A real-time reservation and dispatch system for shared parking

Abstract

The application discloses a kind of real-time reservation and dispatching systems for shared parking, it is related to intelligent transportation and wisdom city management technical field, including demand analysis module, parking scene type, time requirement and priority information are extracted from user parking request data, generate standardized demand description by dynamic weight distribution, obtain user demand analysis result, process decomposition module, according to parking scene type and time requirement in user demand analysis result, decompose parking dispatching process into demand receiving, resource matching and path planning business unit, determine business unit set;The real-time reservation and dispatching system for shared parking, through dynamic weight distribution and real-time response optimization, ensure efficient resource allocation and path planning, significantly improve parking dispatching efficiency and user experience.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent transportation and smart city management, and particularly relates to a real-time reservation and scheduling system for shared parking. BACKGROUND

[0002] Parking scheduling, as a key link in the field of smart city and intelligent transportation, plays an irreplaceable role in improving urban transportation efficiency, optimizing resource utilization, and improving user experience. With the acceleration of urbanization, parking demand is becoming increasingly complex, involving multiple scenarios such as parking in business districts during peak hours, parking in residential areas at night, and temporary parking for special events, etc. Traditional scheduling methods have been difficult to meet the dynamic and personalized needs.

[0003] Current solutions usually rely on fixed process design, lacking flexibility, especially when dealing with multi-scenario integration and real-time changes, often leading to low efficiency or resource waste due to the inability to quickly adapt to different user needs. For example, existing systems may not effectively coordinate multiple parking requests during peak hours, resulting in long waiting times for users or uneven allocation of parking resources. The core challenge lies in how to achieve dynamic combination and real-time adjustment of scheduling processes. Dynamic combination requires breaking down complex parking scheduling processes into standardized business units, such as demand reception, resource matching, path planning, etc., and flexibly combining them according to specific scenarios. However, this decomposition and combination need to strike a balance between real-time and accuracy. Too complex a combination logic can lead to prolonged system response time, while too simple a combination cannot meet individual needs. For example, a commercial complex's parking needs may involve visitor reservations, VIP priority parking, and temporary parking, etc. If the system cannot quickly match resources and coordinate time, users may queue at the entrance or miss the reservation period. In addition, the lack of real-time monitoring and exception handling mechanisms further exacerbates the problem. Based on dynamic combination, if an exception occurs in a certain link, such as path planning being blocked by sudden traffic conditions, the system needs to quickly identify and adjust the process, but existing technologies are difficult to effectively handle such exceptions while ensuring real-time performance. SUMMARY

[0004] The present application aims to provide a real-time reservation and scheduling system for shared parking, designing a flexible scheduling process orchestration mechanism that supports breaking down parking scheduling processes into standardized business units and dynamically combining them according to different scenarios and user needs, while achieving real-time monitoring and exception handling of the process.

[0005] To achieve the above object, the present application provides the following technical scheme: A real-time reservation and scheduling system for shared parking, comprising: a demand analysis module that extracts parking scene type, time requirement and priority information from user parking request data, generates a standardized demand description through dynamic weight distribution, and obtains user demand analysis results; a process decomposition module that decomposes the parking scheduling process into demand receiving, resource matching and path planning business units according to the parking scene type and time requirement in the user demand analysis results, and determines a business unit set; a resource acquisition module that acquires parking lot resource state and traffic flow data for the resource matching unit, and obtains resource matching input data; an initial scheduling generation module that generates an initial scheduling process through resource state analysis and fusion accuracy check according to the resource matching input data, and determines an initial scheduling process configuration scheme if the fusion accuracy check deviation value is lower than a preset scheduling response speed threshold, wherein the fusion accuracy check is performed in units of resource numbers, the state values of the same resource item in different data sources are retrieved, and the deviation of the two indicators of idle parking space number and average speed is calculated, the difference between the idle parking space numbers of the two data sources is taken as an absolute value, and if the difference is less than or equal to 1, it is considered that the deviation meets the accuracy requirement; at the same time, the difference between the average speeds of the two data sources is calculated, and if the difference is less than or equal to 5 kilometers per hour, it is considered that the speed deviation meets the accuracy requirement; the scheduling response speed threshold is that the idle parking space number deviation does not exceed 1 unit, and the average speed deviation does not exceed 5 kilometers per hour; a resource state update module that updates the real-time resource state data set through dynamic weight distribution and abnormal data elimination if the fusion accuracy check deviation value is higher than the preset scheduling response speed threshold, and obtains adjusted resource matching input data; a path optimization module that generates an optimal path planning scheme through path smoothing processing and real-time response delay optimization according to the adjusted resource matching input data, and determines a parking resource allocation result; a conflict detection module that generates an alternative scheduling scheme through abnormal trigger threshold judgment if a path blocking or resource conflict event is detected for the parking resource allocation result; a scheduling adjustment module that updates the scheduling process through event priority evaluation and resource reallocation according to the alternative scheduling scheme, and determines a final scheduling execution scheme in combination with dynamic response adjustment; and a process tracking module that obtains an execution state from the final scheduling execution scheme, tracks the user parking process, and generates a scheduling completion confirmation if the real-time response delay meets the preset scheduling response speed threshold.

[0006] Preferably, the demand analysis module extracts parking scene types, time requirements and priority information from user parking request data, generates a standardized demand description through dynamic weight allocation, and obtains user demand analysis results including: parsing parking scene types, time requirements and priority information from user request data, using a structured data extraction method to obtain an initial demand data set; if the initial demand data set includes missing values, fill in the missing values through a pre-set default value filling rule to obtain a complete demand data set; according to the parking scene types and time constraints in the complete demand data set, a decision tree algorithm is used to generate scene classification labels to obtain classified demand descriptions; through the classified demand descriptions, combined with priority information and weighted allocation rules, the demand weight values are calculated to obtain a weighted demand vector; if the weight value of the weighted demand vector exceeds a pre-set threshold, the parking resources are preferentially allocated, and a resource allocation scheme is generated; according to the resource allocation scheme and the time constraint conditions, a scheduling algorithm is used to optimize the parking resource allocation sequence to obtain the final parking allocation result; through the final parking allocation result and the user identification information, a personalized parking demand description is generated to obtain user notification data.

[0007] Preferably, the process decomposition module decomposes the parking scheduling process into demand receiving, resource matching and path planning business units according to the parking scene types and time requirements in the user demand analysis results, and determines the business unit set including: obtaining parking scene types and time requirements from user demand analysis results, extracting demand feature data using a structured analysis method, and generating an initial demand set; if the parking scene types or time requirements in the initial demand set are missing, fill in the missing values through a pre-set default rule to obtain a complete demand set; according to the parking scene types in the complete demand set, a K-means clustering algorithm is used to divide the scene categories to generate classified scene labels; through the classified scene labels and time requirements, combined with traffic pattern recognition, a random forest algorithm is used to predict traffic flow distribution to obtain a traffic pattern set; if the traffic distribution in the traffic pattern set exceeds a pre-set threshold, the resource matching sequence is preferentially adjusted to generate an optimized resource allocation scheme; according to the optimized resource allocation scheme and the path planning requirements, a Dijkstra algorithm is used to calculate the optimal parking path to obtain the path planning result; through the path planning result and the complete demand set, a scheduling instruction set is generated to determine the final parking scheduling scheme.

[0008] Preferably, the resource acquisition module acquires the parking lot resource state and traffic flow data for the resource matching unit to obtain resource matching input data, including: acquiring the parking lot resource state and traffic flow data in real time, integrating multi-source data through a data stream synchronization mechanism, eliminating duplicate records by using a data redundancy filtering method, generating a real-time resource state data set, and obtaining resource matching input data; acquiring original data of different monitoring points from multiple data sources, collecting time sequence input streams including parking lot resource state and regional traffic flow information; applying a preset filtering rule to process the time sequence input stream, filtering out redundant information and format error data in the time sequence input stream to obtain a purified data set; according to the time stamp of each data item in the purified data set, using a time window alignment technology, calibrating data from different monitoring points to a unified time reference to form a multi-source synchronous data frame; by analyzing the multi-source synchronous data frame, aggregating the associated parking lot resource state and traffic flow data in the multi-source synchronous data frame, and constructing a structured real-time resource snapshot; using an exponential smoothing method to analyze a continuous real-time resource snapshot sequence, calculating the change rate of key indicators, predicting the short-term dynamic trend of the resource state, and obtaining a resource trend data set; combining the real-time resource snapshot and the resource trend data set, if the predicted change rate of the resource exceeds the preset fluctuation threshold, adjusting the weight of the resource in the matching domain to determine the final resource matching input data.

[0009] Preferably, the initial scheduling generation module generates an initial scheduling process through resource state analysis and fusion accuracy verification according to the resource matching input data, and if the fusion accuracy verification deviation value is lower than the preset scheduling response speed threshold, determines the initial scheduling process configuration scheme, including: obtaining multi-source data streams from resource matching input, pre-processing the data using a data stream integration method, and generating a unified format resource state data set; according to the resource state data set, using a resource state analysis technology, extracting key state features to obtain a resource state feature set; through fusion accuracy verification, performing deviation analysis on the resource state feature set, if the deviation is lower than the preset scheduling response threshold, determining the initial scheduling process configuration scheme; according to the initial scheduling process configuration scheme, using a dynamic combination mechanism to generate a resource allocation priority list; through the resource allocation priority list, using Dijkstra algorithm to calculate the optimal resource scheduling path to obtain a path planning result set; if there is a scheduling conflict in the path planning result set, sorting the priority through a conflict resolution rule to generate an optimized scheduling instruction set; according to the optimized scheduling instruction set, using a real-time data update mechanism to update the resource state data set synchronously to generate a final scheduling execution scheme.

[0010] Preferably, the resource state updating module, if the fusion accuracy check deviation value is higher than the preset scheduling response speed threshold, updates the real-time resource state data set through dynamic weight distribution and abnormal data elimination to obtain the adjusted resource matching input data, comprising: if the fusion accuracy check deviation value is higher than the preset threshold, adjusting the resource matching input data through dynamic weight distribution, eliminating abnormal data, and generating an updated resource state data set; according to the updated resource state data set, using a data flow integration method to preprocess multi-source data to generate a standardized resource state set; using a K-means algorithm to cluster analyze the resource state through the standardized resource state set to obtain a resource state classification set; if there is an abnormal classification in the resource state classification set, eliminating the abnormal classification through a preset abnormal detection rule to generate an optimized resource state classification set; using a linear regression algorithm to predict resource allocation priority according to the optimized resource state classification set to obtain a resource allocation priority list; using a Dijkstra algorithm to calculate the optimal scheduling path through the resource allocation priority list to generate a scheduling path planning result; using a real-time data updating mechanism according to the scheduling path planning result to synchronize the resource state data set to generate a final scheduling execution scheme.

[0011] Preferably, the path optimization module, according to the adjusted resource matching input data, generates an optimal path planning scheme through path smoothing processing and real-time response delay optimization to determine the parking resource allocation result, comprising: obtaining the adjusted input data set from the resource matching data, and generating an initial path planning set through path smoothing processing; according to the initial path planning set, using a real-time response delay optimization algorithm to adjust the path to obtain an optimized path planning set; if there is a path in the optimized path planning set whose delay exceeds the preset threshold, eliminating the unqualified path through a preset path screening rule to generate a path set; according to the path set, using a Dijkstra algorithm to calculate the resource allocation priority of each path to obtain a resource allocation priority list; through the resource allocation priority list, combining the parking resource availability data to generate a preliminary parking resource allocation scheme; according to the preliminary parking resource allocation scheme, using a real-time data synchronization mechanism to update the resource state data set to obtain a final parking resource allocation result; through the final parking resource allocation result, generating an optimal path planning scheme.

[0012] Preferably, the conflict detection module, for the parking resource allocation result, if a path blocking or resource conflict event is detected, generates an alternative scheduling scheme through an abnormal trigger threshold judgment, including: obtaining dynamic data flow of parking resource allocation through real-time monitoring, detecting path blocking events and resource conflict events using time series analysis method to obtain event trigger records; according to the event trigger records, using a pre-set abnormal trigger threshold for screening, if the trigger value exceeds the threshold, an initial alternative scheduling scheme set is generated; from the initial alternative scheduling scheme set, resource state update data is obtained, and the real-time response capability is combined to evaluate the feasibility of each scheme to obtain a scheduling scheme set; through the scheduling scheme set, the A algorithm is used to calculate the path optimization priority of each scheme to obtain a priority sorting list; according to the priority sorting list, the parking resource availability data is obtained to generate a dynamically adjusted resource allocation scheme; through the dynamically adjusted resource allocation scheme, the data synchronization mechanism is used to update the resource state data set of the real-time monitoring system to obtain the final scheduling scheme; according to the final scheduling scheme, a solution to the path blocking and resource conflict is generated to determine the optimized parking resource allocation result.

[0013] Preferably, the scheduling adjustment module, according to the alternative scheduling scheme, updates the scheduling process through event priority evaluation and resource reallocation, and determines the final scheduling execution scheme combined with dynamic response adjustment, including: obtaining event priority data from the alternative scheduling scheme set, arranging the event priority using a sorting algorithm to obtain an initial event sorting list; according to the initial event sorting list, obtaining real-time resource state data flow, if the resource state data flow shows that there is a resource allocation conflict, generating a conflict event record through a conflict detection mechanism; from the conflict event record, using a pre-set response time threshold for screening, if the response time of the conflict event exceeds the threshold, a resource reallocation scheme set is generated; according to the resource reallocation scheme set, obtaining path occupation information in the real-time data flow, using a path optimization algorithm to calculate the path smoothness of each scheme to obtain an optimized path scheme set; from the optimized path scheme set, the resource state data set of the monitoring system is updated through the data synchronization mechanism, if the updated resource state meets the feasibility evaluation conditions, a feasible scheduling scheme set is determined; according to the feasible scheduling scheme set, real-time event trigger data is obtained, and the resource utilization rate evaluation method is used to calculate the resource occupation efficiency of each scheme to obtain the final scheduling execution scheme; from the final scheduling execution scheme, the resource allocation state of the real-time monitoring system is updated through the data synchronization mechanism to determine the optimized scheduling execution process.

[0014] Preferably, the process tracking module obtains the execution state from the final scheduling execution scheme, tracks the user parking process, and generates a scheduling completion confirmation if the real-time response delay meets the preset scheduling response speed threshold, including: obtaining execution state data from the scheduling execution scheme, obtaining real-time updates of the execution state by analyzing task allocation and execution timestamps in the scheduling scheme; according to the real-time updates of the execution state, using data flow synchronization technology, matching state data with sensor data of the user parking process to determine the current stage of the parking process; obtaining key event timestamps and state changes in the user parking process through event log recording, and generating a structured event sequence; for the structured event sequence, if the real-time response delay is lower than the preset scheduling response speed threshold, calculate the average delay by the delay evaluation method to determine whether the scheduling task meets the response requirement; according to the calculation result of the delay evaluation method, using the state tracking mechanism, continuously monitoring the state changes in the user parking process to obtain dynamic updates of the parking process; through the confirmation generation logic, extracting state data meeting the preset threshold from the dynamic updates to generate a scheduling completion confirmation; if the scheduling completion confirmation is generated, write the confirmation information into the event log recording system through data flow synchronization to update the final scheduling state.

[0015] From the above technical solution, the present application has the following beneficial effects:

[0016] The real-time reservation and scheduling system for shared parking generates a standardized description by extracting user demand, decomposes the scheduling process into demand receiving, resource matching and path planning units, and generates an initial scheduling process in combination with real-time monitoring data and dynamic combination mechanism. When the fusion accuracy deviation exceeds the standard, the resource state data set is optimized by excluding abnormal data and adjusting the weight, the optimal path scheme is generated by using the path planning optimization algorithm, and the alternative scheduling scheme is generated by path blocking detection and resource conflict identification. Finally, the execution scheme is determined and continuously tracked through the real-time adjustment mechanism. The present application ensures efficient resource allocation and path planning through dynamic weight allocation and real-time response optimization, significantly improves the parking scheduling efficiency and user experience. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The system connection diagram of the present application. DETAILED DESCRIPTION

[0018] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all the other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0019] As Figure 1 shown, the present application provides a technical solution: a real-time reservation and scheduling system for shared parking, comprising: a demand analysis module, extracting parking scene type, time requirement and priority information from user parking request data, generating a standardized demand description through dynamic weight distribution, and obtaining user demand analysis results; a process decomposition module, decomposing the parking scheduling process into demand receiving, resource matching and path planning business units according to the parking scene type and time requirement in the user demand analysis results, and determining a business unit set; a resource acquisition module, acquiring parking lot resource state and traffic flow data for the resource matching unit, and obtaining resource matching input data; an initial scheduling generation module, generating an initial scheduling process through resource state analysis and fusion accuracy check according to the resource matching input data, and determining an initial scheduling process configuration scheme if the fusion accuracy check deviation value is lower than a preset scheduling response speed threshold; a resource state update module, updating the real-time resource state data set through dynamic weight distribution and abnormal data elimination if the fusion accuracy check deviation value is higher than the preset scheduling response speed threshold, and obtaining adjusted resource matching input data; a path optimization module, generating an optimal path planning scheme through path smoothing processing and real-time response delay optimization according to the adjusted resource matching input data, and determining a parking resource allocation result; a conflict detection module, generating an alternative scheduling scheme through abnormal trigger threshold judgment if a path blocking or resource conflict event is detected for the parking resource allocation result; a scheduling adjustment module, updating the scheduling process through event priority evaluation and resource reallocation according to the alternative scheduling scheme, and determining a final scheduling execution scheme in combination with dynamic response adjustment; a process tracking module, obtaining an execution state from the final scheduling execution scheme, tracking the user parking process, and generating a scheduling completion confirmation if the real-time response delay meets the preset scheduling response speed threshold.

[0020] The system processes user parking request data in stages based on a modular architecture. First, the demand analysis module extracts the key factors of the user's parking scenario and standardizes the processing with a dynamic weight mechanism, improving the uniformity and processing efficiency of the demand data. Subsequently, the process decomposition module decomposes the scheduling task according to the scenario type and time dimension, ensuring the refinement of the scheduling strategy. The resource acquisition module introduces real-time parking resource status and traffic flow information, providing a data basis for subsequent scheduling strategy formulation. The initial scheduling generation module generates a preliminary scheduling process using resource status information and checks its effectiveness by fusing precision. If the checking result meets the expectation, i.e., the deviation value is less than the threshold, the scheduling configuration scheme is determined; otherwise, the resource status update module is triggered for dynamic adjustment, ensuring the timeliness and accuracy of the scheduling data. The path optimization module then optimizes the path smoothing and delay minimization based on this, obtaining the optimal parking path planning result. If the generated resource allocation result has path blocking or resource conflict, the conflict detection module will determine and generate alternative schemes according to the abnormal event threshold. The scheduling adjustment module reconfigures the resources based on the event priority to optimize the final scheduling execution path. The entire scheduling process is monitored by the process tracking module in a closed loop to ensure that the execution efficiency and response time are controlled within the expected range, and finally a scheduling completion confirmation is generated.

[0021] This embodiment improves the adaptability of the shared parking system to complex and variable scenarios. Through the modular architecture and dynamic feedback mechanism, it can efficiently respond to user demand while improving the utilization of parking resources and the rationality of path scheduling, significantly enhancing the stability and reliability of the system in real-time scheduling.

[0022] The demand analysis module extracts the parking scenario type, time requirement, and priority information from the user parking request data, generates a standardized demand description through dynamic weight allocation, and obtains the user demand analysis result, including:

[0023] From the user request data, parse the parking scenario type, time requirement, and priority information, use structured data extraction methods to obtain the initial demand data set;

[0024] If the initial demand data set includes missing values, fill them in according to the default value filling rules to obtain the complete demand data set;

[0025] According to the parking scenario type and time constraint conditions in the complete demand data set, use the decision tree algorithm to generate scenario classification labels, and obtain the classified demand description;

[0026] Through the classified demand description, combine the priority information and weighted allocation rules to calculate the demand weight value, and obtain the weighted demand vector;

[0027] If the weight value of the weighted demand vector exceeds a preset threshold, the parking resource is preferentially allocated, and a resource allocation scheme is generated;

[0028] According to the resource allocation scheme and the time constraint condition, a scheduling algorithm is used to optimize the parking resource allocation sequence, and a final parking allocation result is obtained;

[0029] Through the final parking allocation result and the user identification information, a personalized parking demand description is generated, and user notification data is obtained.

[0030] In this embodiment, the demand analysis module receives parking request data submitted by users through the client, which includes parking location preferences, estimated arrival time, parking duration, user identification, and historical preference data. The system first processes the request data using a structured data extraction method. This method converts raw text or semi-structured input into a uniform format data table based on fixed field parsing rules. When the system receives user-submitted parking request data, it first performs field-level processing on the data. User-submitted data may come from mobile applications, web interfaces, or voice inputs, with varying degrees of structure. To ensure data uniformity and readability in the system, the system predefines a set of field parsing templates, which specify key fields to be identified, including "parking location preference," "estimated arrival time," "expected parking duration," "user type," "historical overdue record," and "registration time." Each field has specific identification rules and data format requirements. For example, for the "parking location preference" field, the system requires extracting phrases with spatial positioning characteristics such as "XX parking lot," "certain square," or "certain road segment" from the text. For the "estimated arrival time" field, the system requires identifying standard time formats such as "2025-07-28 15:30" or "15:30." For the "expected parking duration" field, the system identifies "hour" and "minute" units and converts them to a unified time value in minutes. During the parsing process, the system performs natural language segmentation on the text information, locates the field content through keyword dictionaries and regular matching methods, and fills it into the standardized data structure to form a structured record. If a field is missing or fails to be identified, the system marks it as null and waits for subsequent steps to complete. After processing, each parking request is represented as a unified data table item with a fixed column field structure, making it easy for the system to call and process in subsequent demand analysis, weight calculation, and resource allocation processes, ensuring data consistency, readability, and computability. Based on fixed field parsing rules, raw text or semi-structured input is converted into a uniform format data table, including parking scenario type field, time requirement field, and priority information field. The parking scenario type field is extracted by comparing the system's pre-set scenario template, such as identifying "residential area night parking" as "long-term low-frequency" type and "commercial district peak parking" as "short-term high-frequency" type, with field values being a fixed set of predefined strings. The time requirement field is calculated by comparing the user's "estimated arrival time" and "parking duration," and if the interval between the estimated arrival time and the current system time is less than or equal to 10 minutes, it is defined as "immediate"; if the interval is greater than 10 minutes but not more than 30 minutes, it is defined as "short-term"; and if it exceeds 30 minutes, it is defined as "reservation."The priority information field is quantitatively assigned according to user account level, whether it is a historical high-frequency user, whether there is a historical overdue record, and other indicators. High-frequency users are assigned a value of 3, ordinary users are assigned a value of 2, and users with overdue records are assigned a value of 1. The system generates an initial demand data set based on this.

[0031] If any field in the initial demand data set is empty or illegal, the system will automatically complete it according to the preset default value filling rule. After the system completes the structured analysis of the user's original parking request data and generates the initial demand data set, it will immediately perform field integrity checking on the data set. The system calls the field checking module to check the key fields one by one, including whether the "parking scene type", "time requirement" and "priority information" fields are missing or invalid. If any field is empty or the format does not match the preset rule, the default value filling process is started. Specifically, if the "parking scene type" field is missing, the system defaults to "public parking lot", which is based on the most common scene type selected by statistics and has general adaptability; if the "time requirement" field is missing, the system automatically infers the time difference between the current system time and the request submission time, and if it cannot be calculated, it fills in "reservation"; if the "priority information" field is missing, the system sets a default priority score based on the user's registration time and historical behavior records, and if there is no historical information available, it is uniformly assigned a value of 1 to represent the ordinary user level. When the "priority information" field is missing, the user's registration time and historical behavior records are automatically scored to set a default priority score, with a full score of 3 points, and a higher score indicates that the user's request should be prioritized in scheduling. The system first extracts the user's registration time corresponding to the current request and calculates the difference from the current system time to obtain the user's registration length (in days). If the registration time is greater than or equal to 365 days, the system assigns a score of 1, indicating that the user has a stable usage background, otherwise it does not score. Second, the system counts the number of valid parking requests made by the user in the last 30 consecutive days. If the number of requests is greater than or equal to 20, the system considers the user to be a high-frequency user and assigns a score of 1, otherwise it does not score. Next, the system queries whether the user has a history of violations, including having made a reservation but not using it, maliciously canceling it, or occupying it for too long, etc. If there is no record of any violation, the system assigns a score of 1, otherwise the item does not score. The above three scoring items are the registration length score, usage frequency score and behavior compliance score, which are directly added together to form the priority score total value, with a maximum of 3 points and a minimum of 0 points. The scoring value is filled into the missing field as the default priority information for subsequent weighted demand vector calculation and scheduling sorting process, ensuring the determinism, reproducibility and system configuration controllability of the scoring logic. The filling process is performed in the data table structure in a direct overwrite manner, ensuring that each structured request record has complete, valid and identifiable data content before entering the subsequent weight calculation and scheduling analysis link, thereby forming a complete demand data set and ensuring the normal execution and consistency of the system's subsequent processing logic.

[0032] Next, according to the two fields of "parking scene type" and "time constraint condition" in the complete demand data set, the preset decision tree classification logic is used to generate scene classification labels to accurately describe the parking dispatch situation of the current user request. The decision tree model is composed of multiple layers of judgment nodes, and each layer of nodes corresponds to a judgment condition, and the output is a fixed scene label. The first layer of nodes is judged as the parking scene type, and the system sets five basic types, namely residential parking, commercial parking, office building parking, hospital area parking and school surrounding parking; according to the comparison of the "parking location preference" field in the user request with the system geographic information database, the corresponding scene type can be determined. The system enters the second layer of nodes, that is, the time constraint condition, and according to the comparison of the "expected arrival time" field with the current system time, if the time interval is less than or equal to ten minutes, it is defined as an instant scene; if the interval is more than ten minutes and less than or equal to thirty minutes, it is defined as a short-time scene; and if it exceeds thirty minutes, it is a reservation scene. The system combines the two layers of judgment paths to generate corresponding nine scene classification labels, for example, the residential area plus instant scene is classified as "residential area high emergency scene", the commercial area plus short-time scene is classified as "commercial area medium emergency scene", the hospital area plus reservation scene is classified as "hospital area low emergency scene", etc. Each combination corresponds to a fixed label name. The label is written into the processing record of the user request in the form of a string as the classified demand description, which is called by the subsequent weight calculation and dispatch sorting steps to realize accurate identification and hierarchical processing of the user demand situation, and to ensure that the classification logic is clear, the calculation process is transparent, and the output result is unique.

[0033] Subsequently, the system introduces priority information to the classified demand description to calculate the weight value. After completing the generation of the classified demand description, the system immediately enters the demand weight value calculation link to evaluate the processing priority of the current request. This calculation process is based on two explicit scoring factors: one is the scene emergency degree score determined by the scene classification label, and the other is the priority score derived from the user registration time and historical behavior records. First, the system reads the score value corresponding to the current classification label from the preset scene emergency degree score table. The score is an integer ranging from 1 to 5, for example, the score of "high emergency scene in residential area" is 5, the score of "medium emergency scene in commercial area" is 3, and the score of "low emergency scene around school" is 1. Then, the system obtains the previously generated priority score, which ranges from 0 to 3, for example, the total score of an old user, a high-frequency user, and no default record is 3. Next, the system applies a fixed weighting allocation rule, i.e., the scene emergency degree score is multiplied by 3, the priority score is multiplied by 2, and then the two calculation results are added to obtain the final demand weight value. For example, if the scene emergency degree score of a request is 4 and the priority score is 2, the system calculates that the scene weight part is 4 multiplied by 3 equal to 12, and the priority weight part is 2 multiplied by 2 equal to 4, and the sum of the two is the weighted demand vector 16. The system writes this value into the request record as an important basis for the sorting of the dispatch module. The scoring source, weight coefficient, and operation steps in the entire calculation process are fixed values set by the system, ensuring the standardization of the scoring process, the consistency of the results, and the repeatability of the dispatch priority determination.

[0034] The system sets a preset threshold of 18, which is based on the 75th percentile of the distribution of user average weight values in historical dispatch data. If the weight value of the weighted demand vector is greater than 18, the system automatically marks the user request as "high priority" and enters the priority dispatch channel to generate a resource allocation scheme. The resource allocation scheme is obtained by the system calling the current parking resource state module to obtain a list of available parking spaces, and sorting and matching from high to low according to the weighted demand value. If there are multiple parking resources that meet the conditions, the system selects them in order according to the geographical location distance and accessible time.

[0035] After generating the preliminary resource allocation scheme, the system combines the "estimated arrival time" field in the request with the current traffic flow data of the parking lot to call the dispatch algorithm module to optimize the final allocation order. The optimization process uses a fixed dispatch priority rule, for example, priority is given to requests with an estimated arrival time within 10 minutes of the current system time, followed by requests within 500 meters of the parking lot. After sorting, the final parking allocation result is obtained.

[0036] Finally, the system combines the final allocation result with the user identification information to generate a personalized parking demand description, including the allocated parking space number, parking lot name, estimated arrival time window, entrance navigation information, etc., formats it into structured notification data, and pushes it to the user terminal client for prompt, realizing visual reception by the user.

[0037] The flow decomposition module decomposes the parking dispatch flow into demand reception, resource matching, and path planning business units according to the parking scene type and time requirement in the user demand analysis result, and determines the business unit set to include:

[0038] The parking scene type and time requirement are obtained from the user demand analysis result, structured analysis method is used to extract demand feature data, and an initial demand set is generated;

[0039] If the parking scene type or time requirement in the initial demand set is missing, it is filled by default rules to obtain a complete demand set;

[0040] According to the parking scene type in the complete demand set, the K-means clustering algorithm is used to divide the scene categories and generate classified scene labels;

[0041] Through the classified scene labels and time requirements, combined with traffic pattern recognition, the random forest algorithm is used to predict the traffic flow distribution to obtain a traffic pattern set;

[0042] If the traffic distribution in the traffic pattern set exceeds the preset threshold, the resource matching order is adjusted in priority to generate an optimized resource allocation scheme;

[0043] According to the optimized resource allocation scheme and path planning requirements, the Dijkstra algorithm is used to calculate the optimal parking path to obtain the path planning result;

[0044] Through the path planning result and the complete demand set, a dispatch instruction set is generated to determine the final parking dispatch scheme.

[0045] In the present embodiment, after completing the demand receiving, the flow decomposition module first obtains two core fields from the user demand analysis results, which are parking scene type and time requirement respectively. The system processes this information through a structured analysis method. This method extracts keywords from the "location description" field in the user input and performs geographic matching. Descriptions containing fixed words such as "hospital", "commercial district", "school", "office building", and "residential area" are mapped to standard scene types according to matching rules, which are hospital area, commercial area, school area, office area, and residential area respectively. At the same time, the time requirement field is calculated by the system by subtracting the "expected arrival time" field from the current system time. If the time difference is less than or equal to 10 minutes, it is defined as immediate; if the time difference is greater than 10 minutes and less than or equal to 30 minutes, it is defined as short; if the time difference is greater than 30 minutes, it is defined as reservation. If any field is missing during the analysis process, the system immediately applies the default filling rule. The parking scene type is filled by default as "public area parking", and the time requirement is filled by default as "reservation". The filling value comes from the most widely distributed default type in the past dispatch, ensuring that it does not affect the subsequent process.

[0046] After data integrity verification is completed, the system enters the clustering processing step, and uses the K-means algorithm to divide all parking requests into scene categories. The system sets the number of cluster centers to 5, which corresponds to residential area, commercial area, office area, hospital area, and school area. The number of cluster centers is determined based on the statistical classification of the last 180 days of parking request data, and is fixed as a constant value during model training. The scene type field of each parking request is converted into a feature vector through natural language vector conversion. The system calculates the Euclidean distance between this feature vector and the vectors of the 5 cluster centers, and the minimum distance is the category to which it belongs, thereby generating the scene classification label of each request. The system takes this classification label and the original request time requirement field as input and enters the traffic flow prediction stage.

[0047] In the traffic prediction phase, the system calls the historical traffic data of the area to which the scene belongs in the last 60 minutes, including the average number of vehicles passing through the road section, the average speed, the signal light density, the road grade and other elements, as the training sample input into the trained random forest model. The model contains 200 decision trees, each tree has a maximum depth of 10, and the training data is derived from the measured data of the urban traffic management system in the past 365 days. The system classifies the scene of the current request, the time requirement field, the current date holiday identification and the weather conditions at the time point as features for prediction, and the output result is the vehicle passing density of the corresponding area in the next 30 minutes, with the unit of vehicle per hour. The system adds this value to the traffic pattern set as the traffic prediction value. The system presets the traffic density threshold value as 500 vehicles per hour, which is based on the 80th percentile of the congestion distribution statistics of all areas in the past year to identify the critical point of entering the congestion state. If the traffic prediction value corresponding to a certain request is greater than or equal to 500, the system will lower the parking space dispatch priority in the associated resource pool and arrange it after the low congestion area resources to generate an optimized resource allocation scheme.

[0048] After completing the resource priority adjustment, the system performs path planning calculation for each optimized dispatch request, and uses Dijkstra algorithm for shortest path query. The system takes the latitude and longitude of the user's current location and the target parking space as the starting point and ending point coordinates, calls the adjacency matrix constructed by the city road graph database, and the weight of each edge in the matrix is composed of the following three parts: the actual length of the road (in meters), the current road traffic density value and the number of signal lights. The system takes the road length as the basic weight, and if the road traffic density is greater than 500 vehicles per hour, the weight of the section will be increased by 50 meters equivalent value; for each additional signal light, add 50 meters equivalent weight. After all possible paths are calculated, the system selects the one with the smallest cumulative weight value as the optimal path to generate the path planning result. The path planning result includes the total path distance, the estimated travel time, the path node number sequence, the turning type of each node and other information.

[0049] The system merges the path planning result with the complete demand set to generate a dispatch instruction set, each instruction including dispatch number, user identification, resource number, path node list, estimated arrival time, path total time and other fields. The set is written into the system dispatch queue as the final parking dispatch scheme for subsequent module calling and execution. The parameters involved in the above process include the number of clustering centers, the clustering distance judgment method, the traffic density threshold, the path weight composition item, the signal light influence value, etc., which are fixedly set or determined based on city historical data statistics to ensure consistent calculation method, traceable process and stable and reliable results.

[0050] The resource acquisition module acquires the parking lot resource state and traffic flow data for the resource matching unit to obtain resource matching input data including:

[0051] The real-time monitoring acquires the parking lot resource state and traffic flow data, integrates multi-source data through a data flow synchronization mechanism, eliminates duplicate records using a data redundancy filtering method, generates a real-time resource state data set, and obtains resource matching input data;

[0052] The original data of different monitoring points is acquired from multiple data sources, and time series input streams including parking lot resource state and regional traffic flow information are collected;

[0053] A preset filtering rule is applied to process the time series input stream, and redundant information and format error data in the time series input stream are screened out to obtain a purified data set;

[0054] According to the time stamp of each data entry in the purified data set, a time window alignment technology is used to calibrate the data from different monitoring points to a unified time reference to form a multi-source synchronous data frame;

[0055] By analyzing the multi-source synchronous data frame, the associated parking lot resource state and traffic flow data in the multi-source synchronous data frame are aggregated to construct a structured real-time resource snapshot;

[0056] An exponential smoothing method is used to analyze a sequence of continuous real-time resource snapshots, calculate the change rate of key indicators, predict the short-term dynamic trend of the resource state, and obtain a resource trend data set;

[0057] In combination with the real-time resource snapshot and the resource trend data set, if the predicted change rate of the resource exceeds a preset fluctuation threshold, the weight of the resource in the matching domain is adjusted to determine the final resource matching input data.

[0058] In the embodiment, the resource acquisition module is used to construct the resource matching input data of the scheduling system, and the implementation process includes six continuous steps of data collection, cleaning, synchronization, aggregation, analysis and weight adjustment, each of which has specific parameter setting and calculation logic. First, through the deployment of sensing devices in various parking lots and urban road nodes, real-time collection of parking resource status and traffic flow data is performed to form multi-source raw data streams. Each monitoring point periodically reports data records containing resource number, parking space occupancy status, reporting timestamp, road segment number, current vehicle passing number and average passing speed, etc. After the system receives the data, it first performs unified structure analysis on all original time series data from different sources, converts them into system standard field format, and enters the stream processing channel. On this basis, the system applies a data stream synchronization mechanism to preliminarily aggregate all data streams in chronological order. Since different data sources may repeatedly report records with consistent status in a very short time, the system uses a data redundancy filtering method to remove duplicate records. In specific operation, the system generates a unique identifier for each record, and compares the timestamp of the current record with the last valid record of the resource in the cache. If the time difference is less than or equal to 30 seconds and all status fields are identical, it is determined that the record is redundant and does not enter the subsequent process. If the current record and the data in the cache have any state field difference, or the time interval exceeds 30 seconds, the data is retained and written into the cache to update the latest status. After redundancy filtering, the system aggregates all valid data in a fixed time window of 30 seconds, classifies and aggregates them by resource number, and generates a structured real-time resource status data set. Each record in the data set represents a resource item's status snapshot in the current time window, including its availability, location, traffic parameters and reporting time, etc. Finally, the system provides this structured real-time resource status data set as resource matching input data to the subsequent scheduling calculation module, ensuring the timeliness, uniqueness and accuracy of the input data, and providing high-quality basic data support for path planning and resource allocation. By establishing real-time connections with multiple data sources, including fixed parking lot detection devices, geomagnetic sensors, high-definition video recognition terminals and traffic flow radar devices, the system receives original time series data streams from different monitoring points. Each monitoring point sends complete data records containing resource status and traffic flow information at fixed intervals, each record containing at least resource number, parking space status value, reporting timestamp, parking lot number, road segment number, passing vehicle number, average passing speed, etc. The system unifies these data into a data stream processing channel and configures a unique identifier for each data source to distinguish the source. After access is completed, the system starts the preset filtering rules to process all time series input streams in real time, and removes redundant records and format error information.The specific processing steps are as follows: first, the system establishes a cache index according to the resource number and the timestamp, compares each new access record with the last record of the same resource number in the cache, and if the state values are completely consistent and the timestamp interval is less than or equal to 30 seconds, it is marked as redundant information and not written into the main memory; if there is any difference in the state field or the time interval is greater than 30 seconds, the current record is retained and the cache is updated. Secondly, the system performs format checking on each record, including whether the timestamp is a valid standard time format, whether the state field is a specified value, whether the field is missing or exceeds the valid range, and if the missing field or abnormal value, such as null, negative, and out-of-limit value, is detected, the record is immediately rejected. After the above two processes are completed, the system collects all records that pass the redundancy check and format check into a set of purified data, and all data in this set have uniqueness, integrity and format specification, providing reliable input for subsequent data synchronization, aggregation and trend analysis. The entire process is executed once every 30 seconds, ensuring that the system only processes valid data every cycle, improving computing efficiency and result accuracy.

[0059] The purified data enters the time alignment stage, and the system sets a uniform time alignment window of 30 seconds, and takes all data records' timestamps down to the nearest 30-second value before the current time point, ensuring that all records are synchronized on a unified reference time. For example, if a record timestamp is 13:05:47, its time alignment value is 13:05:30. The system aggregates all data of the same alignment time to form a frame, called a multi-source synchronized data frame, which contains all the effective data of the monitoring points in a complete time window. Then, the system classifies and merges the information in the data frame according to the resource number and the road section number in the data record, and constructs a structured real-time resource snapshot. Each snapshot record includes resource number, number of empty parking spaces, average vehicle speed, current traffic flow, corresponding timestamp, etc. The system takes this snapshot as a complete description of the resource state of the area at the current time point.

[0060] The system stores each resource snapshot continuously for 10 minutes, generates a snapshot record every 30 seconds, and a total of 20 consecutive snapshot data. On this basis, the system applies exponential smoothing method to analyze the change trend of resource state. The exponential smoothing coefficient is set to 0.6, which is an empirical value calculated by minimizing the prediction error of historical data. In the specific calculation, the system first multiplies the value of the latest snapshot by 0.6 as the current time weight, then multiplies the snapshot value of the previous time by 0.6 by 0.4, which is equal to 0.24, and then multiplies the snapshot value of the previous time by 0.6 by 0.4 by 0.4, which is equal to 0.096, and so on. Finally, all the weighted results are added to obtain the smoothed state estimate value. The difference between the two consecutive smoothed values is divided by the previous smoothed value to obtain the resource state change rate. The system processes two core indicators in this way, i.e. the change rate of the number of idle parking spaces and the change rate of the average vehicle speed.

[0061] The system compares the change rate with the preset fluctuation threshold to determine the state of the resource. The fluctuation threshold of the number of idle parking spaces is set to 10%, and the fluctuation threshold of the average vehicle speed is set to 15%. These thresholds are determined based on the statistical analysis of the monitoring data of the past year, and the 75th percentile value of the fluctuation amplitude distribution of each indicator is extracted, which can effectively reflect the critical change state. If the change rate of the number of idle parking spaces of a resource is greater than 10% or the decrease amplitude of the average vehicle speed is greater than 15%, the system determines that the resource state is in a high fluctuation interval. For the resource in a high fluctuation state, the system reduces the initial scheduling weight of the resource in the resource matching domain by 20% from the default value of 100, i.e. modifies it to 80, to reduce the probability of being selected in the subsequent scheduling algorithm, so as to ensure the stability and prediction reliability of resource allocation. The system finally integrates all records containing resource snapshot information, trend prediction data and adjusted weight values into the resource matching input data structure, which is used as the direct input basis for the subsequent allocation strategy of the scheduling engine. Each parameter in the above process is clearly defined, the source is reliable, all calculation processes are linear or weighted operations, and the logic sequence and repeatability are clear, which ensures the accuracy, stability and prediction accuracy of the scheduling system.

[0062] The initial scheduling generation module generates an initial scheduling process based on the resource matching input data, through resource state analysis and fusion accuracy verification. If the fusion accuracy verification deviation is lower than the preset scheduling response speed threshold, the initial scheduling process configuration scheme includes:

[0063] The initial scheduling generation module generates an initial scheduling process based on the resource matching input data, through resource state analysis and fusion accuracy verification. If the fusion accuracy verification deviation is lower than the preset scheduling response speed threshold, the initial scheduling process configuration scheme includes:

[0064] According to the resource state data set, the resource state analysis technology is used to extract the key state characteristics, and the resource state feature set is obtained;

[0065] By fusing the precision check, the resource state feature set is subjected to deviation analysis, and if the deviation is lower than a preset scheduling response threshold, an initial scheduling flow configuration scheme is determined;

[0066] According to the initial scheduling flow configuration scheme, a dynamic combination mechanism is adopted to generate a resource allocation priority list;

[0067] Through the resource allocation priority list, a Dijkstra algorithm is used to calculate an optimal resource scheduling path, and a path planning result set is obtained;

[0068] If there is a scheduling conflict in the path planning result set, the priority is sorted through a conflict resolution rule, and an optimized scheduling instruction set is generated;

[0069] According to the optimized scheduling instruction set, a real-time data updating mechanism is adopted to synchronously update the resource state data set, and a final scheduling execution scheme is generated.

[0070] In the embodiment, the initial scheduling generation module is used to generate an executable scheduling scheme under the premise that the resource state data is stable and the fusion accuracy meets the requirements. The implementation process includes eight main steps of data integration, state feature extraction, fusion accuracy verification, scheduling flow configuration, priority sorting, path planning, conflict processing and resource state updating. Each step has clear calculation basis and parameter control. First, the system receives multi-source data streams from resource matching input. These data streams come from different monitoring points and device platforms, including parking lot management system, geomagnetic sensor, traffic flow detector and urban road camera, etc. The original record format provided by each data source is different, and the field naming, unit standard and time format are different. After receiving the data, the system immediately enters the data stream integration stage, and specifically performs the preprocessing steps of unified field mapping, timestamp standardization, unit conversion and field completion. First, the system matches the field name in each data source with the corresponding relationship of the standard field through the field mapping table, for example, the "empty parking space number" field in a certain data source is mapped to the system standard field "free parking space number". Then, the system reads the timestamp field of each record, and uniformly converts the non-standard format to the six-segment time format of year, month, day, hour, minute and second, to ensure that the timestamp has a unified structure. For fields with inconsistent units, such as mixed use of "meters per second" and "kilometers per hour" in the vehicle speed field, the system uniformly converts all values to "kilometers per hour" as the standard unit. If some data sources do not provide complete fields due to interface limitations, the system will complete them according to the field missing filling rules, for example, when the traffic flow field is missing, it is filled according to the historical average value of the monitoring point; when the number of traffic lights field is missing, it is supplemented according to the road segment basic data. All the above operations are completed within the first 30 milliseconds after each data access, and the system ensures that the integration and format standardization of all stream data are completed within each 30-second time window period. After integration, all records are organized into standardized data rows, and the system stores them according to resource number, forming a unified format resource state data set, each record containing resource number, free parking space number, road segment number, traffic flow, average speed, standard timestamp and other fixed fields, for subsequent analysis, verification and scheduling flow calling, to ensure consistent data structure, complete fields and standardized format.

[0071] Subsequently, the system immediately calls the resource state analysis technique after obtaining the resource state data set in a unified format, and performs feature extraction operations on each data record, aiming to identify key state elements that affect resource scheduling decisions and build a resource state feature set. First, the system reads the "number of free parking spaces" field in each record and standardizes it to an integer type as the first type of state feature "resource availability"; this value is directly used to determine whether the current resource has allocation conditions. Second, the system extracts the "average traffic speed" field as the second type of state feature "traffic efficiency", which is in units of kilometers per hour, reflecting the vehicle traffic conditions of the corresponding road section in the time window. The higher the value, the more unobstructed the road section is. Third, the system calculates the resource accessibility score, which is composed of three factors: road grade, number of traffic lights, and traffic flow. The road grade is assigned according to the preset rules, with expressways assigned a score of 3, main roads assigned a score of 2, and branch roads assigned a score of 1. The number of traffic lights is provided by the static properties of the road section, with a deduction of 0.2 points for each additional traffic light. The traffic flow field is in units of vehicles per hour, and if the value exceeds 100, a deduction of 0.1 points is made for each 100 vehicles. The final score is rounded to two decimal places and defined as the third type of state feature "resource accessibility level". After the system completes the extraction of the above three types of features for each record, it combines the "resource availability", "traffic efficiency", and "resource accessibility level" into structured items and writes them into the resource state feature set. The entire process is carried out in memory, and the system can process more than 1000 data records per second, ensuring real-time and efficient feature extraction and providing a data foundation for subsequent fusion accuracy verification and scheduling decisions.

[0072] After completing feature extraction, the system performs fusion accuracy verification. This process is based on resource number and retrieves the corresponding state values of the same resource item in different data sources, and calculates the deviation of the number of free parking spaces and the average speed. The system takes the absolute value of the difference between the number of free parking spaces from the two data sources, and if the difference is less than or equal to 1, it is considered that the deviation meets the accuracy requirements. At the same time, the system calculates the difference between the average speeds of the two data sources, and if the difference is less than or equal to 5 kilometers per hour, it is considered that the speed deviation meets the accuracy requirements. The scheduling response speed threshold defined by the system is that the deviation of the number of free parking spaces does not exceed 1 unit, and the deviation of the average traffic speed does not exceed 5 kilometers per hour. These thresholds are set based on the 95th percentile of the error distribution extracted from the analysis of the fusion failure records in the past year, to ensure the consistency of data fusion. If both indicators meet the above conditions, it is determined that the fusion accuracy is qualified, and the scheduling process configuration is started.

[0073] In the configuration of the scheduling process, the system generates a resource allocation priority value according to the user request urgency, resource accessibility score and road traffic efficiency. The urgency score is obtained by subtracting the user's expected arrival time from the current system time. If the time difference is less than or equal to 10 minutes, the score is 5; if the time difference is greater than 10 minutes and less than or equal to 30 minutes, the score is 3; and if the time difference exceeds 30 minutes, the score is 1. The traffic efficiency score is divided according to the average speed, with a speed greater than or equal to 30 kilometers per hour scoring 5, a speed between 10 and 30 kilometers per hour scoring 3, and a speed less than 10 kilometers per hour scoring 1. The system weights the three scores by multiplying them by their respective weights and summing them up. The urgency weight is 0.5, the accessibility weight is 0.3, and the traffic efficiency weight is 0.2. The resulting score is the priority value corresponding to the resource. The system ranks all schedulable resources in descending order of priority value to generate a resource allocation priority list.

[0074] After generating the resource allocation priority list, path calculation is immediately performed. The Dijkstra algorithm is used to calculate the optimal scheduling path between each user request and the target resource. First, the system reads the resource number corresponding to each request from the resource allocation priority list and extracts the latitude and longitude coordinates of the user's current location and the target resource location. Then, the system constructs a directed graph structure on the city road topology map, with road segments as edges and intersections as nodes. Each edge in the graph is attached to a path weight value, which represents the traffic cost. The weight value is composed of three parts: the actual length of the road, the number of traffic lights, and the traffic flow factor. The length of the road is recorded in meters as the base weight; if there is a traffic light on the road segment, an additional 50 meters of equivalent weight is added for each traffic light; if the number of vehicles passing through the road segment per unit time exceeds 100, an additional 30 meters of equivalent weight is added for each 100 vehicles. After calculating the cumulative weight of each edge, the system uses the Dijkstra algorithm to search for the path from the user's current location node to the target resource location node, records all the path nodes and cumulative weight values, and selects the path with the smallest weight value as the optimal path. The system repeats this calculation process until all requests in the resource allocation priority list have been calculated. Each path result contains the path node sequence, the estimated travel distance, the estimated time, the road segment identifiers and the number of traffic lights on the path. All path planning results are organized into a path planning result set. This set is an important part of the subsequent scheduling execution instructions, ensuring that resource allocation has the smallest traffic cost and optimal response efficiency.

[0075] After the generation of the path planning result set, conflict detection is immediately performed on all path endpoints in the set to determine whether multiple user optimal path endpoints point to the same resource number. The system first constructs a reverse index table indexed by resource number, and counts how many path planning results each resource is referenced by. If the number of paths corresponding to a resource number is greater than 1, it is determined that there is a scheduling conflict for the resource. The system sorts the requests associated with each conflict resource in descending order according to their priority values in the resource allocation priority list. The priority value is calculated by weighting the user urgency, resource accessibility, and traffic efficiency, with specific weights of urgency 0.5, accessibility 0.3, and traffic efficiency 0.2. For the user request with the highest priority value in the conflict resource, the system retains its original path and resource allocation result. For other user requests, the system marks them as "resource conflict to be reselected" and reassigns the target resource according to the next available resource in the priority list that is not occupied. The path planning process is then executed again, and the above steps are repeated until all requests do not overlap in resource endpoints. During each conflict reselection process, the system records the final selected path node sequence, traffic distance, estimated time consumption, and reassigned resource number. Finally, the system integrates all non-conflicting path planning results and the paths selected after conflict processing to generate an optimized scheduling instruction set. Each instruction in the set explicitly includes the user number, allocated resource number, path node sequence, estimated arrival time, and path traffic parameters, ensuring that each resource is only occupied by one request during the scheduling period, achieving dynamic controllability of scheduling conflicts and final effectiveness of the scheduling scheme.

[0076] Finally, the system immediately executes the real-time data updating mechanism after obtaining the optimized scheduling instruction set to synchronously update the resource state data set and generate the final scheduling execution scheme. First, the system parses each instruction in the scheduling instruction set one by one, extracts the resource number, user number, expected arrival time, and specified parking duration, and other key fields. The system looks up the corresponding record in the current resource state data set according to the resource number, and performs an update operation on the "free parking space number" field, reducing the value by 1, and the update result must not be less than 0; if the updated value is 0, the resource state is marked as "unallocatable". Then, the system creates two new fields, "expected start time" and "expected release time". The "expected start time" is directly assigned the expected arrival time in the scheduling instruction, and the "expected release time" is calculated by adding the user-specified parking duration to the "expected start time", for example, if the parking duration is 90 minutes, the "expected release time" is 90 minutes later than the "expected start time". The system also adds the "occupied user number" field to the resource record and writes the corresponding user number in the current scheduling instruction. After completing the update of the above fields, the system writes the modified record to the current effective version of the resource state data set, ensuring that the resource is locked during the current scheduling period and does not participate in the matching process of other scheduling requests. The update process is executed synchronously after each instruction is processed, without buffer delay, ensuring that the resource state data set remains real-time and effective in the scheduling system. After completing the update of all resource states, the system packages the updated resource records with the original scheduling instruction information to form the final scheduling execution scheme, which serves as the basis for execution during the current scheduling period and is called by the system scheduling control logic for implementation, completing the full-process closed loop from scheduling data planning to execution.

[0077] The resource state updating module updates the real-time resource state data set by dynamic weight distribution and abnormal data elimination if the fusion accuracy check deviation value is higher than the preset scheduling response speed threshold, and obtains the adjusted resource matching input data, including:

[0078] If the deviation value of the fusion accuracy check is higher than the preset threshold, the resource matching input data is adjusted by dynamic weight distribution, abnormal data is eliminated, and the updated resource state data set is generated;

[0079] According to the updated resource state data set, a data stream integration method is used to preprocess the multi-source data to generate a standardized resource state set;

[0080] Through the standardized resource state set, a K-means algorithm is used to perform cluster analysis on the resource state to obtain a resource state classification set;

[0081] If there is an abnormal classification in the resource state classification set, the abnormal classification is eliminated by a preset abnormality detection rule to generate an optimized resource state classification set;

[0082] According to the optimized resource state classification set, a linear regression algorithm is used to predict the resource allocation priority, and a resource allocation priority list is obtained;

[0083] Through the resource allocation priority list, a Dijkstra algorithm is used to calculate the optimal scheduling path, and a scheduling path planning result is generated;

[0084] According to the scheduling path planning result, a real-time data updating mechanism is used to update the resource state data set synchronously, and a final scheduling execution scheme is generated.

[0085] In further implementation process, the system first calculates the deviation value of resource state characteristics in multiple data sources through fusion accuracy verification, which is the average of the absolute value of the numerical difference between all characteristic values, with the same dimension as the characteristic value, for example, the number of empty parking spaces is the number, the traffic flow is the number of vehicles per minute. When the deviation value exceeds the preset threshold set by the system, that is, the consistency of resource state between different data sources is insufficient, it cannot be directly used as scheduling basic data, the system immediately triggers the dynamic weight distribution process. Weight distribution is based on three parameters: data source type, historical accuracy rate and data reporting frequency. Data source type is divided into four categories, namely road geomagnetic detection equipment, parking lot infrared equipment, urban traffic camera system and user mobile phone application feedback, and the weights are set as 0.4, 0.3, 0.2 and 0.1 respectively. Historical accuracy rate refers to the proportion of the device that is consistent with the actual parking state in the past 24 hours, if the accuracy rate is more than 95%, an additional weight of 0.05 is added; if the accuracy rate is less than 85%, a weight of 0.05 is reduced. The data reporting frequency is calculated by the number of reports per minute, if it is more than 6 times, a weight of 0.05 is added, if it is less than 2 times, a weight of 0.05 is reduced. The system calculates the maximum weight value of each record according to the above weight addition and subtraction rules, and performs weighted average on the resource state value of each record to obtain the fusion data result.

[0086] Then the system performs outlier rejection on the fusion data result, and the rejection basis is whether the difference between the key field in each record and the weighted average value exceeds 5 times the historical standard deviation of the field. For example, if the difference between the number of empty parking spaces in a record and the weighted average value is 10, and the historical standard deviation is 2, the record is an abnormal record and should be deleted. This judgment standard is a fixed rule set by the system and does not change with the situation, ensuring the uniformity of the judgment.

[0087] After the completion of the exception elimination, the system obtains a preliminary updated resource state dataset, which still contains problems such as diverse sources, inconsistent formats, and incomplete fields, and therefore needs to be further preprocessed by a data stream integration method. The integration process includes three steps: the first step is field structure standardization, that is, uniformly naming the fields in all records and converting the data types to a uniform format, for example, converting the "free parking space number" field to an integer type and the "update time" field to a year-month-day-hour-minute-second format. The second step is time benchmark alignment, which aligns different source data according to the timestamp and uses a window alignment method to uniformly record the last valid data in the window. The third step is field completion and unit consistency verification, which uses the last valid data completion method to fill in the missing fields and converts different units to a uniform unit, such as converting kilometers per hour to meters per second. After the three steps, a standardized resource state set is generated, which has consistent data sources, uniform field structure, and synchronized time and meets the requirements of being directly used for subsequent resource matching and path calculation. The entire process is executed in batches inside the system, and the processing efficiency is improved through multi-thread parallel computing to ensure that the updated data is processed within 3 minutes.

[0088] The system receives the resource state set that has completed the standardization processing, and each data record contains consistent field information, mainly including resource number, free parking space number, current traffic flow, average travel speed, and timestamp. The system constructs a resource feature vector based on each resource number, and takes the free parking space number, traffic flow, and average speed as the clustering input dimensions. The system sets the initial clustering number of the K-means algorithm to 5, representing that all resources are divided into 5 categories, each representing a resource state mode, such as high availability and high smoothness, low availability and high congestion, etc. In the initial stage of the algorithm, 5 data points are randomly selected from the samples as the cluster centers, and then all resource data are classified according to the principle of minimum distance to the cluster center. The center value of each cluster category is calculated, and then the cluster center is iteratively updated until all data points are classified or the maximum iteration number reaches 100.

[0089] After clustering, the system counts the number of resources in each category, the average value and variance of each feature field, and generates a resource state classification set. Each record in the set records the typical resource state characteristics of a certain category. Next, the system performs anomaly detection on the clustering results to determine whether there is an abnormal category division. The specific rules are as follows: if a clustering category contains more than 60% of the total number of resources, and the average free parking space number or average travel speed of the category is less than 50% of the average value of other categories, then the category is determined to be abnormal, and the category will be excluded from the classification set. After the exclusion, the system reorganizes the resource state classification set to obtain an optimized classification set, ensuring that the clustering results are representative and free from the influence of excessively biased data.

[0090] Subsequently, the system utilizes the optimized resource classification results to construct a linear regression model for predicting the allocation priority of each resource. The model input parameters include the number of empty parking spaces, the average traffic speed, and the signal light density, and the weights of each parameter are determined by analyzing the past dispatch success rate, which are set to 0.5, 0.3, and 0.2, respectively. The system trains on historical dispatch data to establish a regression relationship between the input and the final dispatch effect, thereby obtaining the fitted weight coefficients. For each current resource data input into the model, the corresponding priority value is calculated, ranging from 0 to 100, with a higher priority value indicating that the resource should be prioritized in the current dispatch cycle. After sorting all resource priority values, a resource allocation priority list is generated for subsequent path calculation and final dispatch decision. The entire process is automatically completed by the system, with an execution time controlled within 1 minute, supporting high-frequency resource dynamic update scenarios.

[0091] The system determines the dispatch target resource based on the resource allocation priority list. The resources in the priority list are arranged in descending order of priority value, and the system selects the parking resource with the highest priority value and available resource state as the dispatch target. Subsequently, the system constructs a road network graph with the user's current location as the starting point and the target resource parking location as the ending point. The network graph consists of multiple nodes connected by edges, with each edge representing a road. The system sets a path weight for each edge, which is calculated from three indicators: the actual length of the road in meters, the number of signal lights on the section, and the traffic flow per unit time on the section in vehicles per minute. These three indicators are multiplied by weight coefficients, with the road length weight being 0.5, the signal light number weight being 0.3, and the traffic flow weight being 0.2. The calculated weighted total value is taken as the maximum weight of the edge.

[0092] The system uses Dijkstra's algorithm for path calculation. The specific calculation process is as follows: first, initialize the path cost of all nodes to the maximum value, and set the path cost of the starting node to 0; second, select the node with the smallest current path cost as the "current node" and traverse all adjacent nodes from the current node; third, calculate the cumulative path weight from the starting point to each adjacent node, and if the cumulative weight is smaller than the current recorded path cost of the node, update it to a smaller value and record the current node as its predecessor node; fourth, mark the current node as visited and repeat the above process until the end node is visited. Finally, the system backtracks from the end node to the predecessor nodes to generate a shortest path from the starting point to the end point, which is the optimal dispatch path.

[0093] After the optimal path is generated, the system immediately starts the real-time data updating mechanism to synchronize the resource state data set. The process includes: the system obtains the resource number from the end node of the dispatch path, locates the resource record in the resource state data set; reduces the "free parking space number" field by 1, if the result is 0, then update the "resource state" field to "unavailable"; write the "bound user number" field to record the unique identifier of the current user; write the "expected occupancy start time" field, the value is the current time of the system plus the calculated travel time required by the dispatch path, for example, the travel time is 12 minutes, then determine it by adding 12 minutes to the current time; then write the "expected release time" field, the value is equal to the start time plus the parking duration requested by the user, for example, the user requests for 90 minutes, then the release time is the start time plus 90 minutes. After the update is completed, the record is written back to the resource state data set, the system records the execution scheme of this dispatch and assigns it to the task execution queue, marking the completion of the dispatch task closed loop. The entire path calculation and data updating process is controlled within 3 seconds to ensure real-time performance.

[0094] The path optimization module generates an optimal path planning scheme based on the adjusted resource matching input data through path smoothing processing and real-time response delay optimization, and determines the parking resource allocation result, including:

[0095] An initial path planning set is generated through path smoothing processing based on the adjusted input data set from the resource matching data;

[0096] The initial path planning set is adjusted using a real-time response delay optimization algorithm to obtain an optimized path planning set;

[0097] If there is a path in the optimized path planning set whose delay exceeds the preset threshold, the unqualified path is removed through the preset path screening rule to generate a path set;

[0098] The resource allocation priority list is obtained by calculating the resource allocation priority of each path using the Dijkstra algorithm based on the path set;

[0099] The preliminary parking resource allocation scheme is generated by combining the parking resource availability data based on the resource allocation priority list;

[0100] The final parking resource allocation result is obtained by updating the resource state data set using the real-time data synchronization mechanism based on the preliminary parking resource allocation scheme;

[0101] The optimal path planning scheme is generated based on the final parking resource allocation result.

[0102] In this path optimization process, the system first obtains the adjusted input data set from the resource matching module, each piece of data including the user's current location, the geographic coordinates of the candidate parking resource, the road traffic speed, the signal light distribution, and the current traffic flow. The system uses these data to construct the basic path set between the user and the candidate resource. First, the system searches for paths through the road network model and determines the initial path from the user to each resource using the shortest distance strategy. Each path consists of several nodes and road segments, each containing length, average speed, traffic control information, etc.

[0103] In the path smoothing process, the system constructs the corner for every three consecutive nodes in the path and calculates the included angle. If the included angle is less than 45 degrees, it indicates that there is a sharp turn, and the system replaces this segment with a curve segment with a radius not less than 15 meters and reconstructs the node coordinates using the interpolation method to form a smooth path. At the same time, the system identifies the positions of all traffic signals in the path, sets the average waiting time at each light control point to 60 seconds, and adds it to the total estimated travel time of the path. In addition, the system converts the travel speed and length of each road segment into a ratio, calculates the estimated time for this segment, and adds up the time for all segments to form the preliminary travel time of the path.

[0104] Next, the system applies the real-time response delay optimization algorithm to the initial path planning set. This algorithm calls the latest traffic congestion data from the traffic data platform, sets each road segment to 500 meters, and records the current delay time in seconds for each segment. The system traverses each path, divides the path into several segments, reads the corresponding delay value for each segment, and sums them up to obtain the real-time cumulative delay time for the path. The system adds this cumulative delay time to the basic travel time to obtain the total estimated response time of the path.

[0105] Subsequently, the system performs threshold judgment on the total estimated response time of all paths. The maximum acceptable delay is preset to 300 seconds, which is obtained from user feedback surveys and represents the upper limit of the tolerance of most users to parking scheduling delays. The system traverses the path set, and if the total response time of a path is greater than 300 seconds, it marks the path as unqualified and removes it from the set based on the number. The remaining paths constitute the final optimized path set, which is input into the next stage of the resource priority sorting and allocation module. The entire process ensures that the remaining paths are smooth and meet the real-time response requirements, effectively improving the quality of path planning and user experience.

[0106] In this path resource optimization and allocation process, the system first generates a path set based on the previous step, and uses Dijkstra algorithm to perform shortest path optimization calculation on each path to evaluate its resource allocation priority. Dijkstra algorithm takes the user's current location as the starting point and the candidate parking resource location as the ending point, and models the entire road network as a graph structure, where each road segment is an edge and each intersection is a node. The system sets a comprehensive cost weight for each road in the path, which is composed of three indicators: road length, traffic delay time and number of traffic lights. The system sets the road length to be zero point five weight per meter, the traffic delay to be zero point three weight per second, and the traffic light to be zero point two weight per light. The system adds the weight of all edges of each path to get the total cost of the path, and the smaller the cost value, the higher the priority. All paths are sorted in ascending order of cost to generate a resource allocation priority list.

[0107] Subsequently, the system generates a preliminary parking resource allocation scheme based on the priority list and resource availability data. The resource availability data includes resource number, number of remaining free parking spaces, current state of the resource whether it can be allocated, and whether the resource has been locked. The system starts from the highest priority path, reads its target resource number, judges whether there is a free parking space and whether it is locked by other users, if it meets the requirements, the resource is allocated to the current user, and the user number, resource number, estimated arrival time and parking demand duration are recorded. The estimated arrival time is calculated by the path length and traffic speed, and the parking duration is given by the user's submitted demand, which is in minutes.

[0108] After completing the preliminary resource allocation, the system updates the resource state data set by calling the real-time data synchronization mechanism. The update operation includes three contents: first, the number of free parking spaces of the target resource is reduced by one, if it is reduced to zero, the resource state field is updated to be not allocatable; second, the number of the current allocated user is written into the resource data to record the identity of the user of the resource; third, the occupation time period is calculated according to the estimated arrival time and the parking duration, and written into the resource occupation time field. The update operation is synchronized to the central database and edge node through data write instruction, to ensure the consistency of the state of each node and prevent repeated allocation.

[0109] Finally, the system generates the final parking resource allocation result based on the updated resource state data and the path planning record, and outputs the optimal path planning scheme accordingly. The optimal path planning scheme is composed of the shortest path from the user's current location to the allocated resource location, including path node sequence, road segment traffic information, estimated travel time and signal waiting time. The scheme is pushed to the user terminal by the scheduling system for navigation and final landing execution, to ensure the efficient closed-loop completion of the entire reservation scheduling process.

[0110] The conflict detection module generates an initial candidate scheduling scheme set according to the event trigger records and a preset abnormal trigger threshold value. If the trigger value exceeds the threshold value, the initial candidate scheduling scheme set is generated.

[0111] The dynamic data flow of the parking resource allocation is obtained through real-time monitoring, and the time series analysis method is used to detect the path blocking event and the resource conflict event to obtain event trigger records.

[0112] According to the event trigger records, a preset abnormal trigger threshold value is used for screening. If the trigger value exceeds the threshold value, an initial candidate scheduling scheme set is generated.

[0113] From the initial candidate scheduling scheme set, resource state update data is obtained, and the real-time response capability is combined to evaluate the feasibility of each scheme to obtain a scheduling scheme set.

[0114] Through the scheduling scheme set, the A algorithm is used to calculate the path optimization priority of each scheme to obtain a priority sorting list.

[0115] According to the priority sorting list, the parking resource availability data is obtained to generate a dynamically adjusted resource allocation scheme.

[0116] Through the dynamically adjusted resource allocation scheme, a data synchronization mechanism is used to update the resource state data set of the real-time monitoring system to obtain a final scheduling scheme.

[0117] According to the final scheduling scheme, a solution to the path blocking and resource conflict is generated to determine the optimized parking resource allocation result.

[0118] In the implementation process, the conflict detection module first obtains the dynamic data flow of the parking resource allocation in real time with the help of the resource monitoring mechanism inside the system. The collected data includes resource number, allocation user number, resource path node state, resource current available state, allocation timestamp and other information. The system uses continuous time series analysis method to sort each path node state and resource state according to the timestamp into a sequence, and sets the sliding window parameter. The window length is 60 seconds and the step length is 10 seconds. In each time window, the system checks the path node state field in turn. If the same path node is marked as "closed" or "unpassable" in 3 consecutive windows, the system identifies it as a path blocking event. If a parking resource is concurrently applied for allocation by 2 different users within 10 seconds and its resource state is still "occupied" or "locked", the system determines it as a resource conflict event. Each identified event generates an event trigger record, which includes event type, resource number, involved path node, all related user numbers, event occurrence time and conflict details.

[0119] After the system reads the event trigger record, the key parameters of each event are compared with the preset abnormal trigger threshold. The path blockage event threshold is defined as the cumulative duration of continuous unpassable nodes reaching 60 seconds, and the resource conflict event threshold is two or more conflict requests for the same resource within 10 seconds. This threshold is obtained based on historical log analysis to ensure that real abnormal conditions can be captured in time, and excessive response to transient fluctuations is avoided. Any event that meets any trigger threshold is included in the dispatch recovery process and triggers the generation of an initial set of alternative dispatching schemes.

[0120] When generating the initial set of alternative dispatching schemes, the system provides multiple alternative schemes for each event. Each scheme includes alternative resource numbers, alternative path node sequences, and estimated response times calculated based on current traffic data. The system then performs a feasibility assessment on each scheme. The evaluation parameters include the current number of free parking spaces for the target resource, the estimated travel time, whether the resource is locked, and the current dispatch load of the system. Among them, the number of remaining spaces must be at least 1, the estimated travel time must be less than 300 seconds (based on the average tolerance response time of user survey), the number of concurrent tasks of the system must be less than 100 tasks, and the resource status field must be marked as "available". If a scheme meets all the above parameter requirements, the scheme is marked as a feasible dispatching scheme and enters the dispatching scheme set. The final dispatching scheme set will be used by the subsequent path optimization and resource reallocation modules to complete the dynamic recovery arrangement of path blockage or resource conflict.

[0121] Then the system applies A algorithm to each alternative scheme in the dispatching scheme set for path optimization priority calculation. A algorithm takes the user's current location as the starting point and the geographical location of the alternative parking resource as the target point to establish a path search model. In the search process, the system constructs a graph structure containing road nodes and their connection relationships, where each edge represents a road, and the edge weight is composed of three parts: the actual length of the road, the current predicted travel time, and the historical average waiting time. The system sets the weights for the above three factors, where the road length weight is 0.5, the travel time weight is 0.3, and the waiting time weight is 0.2. The total cost of each scheme's path is obtained by accumulating the weighted values of each edge, and the system sorts all schemes from small to large based on the total cost to obtain a path optimization priority ranking list.

[0122] Next, the system evaluates the resource availability of each high-priority scheme in turn according to the ranking result. The resource availability data is derived from the real-time resource state dataset, and the system determines whether the current number of vacancies of the resource is greater than 0, the resource state is "allocable", and there is no ongoing allocation task. If all the above conditions are met, the resource is confirmed to be available, and the scheme is marked as valid and added to the dynamically adjusted resource allocation scheme. If the resource is not available, the scheme is skipped, and the next scheme in the ranking list is continued to be judged.

[0123] Subsequently, the system starts the data synchronization mechanism, and writes all the resource allocation schemes that pass the feasibility verification into the resource state dataset. Each piece of written data includes the resource number, the allocated user number, the change of the current resource state to "occupied", the expected arrival time, the expected release time, and the change timestamp. The expected arrival time is the current system time plus the path cost converted into time, and the expected release time is the expected arrival time plus the parking duration booked by the user. The system ensures that all resource state information is updated synchronously to ensure data consistency.

[0124] Finally, the system extracts the change information from the resource state data after synchronization is completed, and generates the final scheduling scheme. The scheduling scheme includes path adjustment content, resource replacement situation, the final parking point of each user, expected travel time, and scheduling execution sequence, etc. According to the final scheduling scheme, the system generates the corresponding path blocking and resource conflict resolution scheme, and clearly indicates the new path and new resource that replace the original path or resource due to unreachability or conflict, so as to ensure that the user can successfully complete the entire reservation parking process and complete the optimized parking resource allocation.

[0125] The scheduling adjustment module updates the scheduling process by event priority evaluation and resource reallocation according to the alternative scheduling scheme, and determines the final scheduling execution scheme in combination with dynamic response adjustment, including:

[0126] Obtain event priority data from the alternative scheduling scheme set, and arrange the event priority by using a sorting algorithm to obtain an initial event ranking list;

[0127] According to the initial event ranking list, obtain a real-time resource state data stream, and if the resource state data stream shows that there is a resource allocation conflict, generate a conflict event record by using a conflict detection mechanism;

[0128] From the conflict event record, a preset response time threshold is used for screening, and if the response time of the conflict event exceeds the threshold, a resource reallocation scheme set is generated;

[0129] According to the resource reallocation scheme set, obtain path occupation information in the real-time data stream, and calculate the path unblocked degree of each scheme by using a path optimization algorithm to obtain an optimized path scheme set;

[0130] From the optimized path scheme set, the resource state data set of the monitoring system is updated through a data synchronization mechanism, and if the updated resource state satisfies the feasibility evaluation condition, a set of feasible scheduling schemes is determined;

[0131] According to the set of feasible scheduling schemes, real-time event trigger data is obtained, resource occupation efficiency of each scheme is calculated by using a resource utilization evaluation method, and a final scheduling execution scheme is obtained;

[0132] From the final scheduling execution scheme, the resource allocation state of the real-time monitoring system is updated through a data synchronization mechanism, and an optimized scheduling execution process is determined.

[0133] The system first reads each scheduling task to be processed from the set of alternative scheduling schemes, each task carrying event priority data calculated by the system according to four factors: the urgency of the task submission time, the user credit score, the target resource tension, and the current traffic load. The urgency of the task submission time is used to measure the time difference between the user's submission of the scheduling request and the expected execution time, reflecting the time sensitivity of the scheduling task. The system first reads the "expected start time" in the user request, then compares it with the current system time, and calculates the remaining time interval. If the time interval is less than 5 minutes, it is determined to be "highly urgent" and assigned a value of 3; if the time interval is between 5 and 15 minutes, it is determined to be "medium urgent" and assigned a value of 2; if it exceeds 15 minutes, it is determined to be "low urgent" and assigned a value of 1. This value is used as a reference factor in the scheduling order. The user credit score is calculated based on the user's historical parking behavior data. The system evaluates the user's parking records in the past 30 days, including the success rate of reservation, the cancellation rate, the number of parking violations, and the timeliness of payment. Users with a success rate of more than 90%, a cancellation rate of less than 5%, no parking violation records, and no payment delays are rated as "high credit" with a credit score of 5; users meeting 3 of the above conditions are rated as "medium credit" with a score of 3; users meeting less than 2 of the above conditions or having serious violation records are rated as "low credit" with a score of 1. This score is used in the scheduling resource priority adjustment. The target resource tension is calculated by monitoring the ratio of available parking spaces in the target parking area to the current number of reservation requests. If the ratio is less than 0.5, it indicates that the resource is highly tense, and the tension score is 3; if the ratio is between 0.5 and 1, it indicates that the resource is moderate, and the score is 2; if the ratio is greater than 1, it indicates that the resource is abundant, and the score is 1. This score is used to modify the weight of resource allocation priority to prevent scheduling resources from being too concentrated in tense areas. The current traffic load is obtained in real time by the traffic perception system, which obtains the number of vehicles passing through the target area or path segment per unit time, such as the number of vehicles passing through in 5 minutes. The system sets the traffic load reference threshold to 1000 vehicles per hour. If the current unit time passing amount exceeds 80% of the threshold, i.e. more than 800 vehicles per hour, it is judged to be "high load" with a score of 3; between 500 and 800 is "medium load" with a score of 2; less than 500 is "low load" with a score of 1. This score is used in path selection and scheduling delay tolerance calculation. After standardization, each factor is weighted and summed according to the weight ratio of 30%, 30%, 20%, and 20% to form the total priority score. Then, the system uses a quick sorting algorithm to sort all event priorities, generating an initial event sorting list in descending order, ensuring that tasks with high priority are scheduled first.

[0134] Next, the system reads the top several high-priority tasks in the sorting list and associates them with the real-time resource status data stream. The resource status data stream includes the current usage status, locked identification, whether it has been allocated, and the remaining free number of each resource. If it is found that the associated resource of a task is in a locked or remaining number of 0 state, the system determines that the task has a resource conflict, immediately generates a conflict event record and stores it in the conflict event data table.

[0135] Then, the system analyzes the conflict event records one by one, extracts the difference between the scheduling task creation time and the current system time, and calculates the response time. If the response time exceeds the system set response time threshold of 60 seconds, the task is marked as delayed scheduling and needs to enter the resource reallocation process. The threshold is obtained by user research, and the upper limit of the waiting time of most users is statistically selected as the threshold basis.

[0136] For each task that needs to be reallocated resources, the system regenerates a list of alternative resources and performs path analysis in combination with the current path state data. The path state data includes the travel time, road traffic density, and average waiting time of traffic lights of each path. The system uses a path optimization algorithm to score each path, with travel time weight 40%, traffic density weight 30%, and traffic light delay weight 30%. Each indicator is first normalized, and then weighted sum is calculated to obtain the smoothness score of each path. The system selects the top 5 paths with the highest smoothness score as the optimized path solution set, providing input basis for subsequent scheduling optimization. The whole process ensures that critical tasks are processed first, while high delay and high congestion options are screened out at the path level, improving scheduling response efficiency and user satisfaction.

[0137] From the optimized path scheme set, obtain each dispatch path and the corresponding resource use plan. The system first starts the data synchronization mechanism and uploads these paths and resource allocation information to the monitoring system main database. The update process is divided into three stages: in the first stage, the system reads each path involving road nodes and target berth numbers one by one, writes the current state of these resources, marks the nodes as "planning", and marks the berths as "to be allocated"; in the second stage, the system performs state consistency check on the resources whose state has been written. The check method is to compare the interval between the write timestamp and the latest historical state update timestamp. If the interval is not more than 10 seconds, it is considered that the state synchronization is successful, otherwise the write operation is rolled back and a conflict event is recorded; in the third stage, the system enters the feasibility evaluation process. The system calls all records marked as "planning" and "to be allocated" in the updated resource state data set, counts the number of remaining available berths and the remaining available path bandwidth (i.e. the current unoccupied path capacity), and compares them with the system set minimum available threshold. The minimum berth threshold is usually set to 10% of the total target berths, and the minimum path capacity threshold is set to 20% of the maximum load of each path. If the berth and path capacity under the updated state are both higher than the respective thresholds, it is considered that the feasibility condition is met, the system marks the path and resource scheme as "feasible", and includes it in the feasible dispatch scheme set. This process is executed in real time after each path optimization is completed, ensuring that at any time there is a stable, effective and fast response dispatch scheme set.

[0138] The system retrieves real-time event trigger data based on the set of feasible scheduling schemes, including user arrival time, parking request change records, path change events, and resource re-release records, etc. The system maps these event data to each path and corresponding resource in the scheduling scheme to establish an event association matrix. Then, it enters the resource utilization assessment stage. For each scheduling scheme, the system calculates the ratio of parking space usage time to available time to obtain the parking space utilization rate, and simultaneously calculates the ratio of the used length to the total length of each path to obtain the path utilization rate. The system assigns a weight of 50% to the parking space utilization rate and a weight of 50% to the path utilization rate, and calculates the weighted average as the resource occupation efficiency. If the resource occupation efficiency of a scheme is higher than 70%, it is considered as an efficient scheme and is preferentially retained. Subsequently, the system ranks all scheduling schemes that meet the efficiency threshold in descending order of resource occupation efficiency, and selects the top three schemes for simulation verification. Simulation verification includes simulating user path execution and resource release process within one minute, recording the scheduling response delay time, and if the time is lower than the system set ten seconds, it is confirmed as the final scheduling execution scheme. Finally, the system writes all resource allocation states in the final scheduling execution scheme into the real-time monitoring database through the data synchronization mechanism, updates the related parking space state to "allocated", updates the path node state to "in execution", and records the version number and timestamp of this write operation for subsequent audit and rollback. On this basis, the system determines the optimized scheduling execution process and generates an execution log.

[0139] The process tracking module obtains the execution state from the final scheduling execution scheme, tracks the user parking process, and generates a scheduling completion confirmation if the real-time response delay meets the preset scheduling response speed threshold, including:

[0140] The execution state data is obtained from the scheduling execution scheme, and the real-time update of the execution state is obtained by analyzing the task allocation and execution timestamp in the scheduling scheme;

[0141] According to the real-time update of the execution state, the state data is matched with the sensor data of the user parking process using data flow synchronization technology to determine the current stage of the parking process;

[0142] Through event log recording, the key event timestamp and state change in the user parking process are obtained to generate a structured event sequence;

[0143] For the structured event sequence, if the real-time response delay is lower than the preset scheduling response speed threshold, the average delay is calculated by the delay evaluation method to determine whether the scheduling task meets the response requirement;

[0144] According to the calculation result of the delay evaluation method, the state tracking mechanism is adopted to continuously monitor the state change in the user parking process to obtain the dynamic update of the parking process;

[0145] The dispatch completion confirmation is generated by confirming the generation logic, extracting the state data meeting the preset threshold from the dynamic update, and generating the dispatch completion confirmation;

[0146] If the dispatch completion confirmation is generated, the confirmation information is written into the event log recording system through data flow synchronization, and the final dispatch state is updated.

[0147] In one possible implementation, the system first extracts the detailed information of each dispatch task from the generated dispatch execution scheme, including the task number, the specific parking space number allocated to the user, the task start time, the expected completion time, and the current execution state mark. Each task data is attached with a unique identifier for subsequent state tracking. The dispatch execution scheme is stored in a structured data format, and the system periodically retrieves the state data of all current tasks through a data interface, with a sampling period of every 10 seconds. After the system starts the state analysis process, all task records are first indexed and sorted according to the task number, and then the task start time field of each task is read and compared with the current system clock. If the current time has exceeded the task start time and the task state mark is still "not started", the task state is modified to "in progress", and the actual start time of the task is recorded as the current time. The system continues to monitor the execution process of the task, and if the task state mark is "in progress" and the current time has exceeded the expected completion time, the system further detects whether the task completion confirmation mark or sensor feedback signal has been received; if not, the task state is temporarily marked as "abnormal to be confirmed". In addition, the system continuously calculates the task execution progress according to the execution time of each task, i.e. the current time minus the actual start time of the task, and updates this data to the task execution state field. To maintain the real-time nature of the task state, the system uses a combination of time stamp comparison and logical judgment to perform a complete state detection and update process at each sampling, ensuring that the task state data is refreshed in real time in the dispatch platform. Finally, all updated task state data is packaged as the current execution state data set and uploaded to the dispatch state center with the task number as the main index for use by the subsequent data flow synchronization processing module. Through the above process, the system can accurately monitor and update the execution state of all tasks, providing a basis for real-time dispatch decision-making.

[0148] The system extracts the list of tasks in "in progress" state from the real-time update dataset of execution status, and indexes and arranges them according to the task number. The state data contained in each task includes task number, parking space number, current state marker, actual start time of the task, current system time, and progress percentage. The system also acquires real-time data stream from the sensor data acquisition module of the user parking site. The sensor acquisition items include parking space pressure sensing state, vehicle approaching distance sensing, user identity confirmation signal, and user operation record timestamp, etc. The sensor data is uploaded to the system center database at a sampling period of every 5 seconds. Subsequently, the system aligns the two types of data streams according to a unified time reference through data stream synchronization technology. The system takes the current system clock as the reference, sorts the execution status data and sensor data according to the timestamp respectively, and then uses the sliding time window mechanism to set the time window width to 30 seconds, with each round of window sliding forward by 10 seconds. The state data and sensor data within the time range are extracted. Within the window, the system performs bidirectional matching operation with the task number and parking space number as the key fields, first confirms that the task parking space in the state data is consistent with the sensor detection point, and then compares the signal events in the sensor record with the task state progress logic. For example, if the task progress is between 20% and 40%, the system expects that this stage should correspond to the "user arrives at the parking area but has not completed parking" state, then the system looks for records in the sensor data whether there is a vehicle approaching distance signal less than 1 meter but the pressure plate is not activated. If the matching is successful, it is judged that the user is in the "approaching parking space" stage. If the pressure plate activation signal appears at the same time and the duration exceeds 10 seconds, it is further confirmed that the "has been parked" state is achieved, and the state is written into the task execution status field at the same time. After all the matching is completed, the system updates and writes the identified parking stage state marker and task state into the structured state record table, forming a parking process stage mapping table covering all current tasks. The mapping table takes the task number as the primary key, records the completion time of each stage, sensor event details, and state evolution track, which is used for subsequent event log analysis and dispatch completion confirmation processing. Through the above process, the system can accurately determine the current stage of the user's parking process and realize real-time linkage tracking of the state.

[0149] The system extracts all event entries related to the user's parking task by reading the event log records. Each event log contains task number, event type, event timestamp, state marker, and data source identifier. The system groups these logs according to the task number and sorts them in ascending order according to the timestamp to form a structured event sequence. The key events in the structured event sequence include reservation initiation, navigation start, vehicle approaching, parking space detection activation, parking completion, and user confirmation, etc.

[0150] Subsequently, the system performs a delay analysis on the structured event sequence. A preset dispatch response speed threshold of 10 seconds is set, representing the maximum allowed response time between the system issuing a dispatch instruction and the user actually arriving at the parking space. The system sequentially calculates the time difference between pairs of key events in the event sequence, such as the time length between "dispatch instruction sending" and "parking space approaching detection". All valid time difference values form a set of response delay samples, and the system averages the sample set to obtain the average response delay of the dispatch task. If the average response delay is less than or equal to the preset threshold of 10 seconds, the system determines that the current task response is timely and meets the dispatch requirements, entering the state tracking phase. In this phase, the system continuously collects state change data uploaded by the sensors, including vehicle entry and exit signals, user operation behaviors, system feedback states, etc. The system uses a sliding time window to detect whether the state has changed every 5 seconds, records the change type and timestamp, and updates it to the state evolution record. At the same time, the system determines whether the task meets the completion conditions according to the preset confirmation generation logic. The confirmation logic requires that the last state change is "parking completion" and has lasted more than 15 seconds, accompanied by the generation of a "user confirmation" event. If the conditions are met, the system extracts the state change record and marks it as "dispatch completion confirmation".

[0151] Finally, the system calls the data synchronization mechanism to write the dispatch completion confirmation information into the event log record system, and the record fields include task number, confirmation generation time, state category and confirmation result identifier. The system also updates the final dispatch state of the task in the dispatch management system to "completed", and the task dispatch process loop ends. Through the above whole process, the system ensures that the dispatch response delay is controllable, the state tracking is accurate, and the reliable generation and recording of the dispatch task completion confirmation are realized.

[0152] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A real-time reservation and dispatch system for shared parking, characterized in that, include: The requirements analysis module extracts parking scenario type, time requirement, and priority information from user parking request data, and generates standardized requirements descriptions through dynamic weight allocation to obtain user requirements analysis results. The process decomposition module, based on the parking scenario type and time requirements in the user demand analysis results, decomposes the parking scheduling process into business units of demand reception, resource matching and route planning, and determines the set of business units. The resource acquisition module acquires parking lot resource status and traffic flow data for the resource matching unit, and obtains resource matching input data. The initial scheduling generation module generates an initial scheduling process based on resource matching input data, through resource status parsing and fusion accuracy verification. If the fusion accuracy verification deviation is lower than the preset scheduling response speed threshold, the initial scheduling process configuration scheme is determined. Specifically, the fusion accuracy verification is performed by retrieving the corresponding status values ​​of the same resource item in different data sources, using resource number as the unit. The deviations for the two indicators, namely the number of available parking spaces and average vehicle speed, are calculated. The absolute value of the difference in the number of available parking spaces between the two data sources is taken; if the difference is less than or equal to 1, the deviation is considered to meet the accuracy requirements. Simultaneously, the difference in the average vehicle speed between the two data sources is calculated; if the difference is less than or equal to 5 kilometers per hour, the speed deviation is considered to meet the accuracy requirements. The scheduling response speed threshold is that the deviation in the number of available parking spaces does not exceed 1 unit, and the deviation in the average traffic speed does not exceed 5 kilometers per hour. If the fusion accuracy verification deviation value is higher than the preset scheduling response speed threshold, the resource status update module updates the real-time resource status dataset through dynamic weight allocation and abnormal data removal to obtain the adjusted resource matching input data. The route optimization module generates the optimal route planning scheme based on the adjusted resource matching input data through route smoothing and real-time response delay optimization, and determines the parking resource allocation result. The conflict detection module, based on the parking resource allocation results, generates alternative scheduling schemes by judging the abnormal trigger threshold if path blocking or resource conflict events are detected. The scheduling adjustment module updates the scheduling process based on alternative scheduling schemes by evaluating event priorities and reallocating resources, and determines the final scheduling execution scheme by combining dynamic response adjustments. The process tracking module obtains the execution status from the final scheduling execution plan, tracks the user parking process, and generates a scheduling completion confirmation if the real-time response delay meets the preset scheduling response speed threshold.

2. The real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The demand analysis module extracts parking scenario type, time requirement, and priority information from user parking request data, and generates standardized demand descriptions through dynamic weight allocation. The resulting user demand analysis results include: The parking scenario type, time requirement, and priority information are parsed from user request data, and the initial requirement dataset is obtained by using structured data extraction methods. If the initial requirement dataset contains missing values, it will be filled in using the preset default value filling rules to obtain the complete requirement dataset; Based on the parking scenario types and time constraints in the complete demand dataset, a decision tree algorithm is used to generate scenario classification labels, resulting in a classified demand description. By combining the classified demand descriptions with priority information and weighted allocation rules, the demand weight values ​​are calculated to obtain the weighted demand vector. If the weight value of the weighted demand vector exceeds the preset threshold, parking resources will be allocated first, and a resource allocation plan will be generated. Based on the resource allocation scheme and time constraints, a scheduling algorithm is used to optimize the parking resource allocation order to obtain the final parking allocation result. Based on the final parking allocation results and user identification information, a personalized parking demand description is generated, and user notification data is obtained.

3. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The process decomposition module, based on the parking scenario type and time requirements in the user demand analysis results, decomposes the parking scheduling process into demand reception, resource matching, and route planning business units, and determines the set of business units including: The parking scenario types and time requirements are obtained from the user demand analysis results. The structured parsing method is used to extract the demand feature data and generate an initial demand set. If there are missing parking scenario types or time requirements in the initial demand set, they will be filled in using preset default rules to obtain a complete demand set. Based on the parking scenario types in the complete demand set, the K-means clustering algorithm is used to divide the scenario categories and generate classified scenario labels; By combining the classified scene labels and time requirements with traffic flow pattern recognition, the random forest algorithm is used to predict traffic flow distribution and obtain a set of traffic flow patterns. If the traffic distribution in the traffic pattern set exceeds the preset threshold, the resource matching order will be adjusted first to generate an optimized resource allocation scheme. Based on the optimized resource allocation scheme and path planning requirements, the Dijkstra algorithm is used to calculate the optimal parking path and obtain the path planning result. Based on the path planning results and the complete demand set, a set of scheduling instructions is generated to determine the final parking scheduling scheme.

4. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The resource acquisition module, for the resource matching unit, acquires parking lot resource status and traffic flow data, and obtains resource matching input data including: Real-time monitoring acquires parking lot resource status and traffic flow data. Multi-source data is integrated through a data stream synchronization mechanism. Duplicate records are removed using a data redundancy filtering method to generate a real-time resource status dataset and obtain resource matching input data. Raw data from different monitoring points is obtained from multiple data sources, and a time-series input stream including parking lot resource status and regional traffic flow information is collected. The time-series input stream is processed by applying preset filtering rules to remove redundant information and data with format errors, thereby obtaining a cleaned data set. Based on the timestamps of each data entry in the cleaned data set, time window alignment technology is used to calibrate data from different monitoring points to a unified time base, forming a multi-source synchronous data frame. By parsing multi-source synchronous data frames, we can aggregate the parking resource status and traffic flow data associated in the multi-source synchronous data frames and construct a structured real-time resource snapshot. The exponential smoothing method is used to analyze continuous real-time resource snapshot sequences, calculate the rate of change of key indicators, predict the short-term dynamic trend of resource status, and obtain a resource trend dataset. By combining real-time resource snapshots and resource trend datasets, if the predicted rate of change of a resource exceeds a preset fluctuation threshold, the weight of the resource in the matching domain is adjusted to determine the final resource matching input data.

5. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The initial scheduling generation module generates an initial scheduling process based on resource matching input data through resource status parsing and fusion accuracy verification. If the fusion accuracy verification deviation value is lower than a preset scheduling response speed threshold, the initial scheduling process configuration scheme is determined as follows: The system obtains multi-source data streams from resource matching inputs, preprocesses the data using a data stream integration method, and generates a resource status dataset in a unified format. Based on the resource status dataset, key status features are extracted using resource status parsing techniques to obtain a resource status feature set. By fusion accuracy verification, deviation analysis is performed on the resource status feature set. If the deviation is lower than the preset scheduling response threshold, the initial scheduling process configuration scheme is determined. Based on the initial scheduling process configuration scheme, a resource allocation priority list is generated using a dynamic combination mechanism; Using the resource allocation priority list, the optimal resource scheduling path is calculated using the Dijkstra algorithm to obtain a set of path planning results; If there are scheduling conflicts in the path planning result set, the priorities are sorted according to the conflict resolution rules to generate an optimized set of scheduling instructions. Based on the optimized set of scheduling instructions, a real-time data update mechanism is adopted to synchronously update the resource status dataset and generate the final scheduling execution plan.

6. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: If the fusion accuracy verification deviation value of the resource status update module is higher than the preset scheduling response speed threshold, the module updates the real-time resource status dataset through dynamic weight allocation and abnormal data removal to obtain the adjusted resource matching input data, including: If the deviation value of the fusion accuracy verification is higher than the preset threshold, the resource matching input data is adjusted by dynamic weight allocation, abnormal data is removed, and an updated resource status dataset is generated. Based on the updated resource status dataset, a data flow integration method is used to preprocess the multi-source data to generate a standardized resource status set; By standardizing the resource status set, the K-means algorithm is used to perform cluster analysis on the resource status to obtain a resource status classification set; If there are abnormal categories in the resource status classification set, the abnormal categories will be removed according to the preset abnormal detection rules to generate an optimized resource status classification set. Based on the optimized resource status classification set, a linear regression algorithm is used to predict resource allocation priorities, resulting in a resource allocation priority list. The optimal scheduling path is calculated using the Dijkstra algorithm based on the resource allocation priority list, and the scheduling path planning result is generated. Based on the scheduling path planning results, a real-time data update mechanism is adopted to synchronously update the resource status dataset and generate the final scheduling execution plan.

7. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The path optimization module, based on the adjusted resource matching input data, generates the optimal path planning scheme through path smoothing and real-time response delay optimization, and determines the parking resource allocation result, including: Obtain the adjusted input dataset from the resource matching data, and generate an initial path planning set through path smoothing processing; Based on the initial path planning set, the paths are adjusted using a real-time response delay optimization algorithm to obtain the optimized path planning set. If there are paths in the optimized path planning set whose delay exceeds the preset threshold, then unqualified paths are removed according to the preset path filtering rules, and a path set is generated. Based on the set of paths, the resource allocation priority of each path is calculated using Dijkstra's algorithm to obtain a list of resource allocation priorities; By combining the resource allocation priority list with parking resource availability data, a preliminary parking resource allocation plan is generated. Based on the preliminary parking resource allocation plan, the resource status dataset is updated using a real-time data synchronization mechanism to obtain the final parking resource allocation result. The optimal route planning scheme is generated based on the final parking resource allocation results.

8. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The conflict detection module, based on the parking resource allocation results, if it detects a path blockage or resource conflict event, generates alternative scheduling schemes by determining an abnormal trigger threshold, including: By acquiring dynamic data streams of parking resource allocation through real-time monitoring, and using time series analysis methods to detect path blocking events and resource conflict events, event trigger records are obtained. Based on the event trigger records, a preset abnormal trigger threshold is used for filtering. If the trigger value exceeds the threshold, an initial set of alternative scheduling schemes is generated. From the initial set of alternative scheduling schemes, resource status update data is obtained, and the feasibility of each scheme is evaluated in conjunction with real-time response capabilities to obtain a set of scheduling schemes; Using the set of scheduling schemes, the path optimization priority of each scheme is calculated using the A algorithm, resulting in a priority ranking list; Based on the priority sorting list, obtain parking resource availability data and generate a dynamically adjusted resource allocation plan; By using the dynamically adjusted resource allocation scheme, the resource status dataset of the real-time monitoring system is updated using a data synchronization mechanism to obtain the final scheduling scheme. Based on the final scheduling plan, solutions for path blocking and resource conflicts are generated, and the optimized parking resource allocation results are determined.

9. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The scheduling adjustment module, based on alternative scheduling schemes, updates the scheduling process through event priority evaluation and resource reallocation, and determines the final scheduling execution scheme by combining dynamic response adjustments, including: Obtain event priority data from the set of alternative scheduling schemes, and use a sorting algorithm to arrange the event priorities to obtain an initial event sorting list; Based on the initial event sorting list, obtain the real-time resource status data stream. If the resource status data stream shows that there is a resource allocation conflict, generate a conflict event record through the conflict detection mechanism. From the conflict event records, a preset response time threshold is used for filtering. If the response time of a conflict event exceeds the threshold, a set of resource reallocation schemes is generated. Based on the set of resource reallocation schemes, obtain path occupancy information from the real-time data stream, use path optimization algorithms to calculate the path availability of each scheme, and obtain a set of optimized path schemes. From the set of optimized path schemes, the resource status dataset of the monitoring system is updated through a data synchronization mechanism. If the updated resource status meets the feasibility assessment conditions, the set of feasible scheduling schemes is determined. Based on the set of feasible scheduling schemes, real-time event triggering data is obtained, and the resource utilization evaluation method is used to calculate the resource utilization efficiency of each scheme to obtain the final scheduling execution scheme. From the final scheduling and execution plan, the resource allocation status of the real-time monitoring system is updated through a data synchronization mechanism to determine the optimized scheduling and execution process.

10. A real-time reservation and dispatch system for shared parking according to claim 1, characterized in that: The process tracking module obtains the execution status from the final scheduling execution plan, tracks the user parking process, and generates a scheduling completion confirmation if the real-time response delay meets a preset scheduling response speed threshold. Execution status data is obtained from the scheduling execution plan, and real-time updates of the execution status are obtained by parsing the task allocation and execution timestamp in the scheduling plan; Based on the real-time updates of the execution status, data stream synchronization technology is used to match the status data with the sensor data during the user's parking process to determine the current stage of the parking process. By recording event logs, key event timestamps and status changes during the user's parking process can be obtained, and a structured event sequence can be generated. For structured event sequences, if the real-time response latency is lower than the preset scheduling response speed threshold, the average latency is calculated using a latency evaluation method to determine whether the scheduling task meets the response requirements. Based on the calculation results of the delay evaluation method, a state tracking mechanism is adopted to continuously monitor the state changes of the user during the parking process and obtain dynamic updates of the parking process. By confirming the generation logic, state data that meets the preset threshold is extracted from the dynamic update, and the scheduling completion confirmation is generated. If the scheduling is completed and confirmation is generated, the confirmation information is written to the event log system through data stream synchronization to update the final scheduling status.

Images