical path planning based on the digital twin model and the parking space allocation matrix. The path planning employs an optimization algorithm, outputting a time-stamped set of vehicle guidance paths, including coordinate sequences, driving speeds, and expected time points. Based on this path set, the system constructs a real-time obstacle avoidance and dynamic adjustment mechanism. It uses sensor fusion to perceive obstacles and generates a set of safe driving paths for vehicle dispatch control. Simultaneously, the system monitors the power status of parking lot energy storage units and power distribution network data, combining this with the safe driving path set for energy synergistic optimization, outputting a power distribution control command set, and triggering updates to the digital twin model. Finally, the system acquires vehicle dispatch execution logs, performs knowledge evolution processing, extracts features from successful dispatch cases, and completes end-to-end closed-loop optimization through machine learning to improve system performance.

[0063] See Figure 2 Parking space cameras, as the core component of visual perception, are deployed directly above or adjacent to the supporting structure of each parking space. These cameras employ wide-angle, high-definition lenses to cover the entire area of ​​the parking space, and the real-time video streams are transmitted to edge computing nodes via wired or wireless networks. At the edge nodes, a deep learning model based on a convolutional neural network architecture performs real-time image analysis. This network, trained on a large amount of labeled data, can identify the occupancy status of parking spaces (vacant, occupied, abnormally occupied) and extract vehicle identification information (such as license plate number, vehicle type, and color). The image processing includes background subtraction, object detection, and feature extraction steps to address the challenges posed by different lighting and weather conditions.

[0064] Geomagnetic sensors are embedded beneath the center of each parking space as an auxiliary sensing method. These sensors, based on microelectromechanical systems (MEMS) technology, perform high-frequency sampling of the minute vibrations and magnetic field changes generated when vehicles park and leave. The collected raw signals are filtered and analyzed to determine the type of vehicle parking behavior (e.g., normal parking, emergency braking, prolonged stay) and detect abnormal states (e.g., illegal occupation, equipment failure). Sensor data is transmitted to the gateway device via Bluetooth Low Energy or ZigBee protocols. IoT terminals constitute the environmental sensing layer. These terminal devices are distributed in various key areas of the parking lot and integrate temperature and humidity sensors, light sensors, and road condition sensors. Temperature and humidity sensors monitor environmental thermodynamic parameters to assess the risk of icing or high temperatures; light sensors detect lighting intensity to support energy-saving control; and road condition sensors use optical or resistive principles to detect water accumulation, icing, or oil stains. All environmental parameters are uploaded to the cloud platform via LoRaWAN or NB-IoT networks. The construction of a multidimensional state graph is a data integration and structuring process. The system performs spatiotemporal alignment and format unification of perception data from different sources. Parking space occupancy status and vehicle identity information are organized as node attributes, vehicle parking behavior and abnormal status are used as edge relation weights, and environmental parameters are used as dynamic attribute labels for the environment, thus forming a graph network containing entities, relationships, and attributes.

[0065] High-precision map data is pre-acquired using laser scanning and photogrammetry techniques. It includes the parking lot's 3D geometry, precise lane locations, slope curvature information, and coordinates of fixed obstacles. During the data fusion phase, Kalman filtering and Bayesian estimation methods are used to perform layered matching and calibration between the real-time multi-dimensional state map and the static high-precision map. The final generated digital twin model of the parking lot is visualized using a 3D rendering engine. The model comprises three layers: a physical layer, a data layer, and a rule layer. The physical layer maps the spatial structure of the real parking lot; the data layer integrates real-time perception and historical data streams; and the rule layer embeds traffic flow simulation and optimization algorithms. This model supports the extrapolation and simulation analysis of the parking lot's past, present, and future states.

[0066] See Figure 3The process of generating a parking space allocation matrix with priority and time windows begins with receiving and parsing user parking requests. Users submit parking requests through mobile applications or web interfaces, and the request data includes key fields such as vehicle type, reservation time period, and user status level. Upon receiving the request, the system immediately initiates the parsing process. The vehicle type field distinguishes different categories such as fuel vehicles, electric vehicles, and large vehicles; the reservation time period specifies the start and end times of the desired parking space usage; and the user status level field identifies different levels such as ordinary users, member users, or users with special permissions. Querying the historical parking database is performed simultaneously with request parsing. This database uses a distributed architecture to store users' past parking records. During queries, historical parking frequency, performance status, and outstanding fees are quickly retrieved based on the user ID, thereby calculating the user's credit score. User preferences are also extracted, including preferred parking areas, charging needs, and accessibility requirements—personalized options that collectively constitute important dimensions of the user profile.

[0067] Real-time parking space status is obtained by querying the parking lot's full-domain digital twin model. This model provides the latest real-time data on parking space occupancy, vehicle compatibility, and facility availability. The matching degree calculation algorithm is based on multi-factor weighted evaluation, considering multiple dimensions such as the matching degree between parking space size and vehicle size, the fit between electric vehicle spaces and charging needs, and the alignment of parking space location with user preferences. Each dimension is assigned a different weight coefficient and dynamically adjusted. The dynamic priority ranking algorithm employs a multi-level decision-making model. The first level performs initial screening based on vehicle type, prioritizing electric vehicles for spaces with charging stations and larger vehicles for spacious spaces. The second level adjusts based on user level, with higher-level users having priority. The third level introduces a credit scoring mechanism, giving higher priority scores to users with good credit. The final level comprehensively considers the real-time matching degree results to form the final priority ranking.

[0068] The time window is generated based on reservation time and historical dwell time data. The system analyzes the average dwell time of similar vehicles, current parking lot traffic patterns, and reservation time requirements to calculate a reasonable time window range for each allocation scheme. The time window includes the earliest start time and the latest end time allowed for parking, while reserving a certain buffer time to cope with emergencies. The parking space allocation matrix is ​​constructed using a two-dimensional table structure. Rows represent available parking space numbers, and columns include fields such as allocation time window, priority score, and matching degree coefficient. After the matrix is ​​generated, it enters a dynamic update loop, receiving the latest status data from the digital twin model in real time. When the parking space status changes or a new request arrives, the matrix is ​​immediately recalculated to ensure the optimality of the allocation result. The allocation results are displayed in real time through the visualization interface of the digital twin model and provide input data for the subsequent route planning module.

[0069] The entire allocation process employs a distributed computing architecture, with modules such as request parsing, database querying, and matching degree calculation running in parallel. Data exchange and state synchronization are achieved through message queues. An exception handling mechanism monitors conflicts and anomalies in real time during the allocation process. When multiple users request the same parking space or resources are insufficient, the system automatically resolves conflicts according to priority rules, ensuring the fairness and efficiency of the allocation process. Persistent storage of allocation results is implemented using a time-series database, fully recording the timestamp, decision basis, and execution result of each allocation decision. This historical data is then fed back into the historical parking database, forming a continuously optimized data loop that provides more accurate data support for subsequent allocation decisions.

[0070] For example, at 10:15 AM on a weekday, a Tesla Model 3 owner submits a parking request via the mobile app, with a reservation time from 2 PM to 5 PM that same day. This user is a Platinum member. Upon receiving the request, the system immediately initiates the parsing process, identifying the vehicle type as an electric vehicle, the reservation period as 3 hours, and the user's level as high. Almost simultaneously, the system queries the historical parking database and finds that the user has 12 parking records in the past three months, always leaving on time with no outstanding fees, a credit score of 92 out of 100, and preferences showing that the user frequently chooses charging spaces near elevators.

[0071] The digital twin model shows three available charging spaces in the eastern section of the parking lot: E-C12, E-C15, and E-C18. E-C12 is closest to the elevator (15 meters), E-C15 is in the middle (30 meters), and E-C18 is furthest away (45 meters). The matching algorithm begins running, first assessing space size compatibility (all three spaces meet the size requirements of the Model 3), then analyzing facility availability (all three spaces are equipped with fast charging stations), and finally calculating location convenience (E-C12 scores the highest). The dynamic priority ranking module initiates multi-level decision-making: the first level confirms the need for a charging space for the electric vehicle; the second level identifies the Platinum Member status; the third level applies a credit score of 92 points; and the fourth level calculates the location matching degree. The system calculates that E-C12 has the highest overall score of 8.7 out of 10, E-C15 scores 7.2, and E-C18 scores 6.5. The time window generation module begins working. Based on the 2 PM reservation start time, combined with the average charging time (2.5 hours) of similar vehicles in historical data and the current parking lot reservation status, it generates a time window for this allocation: allowing entry as early as 1:45 PM and exit as late as 5:15 PM, with a 15-minute buffer period. The parking space allocation matrix is ​​then updated, recording in the row of parking space E-C12: allocation time window 13:45-17:15, priority score 8.7, matching coefficient 0.93, and locking the parking space status to "pre-allocated".

[0072] However, at 1:30 PM, the digital twin model detected an anomaly: the vehicle originally parked in the adjacent parking space E-C12 had not left on time, and the system immediately started recalculating. The matching algorithm re-evaluated and found that E-C12's score had dropped to 7.8 points (because the occupancy of the adjacent parking space might cause inconvenience in entering and exiting), while E-C15's score had risen to 8.1 points. The system automatically adjusted the allocation scheme, reassigning parking space E-C15 to the user, and updating the time window (remaining unchanged), priority score (adjusted to 8.1), and matching coefficient (adjusted to 0.89) in the allocation matrix. The entire process was completed in seconds, and the user's mobile application received an update notification showing that the assigned parking space had changed from E-C12 to E-C15. At 1:50 PM, when the user's vehicle entered the parking lot, the digital twin model showed in real time that parking space E-C15 remained vacant, and the allocation matrix once again verified the effectiveness of the time window. When a vehicle passes through the entrance gate, the system performs a final confirmation, changing the parking space status from "pre-allocated" to "allocated," and records the actual start time of occupancy in the matrix. The entire allocation process goes through four stages: initial allocation, anomaly detection, dynamic adjustment, and final confirmation, demonstrating the system's decision-making capability in handling real-time changes.

[0073] The process of outputting a set of timestamped vehicle guidance routes begins with the collection of parameters from QR code hardware devices. These devices are installed in prominent locations at each parking space, and their communication protocol employs a dual-mode transmission mechanism of near-field communication and wireless network. The positioning accuracy parameters include centimeter-level coordinate offset correction values. After reading the device parameters, the system immediately initiates global path planning. Based on the target location determined by the three-dimensional spatial structure provided by the parking lot's full-domain digital twin model and the parking space allocation matrix, an improved ant colony algorithm is used for path optimization. During algorithm initialization, historical optimal path sequences and manually preset baseline paths are injected to form an initial population with prior knowledge. The historical paths are derived from a successful scheduling case library, and the manually preset paths are based on the site topology.

[0074] The path intersection strategy is dynamically adjusted based on parking space priority. Paths to higher-priority parking spaces enjoy priority passage. When different vehicle paths overlap, the system automatically adjusts the passage sequence according to priority. Real-time environmental data is continuously acquired through a digital twin model, including parameters such as lane congestion levels, areas of visual obstruction, and slippery road sections. This data is input into the path variability rate adjustment module, using the following adaptive adjustment formula:

[0075] ;

[0076] in: Represents the dynamic variability coefficient. This is the environmental complexity factor (calculated based on obstacle density and visibility). The path conflict weight is calculated based on the number of intersections and the difference in vehicle speed. This is a time urgency parameter (calculated based on the remaining reservation time). This formula ensures increased path exploration diversity in complex environments and maintains path stability in simple environments. During the generation of the timestamped vehicle guidance path set, the system discretizes the optimized path into a continuous sequence of coordinate points, with each coordinate point appended with the expected driving speed and a precise timestamp. The timestamp calculation comprehensively considers path length, turning angle, speed limits, and expected traffic flow conditions, employing a spatiotemporal consistency verification algorithm to eliminate time conflicts. The final output path set is encapsulated in JSON format, containing fields such as path ID, coordinate sequence, speed curve, and timestamp array.

[0077] The generation of the safe driving path set is based on real-time perception through multi-sensor fusion. LiDAR scans the surrounding environment at a 10Hz frequency, generating point cloud data and identifying the outlines of moving obstacles. Ultrasonic sensors supplement the detection of near-ground obstacles and transparent obstacles. A multi-level obstacle avoidance mechanism establishes a three-level response strategy: Level 1 response implements slight path deviation for obstacles beyond 3 meters; Level 2 response initiates speed adjustment and path replanning for obstacles within 1-3 meters; and Level 3 response triggers immediate braking and emergency avoidance for emergency obstacles within 1 meter. Path smoothing employs a segmented optimization algorithm, dividing the obstacle avoidance path into several sub-segments based on curvature change points. Fifth-order polynomial interpolation is performed on each segment to ensure continuous acceleration while satisfying vehicle kinematic constraints. The vehicle-side communication module maintains millisecond-level latency data exchange with the control center via 5G-V2X technology, uploading local perception data and receiving path adjustment commands in real time. After consistency verification, the generated safe driving path set is synchronously distributed to the control units of relevant vehicles, forming a complete vehicle scheduling and control closed loop.

[0078] For example, at 2:30 PM, a BYD Han electric vehicle enters the underground parking lot of the Science and Technology Park. The user has reserved charging parking space B-07 in Zone B, and the system immediately initiates the route planning process. The QR code hardware installed on the parking space pillar is recognized, and its communication protocol uses the 5.8GHz frequency band for transmission. The calibration parameters with a positioning accuracy of ±5 cm are read. The global route planning module calls the parking lot's digital twin model, which displays the real-time traffic status of each lane: main road A is clear, auxiliary road C connecting Zone B has a temporary construction area, and lane D has oncoming traffic. When initializing the route population using the improved ant colony algorithm, the historical best route successfully used by the user last month (via the east entrance - main road A - auxiliary road E - parking space B-07) is injected, along with two manually preset alternative routes (via the west entrance - roundabout - direct lane). During algorithm calculation, based on the high priority attribute of parking space B-07 (charging parking space reservation), the route intersection strategy is adjusted to grant the vehicle priority passage at the intersection of the main and auxiliary roads. Real-time environmental data is continuously updated through a digital twin model, showing that construction in the auxiliary lane C area has reduced the lane width by 40%. The system automatically increases the path variability coefficient to 0.85, prompting the algorithm to explore alternative routes. After three rounds of iterative calculations, the optimal path is finally generated: East Entrance - Main Road A - Backup Lane F - Auxiliary Lane G - Parking Space B-07. This path is discretized into a sequence of 87 coordinate points, each point with an attached driving speed value (15 km / h speed limit on the main road, 8 km / h speed limit within the lane) and a timestamp accurate to milliseconds, with an estimated total travel time of 4 minutes and 35 seconds.

[0079] As the vehicle began traveling along the guided path, the lidar detected an unexpected situation: a group of moving cleaning equipment (a cluster of robotic vacuum cleaners) appeared at corner F of the backup lane, and the ultrasonic sensors simultaneously detected an unmarked area of ​​standing water on the ground. A multi-level obstacle avoidance mechanism was immediately activated. The first-level response addressed the robotic vacuum cleaners 5 meters away, with the path offset calculation module generating a lateral adjustment of 0.8 meters to the right. The second-level response addressed the standing water area 2 meters away, reducing the system's speed from 8 km / h to 4 km / h and triggering path replanning to bypass the area. A segmented optimization algorithm divided the adjusted path into three consecutive segments: the first segment maintained the first 50 coordinate points of the original path; the second segment generated a smooth curve to bypass the standing water (using Bézier curve interpolation); and the third segment connected the last 20 coordinate points. The vehicle-side communication module uploaded the adjusted path data to the control center in real time via 5G-V2X technology, while simultaneously receiving avoidance commands from other vehicles. The final generated safe driving path contained 92 coordinate points, with a total duration adjusted to 4 minutes and 52 seconds. All data was synchronized to the vehicle control unit to complete this scheduling.

[0080] The process of outputting power distribution control commands begins with comprehensive monitoring of the parking lot's energy infrastructure. The system collects real-time data on the remaining power of energy storage units, including distributed lithium battery packs and flywheel energy storage devices, through smart meters and sensor networks. Simultaneously, it monitors the charging piles' operational status, covering various modes such as idle, charging, and maintenance. The charging pile status monitoring employs a multi-index evaluation system, including parameters such as output voltage stability, current fluctuation range, and interface connection reliability. This data is aggregated to the energy management center via a combination of power line communication and wireless transmission. The energy demand prediction module analyzes the vehicle movement trajectories provided by the safe driving path set, combining vehicle type identification (electric vehicle or fuel vehicle) and battery capacity information to calculate the expected energy consumption for different time periods. For example, if an electric vehicle plans to park and charge between 14:00 and 15:00, the system estimates that it will require 15kWh of electricity during that period based on its remaining battery capacity and charging characteristic curve. The power distribution network topology data is stored in a graph structure, including transformer nodes, distribution cabinet branches, cable connection relationships and capacity constraints. The dynamic load allocation algorithm is calculated based on network flow theory to ensure that the load on each line is balanced and does not exceed the rated capacity.

[0081] The power distribution control instruction set is generated using an event-driven mechanism. When a concentrated charging demand is detected in a certain area, the system generates a charging scheduling instruction to adjust the charging power allocation. When a line overload risk is detected, a power distribution switching instruction is generated to change the power supply route. See Table 1 for a portion of the power distribution control instruction set for a certain period.

[0082] Table 1: Power Distribution Control Command Table

[0083]

[0084] The QR code hardware implementation integrates a multi-color LED light source and a multi-mode communication module. The light source uses a combination of RGB three-primary-color LEDs, with different states distinguished by a color coding system: solid green indicates the parking space is vacant and available for reservation, blue flashing indicates the vehicle is charging, and rapid red flashing indicates the device needs maintenance. The communication module supports near-field communication using the ISO14443 protocol and wireless transmission using the IEEE802.11 protocol, enabling two-way authentication of vehicle identity and secure transmission of charging control commands. When a user scans the QR code, a multi-interaction process is triggered. After the phone's camera recognizes the QR code pattern, the application parses the embedded parking space number information and sends an authentication request to the server. Upon successful authentication, differentiated operations are performed based on the user type: ordinary users trigger a parking space locking mechanism, reserving the space for 15 minutes; electric vehicle users additionally initiate a charging pile docking process, opening the charging interface cover to prepare for power transmission; shared vehicle users activate a dispatch process, notifying nearby shared vehicles to move to the parking space.

[0085] The wireless charging control module works in conjunction with the QR code hardware. When an electric vehicle parks in a space equipped with a wireless charging transmitter, scanning the QR code triggers an authentication sequence. The receiver on the bottom of the vehicle performs positioning calibration with the ground transmitter, and power transmission begins after successful calibration. During charging, the energy state layer monitors the grid load in real time. If peak electricity demand is detected, the charging power is automatically reduced; if it is during off-peak hours, the power output is increased. The color of the indicator light reflects the charging status in real time: purple indicates high-power charging, yellow indicates low-power charging, and solid red indicates a charging fault requiring inspection. The energy state layer of the digital twin model is updated using an incremental update mechanism. Each time a power distribution control command is received, the energy state data of the corresponding device is updated. The update process includes three verification steps: first, verifying the consistency between the command execution result and the expectation; second, calculating the deviation between the actual energy consumption and the estimated value; and finally, adjusting the parameters of the energy prediction model. The entire energy management system forms a closed-loop control, continuously optimizing the power distribution strategy to improve energy utilization efficiency.

[0086] The end-to-end closed-loop optimization process is built upon continuous data collection and analysis. Actual vehicle trajectory data is collected via onboard GPS modules and inertial measurement units, recording latitude and longitude coordinates, velocity vectors, and acceleration data per second. Scheduling results data includes task completion time, energy consumption statistics, and abnormal event records. This raw data, after cleaning and time alignment, is stored in an analysis database, forming a time-series dataset covering several months of operational cycles. Feature extraction of successful obstacle avoidance paths focuses on decision sequences in complex environments, analyzing vehicle path selection patterns when facing dynamic obstacles. The extracted feature vectors include path smoothness indices (cumulative curvature change values), safety margin coefficients (statistical distribution of minimum distance to obstacles), and efficiency parameters (ratio of actual driving time to theoretical time). Feature analysis of efficient scheduling sequences emphasizes resource utilization efficiency, extracting features such as parking space turnover rate, charging pile usage balance, and path conflict resolution time. All feature values ​​are standardized to form a training sample set.

[0087] The neural network model employs a deep reinforcement learning architecture. The input layer receives a 142-dimensional feature vector, the hidden layer contains three fully connected layers using the ReLU activation function, and the output layer generates optimized suggestions for path planning parameters. During model training, an experience replay mechanism is used, storing historical decision data in a replay buffer. Each time, a batch of data is randomly sampled to update the network weights. The training objective is to minimize the loss function that deviates between the actual scheduling result and the expected target. The updated parameters are deployed using a canary release strategy. Initially, a small-scale test is conducted in a specific area of ​​the parking lot. The path planning schemes generated by the new parameters run in parallel with the original schemes. The optimal parameter set is selected by comparing the actual performance of the two sets of schemes. The testing period lasts for seven full working days, covering different weather conditions and traffic flow patterns. The final parameter version is distributed to all computing nodes across the entire domain through the configuration management system.

[0088] The integration of the wireless charging control module is reflected in the collaborative design of hardware and software. An electromagnetic induction coil is embedded at the bottom of each parking space as an energy transmitter, while a receiving coil array is installed in the vehicle chassis. The authentication process triggered by scanning a QR code includes two stages: identity verification and power negotiation. Identity verification exchanges digital certificates via near-field communication, while the power negotiation stage determines the optimal transmission power based on parameters provided by the vehicle's battery management system. The dynamic adjustment of charging power follows a multi-factor decision-making mechanism. The system monitors variables such as the total grid load, the remaining capacity of the energy storage unit, and the usage status of adjacent charging piles in real time. When the grid load exceeds a set threshold, it automatically enters energy-saving mode, gradually reducing the output power of each charging pile. When an abnormal temperature is detected in the cooling system of a charging pile, the system limits the maximum output power of that pile and triggers a maintenance alarm.

[0089] The indicator light system employs a composite coding scheme, conveying rich status information through a combination of color and flashing frequency: solid green indicates standby availability, gradually brightening and dimming blue indicates normal charging in progress, rapid flashing yellow indicates power limitation, and solid red indicates a fault requiring manual intervention. Special color combinations are used for advanced prompts, such as alternating flashing purple and blue indicating high-power charging mode activation, and slow flashing orange indicating a reserved / locked state. The entire optimized closed-loop operation relies on continuous data flow. Detailed operation logs are generated after each scheduled task, recording the complete lifecycle data from request reception to task completion. This log data, after anonymization, is input into the analysis pipeline, driving iterative updates of the neural network model to form a self-improving intelligent management system. The effective period for new parameters is gradually shortened, accelerating from initial monthly updates to weekly updates, ultimately achieving dynamic parameter adjustment based on real-time performance evaluation.

[0090] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A digital twin empowered parking lot digital universe management system, characterized in that, The method comprises the following steps: Generating a global digital twin model of the parking lot, performing multi-source perception and dynamic data fusion on the parking lot area, the multi-source perception including vehicle status recognition, vehicle identity recognition, and environmental parameter acquisition; Obtaining user parking requests and real-time parking space status, analyzing the requests and performing dynamic allocation based on the global digital twin model of the parking lot, and generating a parking space allocation matrix with priority and time window; Collecting device parameters of two-dimensional code hardware installed on parking spaces, performing multi-objective hierarchical path planning based on the global digital twin model of the parking lot and the parking space allocation matrix, and outputting a vehicle guide path set with timestamp; Based on the vehicle guide path set, constructing a real-time obstacle avoidance and dynamic adjustment system, generating a safe driving path set, and performing vehicle scheduling control according to the safe driving path set; Obtaining the power of the energy storage unit in the parking lot and the power distribution network data, combining the safe driving path set, performing energy coordination optimization and dynamic scheduling, outputting a power distribution control instruction set, and triggering the global digital twin model update of the parking lot; Obtaining vehicle scheduling execution logs, performing system knowledge evolution, extracting successful scheduling case features, and completing end-to-end closed-loop optimization; The process of generating the global digital twin model of the parking lot comprises: Collecting real-time images through parking space cameras, and using a deep learning model to recognize parking space occupancy status and vehicle identity information; Obtaining parking space vibration data through geomagnetic sensors to determine vehicle parking behavior and abnormal status; Collecting temperature, humidity, light intensity, and road surface status parameters through Internet of Things terminals; Based on the parking space occupancy status, the vehicle identity information, the vehicle parking behavior, and the environmental parameters, a parking lot multi-dimensional state atlas is constructed; Integrating the parking lot multi-dimensional state atlas and high-precision map data to generate the global digital twin model of the parking lot.

2. The digitally twinned enabled parking facility digitalization omnistore management system of claim 1, wherein, The process of generating a parking space allocation matrix with priority and time window comprises: Analyzing vehicle type, reservation time, and user level information in user parking requests; Querying a historical parking database to obtain user credit scores and preference settings; Based on the real-time parking space status in the global digital twin model of the parking lot, calculating the matching degree of each parking space; According to the vehicle type, the user level, the credit score, and the matching degree, dynamically sorting the priority; Generating the parking space allocation matrix containing the allocation time window and the priority.

3. The digitally twinned enabled parking lot digitalization omnistellar management system of claim 1, wherein, The process of outputting a vehicle guide path set with timestamp comprises: Obtaining the communication protocol and positioning accuracy parameters of the two-dimensional code hardware; Based on the global digital twin model of the parking lot and the parking space allocation matrix, performing global path planning; Using an improved ant colony algorithm to initialize the path population, injecting historical optimal paths and artificial preset paths; Adjusting the path intersection strategy according to the parking space priority; Dynamically adjusting the path mutation rate according to real-time environmental data; Outputting the vehicle guide path set containing coordinate sequence, driving speed, and timestamp.

4. The digitally twinned enabled parking facility digitalization omnistore management system of claim 3, wherein, The process of generating a safe driving path set comprises: Fusing the surrounding obstacles through laser radar and ultrasonic sensor perception; Establishing a multi-level obstacle avoidance mechanism to dynamically adjust the path offset according to the distance of the obstacles; Using a piecewise optimization algorithm to generate a smooth obstacle avoidance path; Real-time exchange path adjustment information through the vehicle end communication module; Generating the safe driving path set and synchronizing to each vehicle control unit.

5. The digitally twinned enabled parking facility digitalization omnistore management system of claim 1, wherein, The process of outputting the power distribution control instruction set includes: Monitoring the remaining capacity of the parking lot energy storage unit and the working state of the charging pile; estimating the energy demand of each period according to the safe driving path set; dynamically allocating load combined with the power distribution network topology; generating the power distribution control instruction set containing charging scheduling instructions and power distribution switching instructions; According to the power distribution control instruction set, update the energy state layer of the parking lot global digital twin model.

6. The digitally twinned enabled parking facility digitalization omnistore management system of claim 1, wherein, The two-dimensional code hardware assembled in the parking space contains multi-color light source and communication module; The light source distinguishes user level and operation state through color change and flashing mode; The communication module supports vehicle identity authentication and wireless charging control instruction transmission; Scanning the two-dimensional code hardware triggers the parking lock, charging start or shared vehicle scheduling process.

7. The digitally twinned enabled parking facility digitalization omnistore management system of claim 6, wherein, The two-dimensional code hardware supports nationwide networking of shared vehicle scheduling; scanning the two-dimensional code activates the cross-regional vehicle reservation function; cooperates with off-site parking space resources based on the parking lot global digital twin model; generates cross-regional path planning instructions and automatic parking control sequence.

8. The digitally twinned enabled parking facility digitalization omnistore management system of claim 1, wherein, The process of completing the end-to-end closed-loop optimization includes: Collecting vehicle actual driving trajectory and scheduling result data; extracting the feature vector of the successful obstacle avoidance path and efficient scheduling sequence; optimizing the path planning parameters through neural network model training; deploying the updated parameters to the global management system.

9. The digitally twinned enabled parking facility digital universe management system of claim 1, wherein, The system also includes a wireless charging control module assembled in the parking space; Scanning the two-dimensional code hardware triggers the charging pile docking authentication and power transmission; Based on the energy state layer of the parking lot global digital twin model, dynamically adjust the charging power; Indicate the charging state and abnormal early warning through the color change of the light source.

Images