Parking space digital dynamic sensing method based on Internet of Vehicles fusion

Abstract

The invention relates to the technical field of Internet of Vehicles perception, and discloses an Internet of Vehicles fused parking space digital dynamic perception method. The method comprises the steps that a parking space state model is constructed based on a target area of an Internet of Vehicles system, and the target area is associated with a parking space sensing process; constructing a sensing strategy model for controlling data acquisition and processing output of the parking space state model; carrying out coupling simulation on the two models according to the selected parking lot scenic spot, and obtaining a simulation result; judging whether sensing errors exist in the parking space sensing process or not according to the simulation result, and if yes, adjusting control parameters of the sensing strategy model; and on the basis of the adjusted parameters, coupling simulation is carried out again until a target simulation result with the sensing error within an allowable range is obtained. According to the method, by means of the multi-source information interaction advantage of the Internet of Vehicles, the perception strategy is optimized, the parking space perception accuracy and applicability are improved, the hardware cost is reduced, different parking lot scenes are adapted, and efficient utilization of parking space resources is assisted.

Description

Technical Field

[0001] This invention relates to the field of vehicle-to-everything (V2X) sensing technology, specifically to a method for digital dynamic sensing of parking spaces that integrates V2X technology. Background Technology

[0002] With the acceleration of urbanization and the continuous growth of motor vehicle ownership, parking difficulties have become a significant factor affecting urban traffic efficiency and residents' travel experience. Traditional parking management methods rely heavily on manual inspections or fixed sensor data collection, which have obvious limitations. Manual inspections not only consume a large amount of manpower but also suffer from information lag, making it difficult to reflect the dynamic changes in parking spaces in real time. When vehicles are densely entering and exiting parking lots, misjudgments can easily occur, leading to drivers not being able to obtain accurate parking information in a timely manner and increasing unnecessary mileage.

[0003] While fixed sensors such as geomagnetic and ultrasonic sensors can monitor parking space status to some extent, they suffer from high installation and maintenance costs and limited coverage. These sensors typically need to be deployed individually under or around each parking space, making construction difficult in large parking lots or open-air parking areas, and troubleshooting and replacement of equipment later on are also quite cumbersome. Furthermore, the sensing range of fixed sensors is limited by their physical location, making it difficult to handle complex scenarios such as temporary vehicle parking and unclear parking space lines, and increasing the likelihood of missed or false detections.

[0004] With the gradual development of vehicle-to-everything (V2X) technology, information interaction between vehicles and infrastructure, and between vehicles themselves, has become possible. However, the integration of existing V2X applications in parking space perception remains low. Most systems can only achieve simple transmission of parking space information, lacking dynamic modeling of parking space status and optimization of perception strategies. When parking scenarios change, such as increased traffic flow during peak hours or temporary traffic control altering parking space accessibility, existing systems cannot adjust perception parameters in a timely manner, leading to significant discrepancies between perception results and actual parking space status, thus affecting the efficient utilization of parking space resources. Furthermore, perception requirements vary across different parking scenarios, such as short-term parking needs in shopping mall parking lots and long-term parking needs in residential areas. Existing methods do not provide personalized perception designs for different scenarios, further reducing the accuracy and applicability of parking space perception. Summary of the Invention

[0005] The purpose of this invention is to provide a digital dynamic perception method for parking spaces that integrates vehicle-to-everything (V2X) technology, in order to solve the problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides a digital dynamic perception method for parking spaces based on vehicle-to-everything (V2X) integration, the method comprising: A parking space status model of the vehicle-to-everything (V2X) system is constructed based on the target area of ​​the V2X system, wherein the target area is the area related to the parking space perception process. A perception strategy model is constructed, which is used to control the data acquisition and processing output in the parking space status model; Based on the selected parking lot attractions, the perception strategy model and the parking space state model are coupled and simulated to obtain the corresponding simulation results; Based on the simulation results, determine whether there are any perception errors in the parking space perception process; In the presence of perception error, the control parameters in the perception strategy model are adjusted; Based on the adjusted control parameters, the perception strategy model and the parking space state model are re-coupled and simulated until the target simulation result is obtained. The target simulation result is the result that characterizes the parking space perception error as being within the allowable range.

[0007] Preferably, when the target area includes a parking space grid, vehicle sensors, and a vehicle communication module, the construction of a parking space status model based on the vehicle-to-everything (V2X) system for the target area includes: Construct a parking space grid model, a vehicle sensor model, and a vehicle communication model based on the target area of ​​the vehicle-to-everything (V2X) system. The input parameters of the vehicle sensor model are initialized using the state data of the parking space grid model; The vehicle communication model is configured with the output data of the vehicle sensor model to construct a parking space status model for obtaining the vehicle network system.

[0008] Preferably, the construction of the perception strategy model includes: Based on the data acquisition frequency and processing delay requirements, the perception strategy is discretized into an equivalent control algorithm model; The control algorithm model is integrated by parameter optimization components, and an update cycle is set to construct a perception strategy model.

[0009] Preferably, the step of performing coupled simulation on the perception strategy model and the parking space state model based on the selected parking lot attractions to obtain corresponding simulation results includes: Based on the selected parking lot attractions, the data acquisition rate of the parking space status model is adjusted by outputting control signals through the perception strategy model in order to obtain the real-time status of the target area. Once the parking space state model reaches a stable state, the data output accuracy of the parking space state model is controlled by the processing coefficients output by the perception strategy model, so as to realize the data interaction between the parking space state model and the perception strategy model and obtain the corresponding simulation results.

[0010] Preferably, determining whether there is a perception error in the parking space perception process based on the simulation results includes: If the simulation results meet the first or second preset conditions, it is determined that there is a perception error in the parking space perception process; The first preset condition is that during the coupled simulation parking perception process, the deviation between the state data of the parking space grid model in the parking space state model and the actual measurement data is greater than or equal to a preset deviation threshold. The second preset condition is that during the coupled simulation parking perception process, noise interference occurs in the output of the vehicle sensor model and the noise intensity is greater than a set threshold.

[0011] Preferably, the method further includes: At each time point, based on the initial occupancy matrix corresponding to each time point, the parking space status is corrected for uniformity to obtain the corrected occupancy matrix of the parking space at each time point. Determine the target occupancy rate that the parking space needs to maintain, control the parking space status perception based on the corrected occupancy matrix corresponding to the target occupancy rate, and measure the actual occupancy rate of the parking space through sensors; If the difference between the target occupancy rate and the actual occupancy rate is greater than or equal to a preset deviation threshold, the occupancy deviation matrix is ​​determined based on the corrected occupancy matrix corresponding to the target occupancy rate and the occupancy matrix corresponding to the actual occupancy rate. Based on the occupancy deviation matrix, the corrected occupancy matrix corresponding to the target occupancy rate is adjusted to obtain the adjusted occupancy matrix corresponding to the target occupancy rate; The adjusted occupancy matrix corresponding to the target occupancy rate is used as the new corrected occupancy matrix corresponding to the target occupancy rate, and the step of controlling the parking space status perception based on the corrected occupancy matrix corresponding to the target occupancy rate is returned until the difference between the target occupancy rate and the actual occupancy rate is less than the preset deviation threshold.

[0012] Preferably, the step of performing uniformity correction on the parking space status based on the initial occupancy matrix corresponding to each of the time points to obtain the corrected occupancy matrix of the parking space at each of the time points includes: For any given time point, based on the initial occupancy matrix corresponding to the time point, the parking space status perception is controlled, and real-time data of the parking space is obtained using vehicle networking devices, and the actual occupancy matrix is ​​extracted from the real-time data. Determine the standard deviation of occupancy uniformity based on the actual occupancy matrix; If the standard deviation of the occupancy uniformity is greater than or equal to a preset standard deviation threshold, the actual state matrix corresponding to the actual occupancy matrix is ​​determined based on a pre-constructed occupancy relationship model. The actual state matrix is ​​iterated using a state matrix iterative model to obtain the iterated state matrix corresponding to the actual state matrix; The iterative state matrix is ​​used as the initial occupancy matrix, and the initial occupancy matrix based on the time point is returned to control the parking space status perception step until the occupancy uniformity standard deviation is less than the preset standard deviation threshold. The iterative state matrix is ​​used as the corrected occupancy matrix.

[0013] Preferably, the method further includes: Based on vehicle movement trajectory data, analyze vehicle movement patterns within the area and generate a movement trajectory prediction model; Based on multi-source data, predict changes in regional parking demand and generate parking demand prediction results; Parking demand features are extracted based on time-varying network models to identify the spatiotemporal variation patterns of parking demand; Based on the scoring system, priorities and resource allocation ratios are adjusted to achieve a forward-looking parking resource layout; Adaptive execution of perception strategy adjustments ensures that the perception process anticipates changes in parking demand.

[0014] Preferably, the step of analyzing vehicle movement patterns within the area based on vehicle movement trajectory data and generating a movement trajectory prediction model includes: Anonymized vehicle location data from multiple time points within the region were collected, and data cleaning, outlier removal, and spatiotemporal feature extraction were performed. The region is divided into grids, and the vehicle density of each grid at different times is statistically analyzed to construct a regional movement heat map. Spatiotemporal trajectory clustering algorithm is applied to analyze vehicle movement trajectories and identify typical movement patterns and critical paths; Based on trajectory clustering results, a path traffic prediction function is established, which combines time factors to predict the traffic of each critical path at a specific time point.

[0015] Compared with the prior art, the beneficial effects of the present invention are: By constructing a parking space status model within a vehicle-to-everything (V2X) system, the advantages of multi-source information interaction brought by V2X technology can be fully utilized. Information related to parking space perception from various aspects, including vehicles and infrastructure, can be incorporated into the model construction process. This breaks through the limitations of information isolation in traditional perception methods, making the parking space status model more closely reflect the dynamic changes of actual parking scenarios. Compared to traditional perception methods that rely on fixed sensors, this model eliminates the need to deploy a large number of hardware devices for each parking space, reducing the cost of hardware installation and subsequent maintenance. Furthermore, the wide coverage of the V2X system expands the range of parking space perception, enabling comprehensive perception of parking scenarios of different sizes, such as large parking lots and open-air parking areas, avoiding the limited coverage problem of fixed sensors.

[0016] The construction of the perception strategy model provides a flexible control mechanism for the data acquisition and processing output of the parking space status model. Traditional perception methods often use fixed acquisition frequencies and processing methods, which cannot be adjusted according to changes in parking scenarios. However, the perception strategy model in this method can dynamically adjust parameters such as data acquisition frequency and data processing priority according to different parking scenario needs, making data acquisition more targeted and data processing more efficient. For example, during peak parking periods, when vehicles frequently enter and exit and parking space status changes rapidly, the perception strategy model can increase the data acquisition frequency to ensure timely capture of every change in parking space status; during off-peak parking periods, the acquisition frequency can be appropriately reduced to decrease unnecessary data redundancy and reduce the system's operating load.

[0017] By selecting parking lot locations for coupled simulation of the perception strategy model and parking space status model, the perception effect can be verified in advance before practical application, and potential problems in the perception process can be identified in a timely manner. Traditional methods often require the perception system to be put into actual use before problems can be discovered, and adjustments at this point require a lot of time and resources. However, this method, through simulation, can simulate the perception process under different parking scenarios in a virtual environment, quickly obtain simulation results, and determine whether there are perception errors. This greatly shortens the cycle of problem discovery and adjustment, and reduces the risks in practical applications.

[0018] When perception errors exist, the control parameters in the perception strategy model are adjusted and coupled simulations are re-performed until the target simulation result is obtained. This process forms a closed loop of continuous optimization. Traditional perception methods often struggle to accurately pinpoint the source of perception errors after they occur, and the adjustment process is largely haphazard. However, this method, by adjusting the control parameters of the perception strategy model, can specifically address perception errors caused by different reasons. For example, missed detections due to insufficient data acquisition frequency can be improved by increasing the acquisition frequency parameter; processing delays caused by unreasonable data processing algorithm priority settings can be optimized by adjusting the processing priority parameter. This precise adjustment method effectively reduces the deviation between the perception results and the actual parking space status, ensuring that the perception error is controlled within the allowable range and improving the accuracy of parking space perception.

[0019] This method achieves personalized perception adaptation for different parking scenarios by coupling simulations and parameter adjustments for different parking locations. Whether it's the short-term, high-frequency parking demand in shopping mall parking lots or the long-term, low-frequency parking demand in residential areas, simulations can be performed at corresponding scene points to adjust the perception strategy parameters to meet the needs of that scenario. This ensures the perception method maintains high performance across different scenarios, avoiding the problem of large differences in perception effects across different scenarios found in traditional methods. This further improves the applicability of parking space perception, helps increase the utilization efficiency of parking space resources, reduces ineffective driving caused by inaccurate parking space information, alleviates traffic congestion, and enhances the overall travel experience. Attached Figure Description

[0020] Figure 1 This is a schematic diagram illustrating the working principle of the vehicle-to-everything (V2X) fusion parking space digital dynamic perception method described in this invention. Figure 2 A flowchart for constructing a parking space status model; Figure 3 This is a flowchart for coupling simulation and data interaction. Detailed Implementation

[0021] 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.

[0022] Please see Figure 1This invention provides a digital dynamic perception method for parking spaces based on vehicle-to-everything (V2X) integration. The method includes: constructing a parking space state model for a target area of ​​the V2X system. This target area is defined as a physical or logical region related to the parking space perception process, including components such as a parking space grid, vehicle sensors, and a vehicle communication module. The parking space state model simulates the occupancy status and dynamic changes of parking spaces through mathematical modeling. A perception strategy model is responsible for controlling the logic of data acquisition and processing output; this model achieves adaptive adjustment through parametric design. Based on a selected parking location, the perception strategy model and the parking space state model are coupled and simulated. The coupling process involves data interaction and state synchronization, and the simulation results are used to evaluate the perception error. If a perception error exists, the control parameters in the perception strategy model are adjusted, such as the data acquisition frequency or processing delay, and the coupled simulation is repeated. The iterative process continues until the perception error is reduced to an acceptable range, thereby obtaining the target simulation result. This method achieves accuracy and real-time performance in parking space perception through model coupling and parameter optimization.

[0023] See Figure 2 The parking space status model is constructed based on the target area of ​​the vehicle-to-everything (V2X) system, which includes components such as a parking space grid, vehicle sensors, and vehicle communication modules. First, a parking space grid model is established, using discrete grid cells to represent the spatial distribution of parking spaces. Each grid cell corresponds to a parking space status, with status data including occupancy status (e.g., vacant or occupied), timestamp, and location identifier. This data is digitally mapped to reflect the physical state of the parking space in real time. Next, a vehicle sensor model is constructed, simulating sensor data acquisition behavior. Input parameters are initialized from the parking space grid model's status data; for example, sensor detection thresholds and sampling intervals are dynamically set according to the grid status to ensure the accuracy and adaptability of data acquisition. The sensor model output includes reading data and time-series information for subsequent processing. The vehicle communication model is responsible for data transmission and communication protocol processing. Its transmission strategy is configured by the output data of the vehicle sensor model, involving packet size, transmission frequency, and error correction mechanisms to ensure reliable data transmission in the V2X environment. By integrating the parking space grid model, vehicle sensor model, and vehicle communication model, the parking space status model forms a complete system representation capable of simulating the dynamic changes and interactions of parking spaces.

[0024] When constructing the sensing strategy model, the process begins by discretizing the sensing strategy into an equivalent control algorithm model based on the data acquisition frequency and processing latency requirements. The data acquisition frequency defines the rate at which sensor data is acquired, while the processing latency requirements specify the timeliness constraints of data processing. The control algorithm model is implemented using a state machine or rule engine, for example, based on IF-THEN logic conditions or PID control principles to adjust sensing behavior, ensuring the flexibility and responsiveness of the strategy. The core function of the control algorithm model is to transform the continuous sensing strategy into discrete executable instructions, which can dynamically adjust the data acquisition and processing flow. A parameter optimization element is then introduced, integrating the control algorithm model. This element includes an optimization function and a constraint handling module. The optimization function may search for the optimal parameter set based on gradient descent or heuristic algorithms, while the constraint handling module ensures that parameter adjustments meet system constraints such as resource availability or real-time requirements. Setting an update cycle is a key step in building the perception strategy model. The update cycle adjusts parameters periodically based on simulation time or event triggering mechanisms, such as triggering re-optimization at certain time intervals or when the system state changes exceed a threshold, thereby ensuring the continuous adaptability and efficiency of the perception strategy. The entire perception strategy model can adaptively adjust control parameters through a cyclic optimization process to cope with the dynamic changes and uncertainties in parking space perception.

[0025] The construction of the parking space state model and perception strategy model involves the coordination of multiple sub-steps. The initialization of the parking space grid model includes defining the grid resolution, cell attributes, and state update logic. The grid resolution is determined based on the size of the target area and the parking space density. Cell attributes include location coordinates, status identifiers, and historical data. The state update logic maintains model synchronization with the real environment through event-driven or timed polling. The initialization of the vehicle sensor model's input parameters depends on the state data of the parking space grid model. For example, when the grid cell state changes, the sensor model's detection threshold automatically adjusts to prioritize high-priority areas, and the sampling interval is dynamically shortened or extended based on occupancy rate to balance data accuracy and resource consumption. The sensor model's output data includes raw readings, filtered data, and timestamps. This data is used to configure the vehicle communication model's transmission strategy. For example, when the sensor output shows a high data volume, the transmission strategy increases the packet size or compression ratio to reduce network load, and the error correction mechanism selects retransmission or forward error correction based on data importance. The transmission strategy configuration of the vehicle communication model also considers network conditions such as bandwidth limitations or latency, and adjusts the transmission frequency and protocol parameters through adaptive algorithms to ensure efficient data transmission in the vehicle network environment. The discretization process of the perception strategy model involves decomposing high-level strategies into low-level operation instructions. The specific implementation of the control algorithm model may use a finite state machine to manage state transitions or a rule engine to execute conditional actions, so that the perception strategy can be executed accurately.

[0026] The integration of parameter optimization components is achieved through an algorithm module. The optimization function traverses the parameter space to find the optimal settings, such as using iterative methods to adjust the data acquisition frequency and processing delay parameters. The constraint processing module applies boundary conditions, such as minimum sampling intervals or maximum allowable delays, to prevent unrealistic parameter values. The update cycle is set based on system monitoring feedback, and the cycle length is dynamically adjusted according to the rate of environmental change. For example, the update cycle is shortened during peak hours to respond quickly to changes, and extended during low-activity periods to save resources. The entire implementation process emphasizes the interactivity and real-time performance of the model. The parking space status model provides basic data, while the perception strategy model applies control logic. The two are tightly coupled through data flow and signal exchange. This design ensures the accuracy and efficiency of parking space perception, laying the foundation for subsequent simulation and error adjustment. The implementation avoids using fixed values ​​or assumed effects, focusing instead on methodological processes and component interactions.

[0027] See Figure 3 Parking locations represent typical real-world parking scenarios, such as dense parking during peak hours in commercial areas or dispersed parking in residential areas at night. Selection criteria include historical parking data, traffic flow patterns, or pre-set experimental conditions, aiming to cover diverse operating environments to verify the robustness of the method. Based on the selected parking locations, the perception strategy model and the parking space status model initiate coupled simulations. The simulation engine coordinates the execution timing and data exchange between the two models. The perception strategy model outputs control signals, which are digital instructions or parameter adjustment commands, directly adjusting the data acquisition rate of the parking space status model. The data acquisition rate affects the activation frequency and sampling period of the embedded vehicle sensor model, thereby achieving dynamic acquisition of the real-time status of the target area. The adjustment mechanism is based on a feedback loop; for example, when the perception strategy detects accelerated state changes, the control signal increases the sampling rate to capture more refined data, and vice versa to save resources. The real-time status of the target area includes parking space occupancy bitmaps, sensor reading time series, and network communication status, which are continuously recorded for subsequent analysis.

[0028] During the operation of the parking space state model, reaching a steady state is a key judgment point. A steady state is defined as the fluctuation amplitude of the model's output variables consistently remaining below a preset threshold for a certain duration, indicating that the system has transitioned from an initial transient state to predictable behavior. Once a steady state is achieved, the perception strategy model further outputs processing coefficients, which are numerical parameters or algorithm selection identifiers used to control the data output accuracy of the parking space state model. Adjusting the data output accuracy is achieved by modifying the configuration of the data processing module, such as increasing the filtering order to improve noise suppression or using data compression algorithms to reduce redundancy, thereby balancing accuracy and processing load. Data interaction between the parking space state model and the perception strategy model relies on a shared memory interface or message passing mechanism. Interaction data includes state queries, parameter updates, and event notifications, ensuring that the two models evolve synchronously on the simulation timeline. The execution of the coupled simulation produces corresponding simulation results, which are output in the form of structured log files or database records, containing timestamps, model state snapshots, performance indicators, and raw sensor data. The simulation results are used to comprehensively evaluate the dynamic behavior of the perception process.

[0029] Based on simulation results, the process for determining whether a perception error exists during parking space perception is based on logical evaluation of preset conditions. The first preset condition checks the deviation between the state data of the parking space grid model in the parking space state model and the actual measurement data. The actual measurement data comes from a calibration dataset of real sensor deployments or a high-precision reference system. The deviation is calculated using Euclidean distance or absolute error measurement. The preset deviation threshold is set according to the fault tolerance requirements of the application scenario; for example, the threshold is lower in a precise navigation scenario. If the deviation exceeds or equals the threshold, the condition is triggered, indicating the existence of a perception error. The second preset condition monitors noise interference in the output of the vehicle sensor model. Noise interference is identified and quantified through signal processing algorithms such as Fourier analysis or statistical variance. The noise intensity is expressed as signal-to-noise ratio or amplitude value. The threshold is set based on sensor specifications and the environmental noise baseline. When the noise intensity continuously exceeds the threshold, the condition is also triggered, indicating a perception error. The determination logic is integrated into the simulation post-processing module, which automatically scans the result data and marks abnormal events. If any condition is met, the system confirms the existence of a perception error and triggers the subsequent parameter adjustment process; otherwise, the simulation continues or the final result is output. The entire implementation process emphasizes objective measurement of conditions and automated decision-making, avoiding subjective judgment, thereby ensuring the reliability and consistency of error detection.

[0030] The technical details of coupled simulation involve the meticulous design of multiple sub-processes. Parking lot attractions are selected using clustering algorithms or random sampling from a scene library to ensure diversity across spatial and temporal dimensions, including different geographical locations, weather conditions, or event-driven scenarios such as sporting events. The simulation engine manages the model's time progression and event scheduling, employing discrete event simulation or time-stepping mechanisms to coordinate the execution of the perception strategy model and the parking space status model. The specific implementation of controlling the data acquisition rate includes modifying the timer interrupt frequency of the sensor model or adjusting the data buffer size. Real-time status acquisition is achieved by periodically querying model state variables and recording them in a log. Steady-state detection is achieved by monitoring the sliding window variance of key output variables such as parking space occupancy rate; when the variance is below a threshold for several simulation time units, a steady state is considered reached. When processing coefficients to control data output accuracy, different algorithm instances may be selected, such as switching between a Kalman filter and a moving average filter, or adjusting the data quantization bit depth. Data interaction mechanisms ensure consistency between models, such as using atomic operations or transaction locks to prevent data races; simulation results are recorded in a standardized format for easy parsing, including CSV files or SQL databases, containing fields such as timestamps, model IDs, parameter values, and error metrics.

[0031] The implementation of the perception error judgment conditions emphasizes configurability and scalability. The deviation calculation in the first preset condition supports multiple measurement methods, and the preset deviation threshold is stored in a configuration file for dynamic adjustment. The input of actual measurement data is handled through a data adapter interface, processing input sources of different formats. The noise interference detection in the second preset condition integrates signal processing library functions. The threshold setting can be calibrated according to the sensor type; for example, ultrasonic sensors and optical sensors have different noise characteristics. The judgment module automatically performs scans and generates reports, listing the time points and detailed context of the trigger conditions, providing a basis for subsequent debugging. The entire process is orchestrated through scripts or a workflow engine to achieve end-to-end automated simulation and error analysis, ensuring efficient execution and repeatability of the method.

[0032] After confirming the existence of perception errors during the parking space sensing process, an optimization procedure is initiated. The control parameters in the perception strategy model are optimized using iterative search algorithms, such as gradient descent or Bayesian optimization methods. These algorithms systematically adjust parameters such as data acquisition frequency, filtering coefficients, or processing latency to minimize the perception error index. Initial values ​​for the control parameters are extracted from the current perception strategy model. The optimization objective function is defined as a weighted sum of error metrics (such as deviation or noise intensity), with constraints including system resource limitations and real-time requirements. The optimization process is executed in a simulation environment, selecting the optimal settings by repeatedly evaluating the performance of parameter combinations. Based on the optimized control parameters, the perception strategy model and the parking space state model are re-coupled and simulated. The re-simulation uses updated parameter configurations, and the simulation process is consistent with the initial run but uses new parameters. Iterative execution continues until the target simulation result is obtained. The target simulation result requires the parking space perception error to be reduced to an allowable range, which is predefined by the accuracy requirements of the application scenario.

[0033] At each time point, uniformity correction is performed based on the initial occupancy matrix corresponding to that time point. The initial occupancy matrix is ​​a two-dimensional array representing the binary occupancy status of parking spaces (1 indicates occupied, 0 indicates vacant). Uniformity correction aims to balance the occupancy distribution of the parking area by adjusting the matrix values ​​through mathematical transformations. For each time point, the initial occupancy matrix at that time point is used to control the perception of parking space status. The perception process activates the vehicle-to-everything (V2X) sensor network to collect real-time data, including vehicle detection signals and location information. The actual occupancy matrix is ​​extracted from the real-time data. The extraction algorithm involves data parsing and state mapping, such as converting sensor readings into matrix element values. Based on the actual occupancy matrix, the standard deviation of occupancy uniformity is calculated. This statistic measures the dispersion of the parking space occupancy distribution, and the calculation formula is:

[0034] in: Indicates the standard deviation of occupancy uniformity. This is the total number of parking spaces. It is the first The occupancy status of each parking space (0 or 1). This refers to the average occupancy rate. If the standard deviation of occupancy uniformity is greater than or equal to a preset standard deviation threshold (set based on area capacity and ideal distribution), the actual state matrix corresponding to the actual occupancy matrix is ​​determined based on a pre-constructed occupancy relationship model. This model uses regression or a neural network to map the occupancy matrix to the state space. The actual state matrix captures deep-seated parking characteristics such as clustering trends or abnormal patterns. An iterative state matrix model is used to iterate the actual state matrix. This model applies numerical methods such as Jacobi iteration or conjugate gradient method to gradually adjust the state values ​​until convergence, resulting in the iterated state matrix. The iterated state matrix is ​​then used as the new initial occupancy matrix, and the process returns to the step of controlling parking space status perception. This loop continues until the standard deviation of occupancy uniformity is less than the preset standard deviation threshold. Finally, the iterated state matrix is ​​output as the corrected occupancy matrix.

[0035] The target occupancy rate for parking spaces is determined, derived from urban planning or real-time demand forecasting, and expressed as a percentage. Based on the corrected occupancy matrix corresponding to the target occupancy rate, the parking space status perception is controlled. The perception system is configured with a sensor network operating according to specified parameters, and measures the actual occupancy rate of parking spaces using sensors, calculated from real-time data. If the difference between the target occupancy rate and the actual occupancy rate is greater than or equal to a preset deviation threshold (set according to operational precision), the difference for each parking space is calculated based on the corrected occupancy matrix corresponding to the target occupancy rate and the occupancy matrix corresponding to the actual occupancy rate, and an occupancy deviation matrix is ​​constructed. This occupancy deviation matrix is ​​a real number matrix, where each element represents the magnitude of the deviation for the corresponding parking space. Based on the occupancy deviation matrix, the corrected occupancy matrix corresponding to the target occupancy rate is adjusted using optimization algorithms such as least squares fitting or heuristic rules to modify matrix elements to reduce the deviation, resulting in an adjusted occupancy matrix corresponding to the target occupancy rate. This adjusted occupancy matrix is ​​then used as the new corrected occupancy matrix corresponding to the target occupancy rate, and the process returns to the parking space status perception control step. This process is repeated until the difference between the target occupancy rate and the actual occupancy rate is less than the preset deviation threshold, thereby achieving dynamic calibration and error compensation.

[0036] The entire implementation process emphasizes automation and iterative optimization. Parameter optimization employs closed-loop control, uniformity correction is based on statistical evaluation, and occupancy rate adjustment is achieved through matrix operations. These steps ensure that the parking space perception system can adapt to environmental changes and maintain high accuracy. The implementation avoids assuming performance data and instead focuses on methodological processes and algorithmic interactions, meeting the detailed and operable requirements of patent applications.

[0037] The initial occupancy matrix corrects for the uniformity of parking space status. This can be illustrated using a hypothetical parking area example. Consider a small parking lot with 9 parking spaces (numbered P1 to P9), arranged in a 3x3 grid. At each time point, the occupancy status of the parking spaces is represented by the initial occupancy matrix, a two-dimensional array where a value of 1 indicates occupancy and 0 indicates vacancy. For a specific time point, the initial occupancy matrix may exhibit an uneven distribution, such as dense occupancy in the central area and vacancy in the peripheral areas. This unevenness affects the effective utilization of parking resources. Based on the initial occupancy matrix corresponding to that time point, the control system initiates parking space status sensing, acquiring real-time data through vehicle-to-everything (V2X) devices deployed within the parking lot (such as geomagnetic sensors or cameras). Real-time data includes vehicle presence signals detected by sensors, timestamps, and location coordinates. The actual occupancy matrix is ​​extracted from this data. The extraction process involves data cleaning and status mapping, such as converting sensor signals into matrix element values ​​to form a matrix reflecting the current true occupancy status.

[0038] Based on the actual occupancy matrix, the standard deviation of occupancy uniformity is calculated. This value quantifies the dispersion of parking space occupancy distribution. Assume the state of the actual occupancy matrix is ​​as shown in Table 1 below, which shows the occupancy status of 9 parking spaces.

[0039] Table 1: Actual Parking Space Occupancy Matrix

[0040] Based on the data in the table above, the occupancy status value is 0 or 1, the average occupancy rate μ is 5 / 9 ≈ 0.555, and the standard deviation σ of occupancy uniformity is calculated as the square root of the average of the squares of the deviations of each status value from the mean. If the calculated σ is greater than or equal to the preset standard deviation threshold (assuming the threshold is 0.2), uniformity correction is required. The preset standard deviation threshold is set according to the parking lot's design capacity and ideal distribution, and is used to determine whether the distribution is too concentrated or dispersed. Based on a pre-built occupancy relationship model, the actual state matrix corresponding to the actual occupancy matrix is ​​determined. The occupancy relationship model may be trained using machine learning methods, mapping the occupancy matrix to a state space containing hidden features (such as vehicle flow patterns or area attractiveness). The actual state matrix captures the deep characteristics of parking spaces, such as some parking spaces being vacant but having high demand potential. An iterative state matrix model is used to iterate the actual state matrix. The iterative model uses numerical optimization techniques (such as relaxation iteration) to gradually adjust the state values ​​to promote uniform distribution. The iteration process continues until the state matrix converges.

[0041] The iteratively obtained state matrix is ​​used as the new initial occupancy matrix, and the system returns to the step of controlling parking space status perception. It then reacquires real-time data and calculates a new standard deviation for occupancy uniformity. This loop continues until σ is less than a preset standard deviation threshold, indicating that the occupancy distribution has reached a sufficiently uniform level. Finally, the iteratively obtained state matrix is ​​output as the corrected occupancy matrix for subsequent parking management and guidance. Throughout the process, the vehicle-to-everything (V2X) device continuously provides real-time data to ensure that the correction is based on the latest information, while iterative optimization automatically handles the problem of uneven distribution. This implementation demonstrates the complete process of uniformity correction through a concrete example, highlighting the characteristics of data-driven and adaptive adjustment, avoiding subjective intervention and relying on mathematical calculations and model reasoning.

[0042] This paper presents a method for proactively allocating parking resources based on vehicle movement trajectory data and analysis. The method is illustrated using a medium-sized city commercial district as an example. The commercial district covers approximately 2 square kilometers and includes multiple shopping malls, office buildings, and entertainment facilities, experiencing high daily vehicle traffic with a clear tidal pattern. Anonymized vehicle location data is collected at multiple time points within the area. Data sources include in-vehicle GPS devices, mobile applications, and roadside units, spanning a continuous 30-day period, covering weekdays, weekends, and holidays. Anonymization removes personal identifiers and encrypts location information to ensure privacy compliance. Data cleaning, outlier removal, and spatiotemporal feature extraction are performed. Data cleaning removes missing values ​​and duplicate records, outlier removal filters out trajectory points with sudden speed changes or location jumps, and spatiotemporal feature extraction involves calculating parameters such as vehicle speed, acceleration, direction of travel, and dwell time.

[0043] The area was divided into 500m x 500m grids, and vehicle density in each grid was statistically analyzed at different times, with the statistics performed hourly, recording the number of vehicles within each grid. A regional movement heatmap was constructed, using color gradients to represent density levels: dark red for high-density areas and blue for low-density areas. Visualization tools helped identify traffic flow aggregation points and idle areas. A spatiotemporal trajectory clustering algorithm was applied to analyze vehicle movement trajectories. The algorithm employed an improved variant of DBSCAN, considering both spatial proximity and temporal continuity. Typical movement patterns and critical paths were identified. Typical movement patterns included morning peak commuter flow (from residential areas to commercial areas), midday short-distance shopping flow, and evening peak discrete flow. Critical paths referred to frequently occurring vehicle routes, such as main roads and connecting roads.

[0044] Based on trajectory clustering results, a path traffic prediction function is established. This function combines time series analysis with machine learning models (such as ARIMA or LSTM networks). Input variables include time, day of the week, weather conditions, and special event markers. It predicts traffic flow on key paths at specific time points, and the output is an estimate of the number of vehicles in the next hour or less. Regional parking demand changes are predicted based on multi-source data, including historical parking records, traffic flow sensor readings, business activity schedules, and social media event information. Parking demand prediction results are generated and output as a probability distribution, indicating the intensity of parking demand in different sub-regions in the future. Parking demand features are extracted based on a time-varying network model, which simulates changes in road network capacity and connectivity at different times. The model identifies spatiotemporal patterns of parking demand, including weekday morning peak demand concentrated in office areas, weekend afternoon demand shifting to commercial areas, and nighttime demand reaching its lowest point.

[0045] The scoring system calculates adjustment priorities and resource allocation ratios. It employs a multi-criteria decision-making method, considering factors such as demand urgency (based on the difference between predicted demand and current capacity), resource availability (number of vacant parking spaces), economic benefits (parking rates), and user satisfaction (historical complaint data). An adjustment priority score is calculated for each sub-area; higher scores indicate higher priority. Resource allocation ratios are weighted according to these scores, distributing parking guidance signals and variable parking space markers. This enables proactive parking resource layout, including dynamic adjustments to parking rates, opening backup parking lots, modifying lane directions, and pushing vacant parking space information via an app. Adaptive execution of perception strategy adjustments is based on real-time feedback loops. When the deviation between predicted and actual demand exceeds a threshold, parameters are automatically recalculated to ensure the perception process adapts to changes in parking demand.

[0046] The entire implementation process was demonstrated through actual data flow from a commercial area. Vehicle movement trajectory analysis revealed commuting patterns with increased vehicle entry into the commercial area between 7-9 AM, a short shopping peak between 2-4 PM, and a departure peak between 5-7 PM. The path flow prediction function accurately captured these patterns, for example, predicting that main road traffic at 8 AM on Monday would be 40% higher than the same time on Sunday. Parking demand prediction results showed that the office area's demand peaked at 85% between 8-10 AM on weekdays, while the commercial area maintained high demand between 2-6 PM on weekends. The time-varying network model identified a 20% capacity decrease in connecting roads during peak hours. The scoring system assigned a priority score of 0.8 (out of 1.0) to the office area during the morning peak, directing 70% of resource allocation signals to this area. After the implementation, the system adaptively adjusted its perception strategy, such as increasing sensor sampling frequency in high-demand areas, thereby optimizing overall parking resource utilization. This method, through a concrete example, demonstrates the complete chain from data collection to implementation, highlighting data-driven and adaptive characteristics, aligning with the development trend of intelligent parking management.

[0047] 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.

[0048] 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 method for digital dynamic perception of parking spaces in the Internet of Vehicles, characterized in that, The method comprises: constructing a parking space state model of a vehicle networking system based on a target area of the vehicle networking system, the target area being an area related to a parking space perception process; constructing a perception strategy model for controlling data collection and processing output in the parking space state model; coupling simulation of the perception strategy model and the parking space state model according to selected parking scene points to obtain corresponding simulation results; determining whether there is a perception error in the parking space perception process according to the simulation results; adjusting control parameters in the perception strategy model in the case of a perception error; re-coupling simulation of the perception strategy model and the parking space state model based on the adjusted control parameters until a target simulation result is obtained, the target simulation result being a result indicating that the parking space perception error is within an allowable range.

2. The IoT converged parking space digitization dynamic perception method according to claim 1, characterized in that, In the case where the target area comprises a parking space grid, a vehicle sensor and a vehicle communication module, the constructing a parking space state model of a vehicle networking system based on a target area of the vehicle networking system comprises: constructing a parking space grid model, a vehicle sensor model and a vehicle communication model based on the target area of the vehicle networking system; initializing input parameters of the vehicle sensor model using state data of the parking space grid model; configuring a transmission strategy of the vehicle communication model using output data of the vehicle sensor model to construct the parking space state model of the vehicle networking system. 3.The IoT converged parking space digitization dynamic sensing method of claim 2, wherein, The constructing a perception strategy model comprises: discretizing a perception strategy into equivalent control algorithm models according to data collection frequency and processing delay requirements; integrating the control algorithm models through a parameter optimization element and setting an update period to construct the perception strategy model. 4.The IoT converged parking space digitization dynamic sensing method of claim 1, wherein, The coupling simulation of the perception strategy model and the parking space state model according to selected parking scene points to obtain corresponding simulation results comprises: adjusting a data collection rate of the parking space state model through a control signal output by the perception strategy model according to the selected parking scene points to obtain a real-time state of the target area; controlling data output accuracy of the parking space state model through a processing coefficient output by the perception strategy model when the parking space state model reaches a stable state to realize data interaction between the parking space state model and the perception strategy model and obtain corresponding simulation results.

5. The IoT converged parking spot digitization dynamic perception method of claim 1, wherein, The determining whether there is a perception error in the parking space perception process according to the simulation results comprises: determining that there is a perception error in the parking space perception process in the case where the simulation results meet a first preset condition or a second preset condition; the first preset condition being that, in the coupled simulation parking perception process, a deviation between state data of a parking space grid model in the parking space state model and actual measurement data is greater than or equal to a preset deviation threshold; the second preset condition being that, in the coupled simulation parking perception process, noise interference appears in the vehicle sensor model output and the noise intensity is greater than a set threshold.

6. The IoT converged parking spot digitization dynamic perception method of claim 1, wherein, The method further comprises: At each time point, based on the initial occupancy matrix corresponding to each time point, the parking space state is uniformly corrected to obtain the corrected occupancy matrix of the parking space at each time point; Determine the target occupancy rate required to maintain the parking space, control the parking space state perception based on the corrected occupancy matrix corresponding to the target occupancy rate, and measure the actual occupancy rate of the parking space through the sensor; In the case where the difference between the target occupancy rate and the actual occupancy rate is greater than or equal to a preset deviation threshold, determine the occupancy deviation matrix according to the corrected occupancy matrix corresponding to the target occupancy rate and the occupancy matrix corresponding to the actual occupancy rate; Based on the occupancy deviation matrix, adjust the corrected occupancy matrix corresponding to the target occupancy rate to obtain an adjusted occupancy matrix corresponding to the target occupancy rate; Take the adjusted occupancy matrix corresponding to the target occupancy rate as the new corrected occupancy matrix corresponding to the target occupancy rate, and return to the step of controlling the parking space state perception based on the corrected occupancy matrix corresponding to the target occupancy rate until the difference between the target occupancy rate and the actual occupancy rate is less than the preset deviation threshold.

7. The IoT converged parking space digitization dynamic perception method according to claim 6, characterized in that, The method further comprises: Based on the vehicle moving trajectory data, analyze the vehicle moving mode in the region, and generate a moving trajectory prediction model; Based on multi-source data, predict the change of regional parking demand, and generate a parking demand prediction result; Based on a time-varying network model, extract parking demand features, and identify the spatio-temporal variation law of parking demand; Based on a scoring system, calculate the adjustment priority and resource allocation ratio to realize forward-looking parking resource layout; Adaptively execute the perception strategy adjustment to ensure that the perception process is pre-adapted to the change of parking demand. The method further comprises: 8.The IoT converged parking space digitization dynamic sensing method of claim 1, wherein, Collect anonymized vehicle location data at multiple time points in the region, perform data cleaning, outlier removal, and spatio-temporal feature extraction; ​ ​ ​ ​ ​ 9.The IoT converged parking space digitization dynamic sensing method of claim 8, wherein, ​ ​ The region is divided into grids, and the vehicle density of each grid at different times is counted to construct a region moving heat map; A space-time trajectory clustering algorithm is applied to analyze the vehicle moving trajectory, identify typical moving modes and key paths; Based on the trajectory clustering results, a path flow prediction function is established to predict the flow of each key path at a specific time point combined with the time factor.

Images