CN121982925A

CN121982925A - Intelligent garage management method and system for collaborative dynamic guidance

Info

Publication number: CN121982925A
Application number: CN202610163669.5A
Authority: CN
Inventors: 李江; 刘双生; 何兴亮; 李喜元
Original assignee: Hubei Anxin Digital Intelligence Information Technology Co ltd
Current assignee: Hubei Anxin Digital Intelligence Information Technology Co ltd
Filing date: 2026-02-05
Publication date: 2026-05-05

Abstract

The invention discloses an intelligent garage management method and system for collaborative dynamic guidance, and aims to improve the utilization rate of parking resources and the guidance efficiency. The method receives a parking request containing user preferences and vehicle attributes, recommends and reserves a target parking lot. After a vehicle enters a field, the system acquires real-time data through the Internet of things, combines historical data and generates in-field dynamic situation information by utilizing a space-time fusion prediction model. Based on the method, the system executes multi-target dynamic optimal parking space allocation, and plans the driving path with the lowest comprehensive passing cost. In the guiding process, the system continuously monitors, and when the system is jammed or the parking space is occupied, the system immediately triggers the rescheduling of the parking space and the path. The corresponding system comprises modules of request processing, in-field sensing, data fusion and situation prediction, dynamic parking space allocation, real-time path planning, state monitoring, re-planning, charging updating and the like, and achieves automatic and intelligent management from reservation, guidance to off-site charging through a cloud, so that parking efficiency and user experience are remarkably improved.

Description

Intelligent garage management method and system for collaborative dynamic guidance

Technical Field

The invention relates to the field of intelligent garage management, in particular to a collaborative dynamic guiding intelligent garage management method and system.

Background

With the acceleration of the urban process and the rapid increase of the quantity of motor vehicles held, the 'parking difficulty' has become a common problem which plagues urban traffic management and daily travel of residents.

The traditional parking lot management mode is extensive and mainly depends on manual guidance and static identification, so that a vehicle owner often faces the dilemma of difficult position finding, slow passing and messy management after entering a garage, and a large amount of time is required to blindly find an idle parking place, thereby not only reducing parking experience, but also exacerbating traffic jam and potential safety hazard in the garage.

To address the above challenges, smart parking systems have been developed and developed rapidly. The prior art mainly evolves along several directions, namely, firstly, the occupied state of the parking space is detected in real time by deploying a sensor network (such as ultrasonic waves, geomagnetism and video piles), and idle parking space information is issued to a vehicle owner in a mode of a parking space guide screen, an indicator light and the like, so that basic parking space guide is realized. Secondly, the technology of the Internet of things and the technology of the mobile Internet are combined, so that a user is allowed to remotely inquire the parking space and reserve the parking space through a mobile phone application program, and navigation is performed in the field through an electronic map or an indicator lamp. Third, more advanced image recognition, artificial intelligence algorithms are introduced in an attempt to more accurately recognize and schedule vehicles. For example, some schemes evaluate parking space suitability by analyzing historical vehicle images, or dynamically optimize garage space layout and vehicle access sequence using algorithms.

However, the existing intelligent garage management system still has a plurality of limitations, and cannot realize real collaborative dynamic guidance. The specific expression is as follows:

the system is isolated, and most of system functional modules are relatively independent. Links such as parking space detection, path guidance, reverse vehicle searching, cost payment and the like are often realized by different subsystems, data flow is broken, and closed loop collaborative optimization is difficult to form. For example, real-time traffic flows, user personalized preferences (e.g., approaching elevators, charging piles) and future occupancy probabilities of the parking spaces are rarely comprehensively considered when the parking spaces are allocated, so that the recommended parking spaces may not be globally optimal.

Static or semi-static guidance has poor dynamic adaptability, and the existing guidance strategy is mostly based on static parking space state data at a certain moment. Once planning is completed, the system is difficult to adjust in real time according to the transient traffic conditions (such as sudden congestion and vehicle crossing) in the field. Studies have shown that incorrect road selection consumes significant driver time, and existing systems lack predictive and responsive mechanisms for dynamic traffic flow.

The data is shallow, lacks situation awareness and prediction capability, and multidimensional data (such as parking space state, vehicle passing speed and history rule) collected by the system are not deeply fused and deeply mined. The prediction capability of the future short-time parking space occupation and the traffic jam situation in the parking lot cannot be constructed. This results in the system only reacting "when it is down" and not being able to proceed with prospective planning and scheduling to prevent congestion or increase the turnover rate of the parking space.

The characteristics of the users and the vehicles are not considered enough, the vehicles are mostly regarded as homogeneous objects in the existing scheme, and whether gaps exist or not is mainly considered when the parking spaces are allocated, so that the individual requirements of the vehicle size, the vehicle type and the users are ignored. Meanwhile, the whole-course personalized service chain from outside to inside is not yet opened, for example, the most convenient parking lot and parking space are intelligently recommended for the user according to the final destination of the user.

In summary, the current intelligent garage management system has obvious defects in the aspect of collaborative dynamic guidance.

Therefore, a novel intelligent garage management method and system capable of integrating multi-source information, realizing on-site and off-site coordination and performing continuous optimization decision based on real-time dynamic situation are urgently needed, so that parking efficiency is fundamentally improved, user experience is optimized, and maximum utilization of garage resources is realized.

Disclosure of Invention

In order to solve the obvious defects of the existing current intelligent garage management method and system in the aspect of collaborative dynamic guiding, the intelligent garage management method and system for collaborative dynamic guiding are provided.

An intelligent garage management method with collaborative dynamic guidance comprises the following steps:

S1, receiving a parking request of a user side, and acquiring information of a plurality of surrounding candidate parking lots based on destination information in the request;

s2, determining and reserving a target parking lot for a user based on user preference information and information of the candidate parking lots;

S3, acquiring real-time multidimensional state data in the target parking lot, and carrying out fusion analysis based on the data and historical data to generate and continuously update in-field dynamic situation information for representing future parking space occupation and traffic conditions;

S4, responding to the vehicle entering the target parking lot, and dynamically distributing the optimal parking space for the vehicle based on multi-target decision by combining the in-field dynamic situation information, the vehicle attribute and the user preference;

S5, dynamically planning a driving path from the current position to the optimal parking space for the vehicle based on the in-field dynamic situation information and the vehicle attribute;

s6, the optimal parking space information and the running path are issued to a user side for guiding, and a guiding strategy is dynamically adjusted according to environmental changes in the guiding process;

S7, updating the parking space state and completing the charging operation.

Further, in the step S3, the fusion analysis is implemented by modeling the parking space topology to capture the spatial dependency relationship, and combining the historical and real-time sequence data analysis to capture the time law.

Further, the step S4 of dynamically allocating the optimal parking space for the vehicle based on the multi-objective decision comprises the steps of screening candidate parking spaces conforming to the physical constraint of the vehicle from the current idle parking spaces;

Calculating a comprehensive evaluation value for each candidate parking space by calculating a weighted average, wherein the comprehensive evaluation value at least comprises estimated traffic time based on real-time and predicted traffic flow, walking distance reaching user preference facilities, future idle probability of the parking space based on the in-situ dynamic situation information and matching degree of the parking space attribute and user preference;

And the parking space with the highest comprehensive evaluation value is allocated as the optimal parking space.

Further, in the step S5, when the driving path is dynamically planned for the vehicle, the path cost evaluation integrates the static physical length of the road section, the traffic time cost acquired based on the real-time sensor data, the predicted congestion influence based on the on-site dynamic situation information, and the traffic difficulty coefficient based on the vehicle attribute.

Further, in the step S6, the dynamically adjusting the guiding strategy according to the environmental change includes continuously monitoring traffic flow and target parking space state in the field during the running process of the vehicle, and triggering a re-planning process if the congestion degree of the planned path is detected to exceed a specified threshold or the parking space state is abnormal, and re-executing the parking space allocation and path planning by taking the current position of the vehicle as a new starting point and combining the latest environmental data.

A collaborative dynamic boot intelligent garage management system for implementing the foregoing method, comprising:

The request processing and recommending module is deployed at the cloud and is used for executing the steps S1 and S2;

the in-field sensing module is distributed in the parking lot and used for executing the step S3;

The data fusion and situation prediction module is deployed at the cloud end and is used for executing the step S3;

the dynamic parking space allocation module is deployed at the cloud end and is used for executing the step S4;

The real-time path planning module is deployed at the cloud end and is used for executing the step S5;

the state monitoring and rescheduling module is deployed at the cloud end and is used for executing the step S6;

the charging and status updating module is deployed at the cloud end and is used for executing the step S7;

The system comprises a request processing and recommending module, an in-field sensing module, a data fusion and situation prediction module, a dynamic parking space allocation module, a real-time path planning module, a state monitoring and rescheduling module and a charging and state updating module, wherein data interaction and instruction transmission are carried out through a network, and the whole-course dynamic guiding and management from receiving a user request to completing charging are completed cooperatively.

Further, the request processing and recommending module comprises a user interface unit and a recommending engine unit;

the user interface unit is used for interacting with a user terminal to receive a structured parking request data packet and transmitting the data packet to the recommendation engine unit;

The recommendation engine unit is used for calculating and determining a target parking lot based on destination information, user preference and multi-source external data in the data packet, and issuing a reservation result to a user side through the user interface unit;

the data fusion and situation prediction module comprises a sensor network unit and a prediction analysis unit;

the sensor network units are distributed in the parking lot and are used for collecting multidimensional state data in real time;

The prediction analysis unit is used for receiving and fusing the real-time data and the historical data uploaded by the sensor network unit, and generating in-field dynamic situation information through a space-time fusion prediction model;

The dynamic parking space allocation module and the real-time path planning module comprise a decision unit and a path calculation unit;

The decision unit is used for receiving the in-field dynamic situation information, the vehicle attribute and the user preference, and outputting an optimal parking space allocation result according to a multi-objective decision model;

the path calculation unit is used for receiving the optimal parking space information, the real-time position of the vehicle and the dynamic situation information in the field and calculating a driving path based on a dynamic cost map;

the state monitoring and rescheduling module comprises an abnormality detection unit and a rescheduling triggering unit;

The anomaly detection unit is used for continuously receiving feedback from the sensor network and the vehicle positioning data and monitoring traffic flow and parking space states;

The re-planning triggering unit is used for triggering the decision unit and the path calculation unit to re-execute calculation when the abnormality detection unit judges that the re-planning condition is met.

Further, the sensor network unit in the in-field sensing module includes:

the parking space state sensing subunit is deployed in each parking space and used for detecting the occupation state of the parking space and identifying the vehicle license plate;

The traffic flow sensing subunit is deployed in the key channel and is used for collecting traffic flow, speed and queuing length data;

The environment and event sensing subunit is used for collecting facility state and safety event data;

and the vehicle positioning beacon subunit is deployed in the field and is used for providing real-time positioning data for the vehicle.

Further, the real-time path planning module plans a travel path by:

Constructing a dynamic cost map, wherein the cost of each road section integrates the static physical length, the passing time based on real-time sensor data, the predicted congestion influence based on the in-field dynamic situation information and the passing difficulty coefficient based on the vehicle attribute;

And searching an optimal path from the current position of the vehicle to the optimal parking space on the dynamic cost map by adopting a D Lite algorithm.

The intelligent garage management method based on the collaborative dynamic guidance has the advantages that firstly, the intelligent garage management method based on the collaborative dynamic guidance breaks through data barriers between the inner part and the outer part of the parking lot and between the inner part and the outer part of the parking lot, fusion and collaborative decision of user preference, parking resources and real-time situations in the parking lot are achieved, and regional traffic operation efficiency is improved. And secondly, the system realizes the capability improvement from the passive display of the current vacancy to the active prediction of the future short-time parking space occupation and traffic jam situation through the modeling analysis of the parking space topology and time sequence data. On the basis, the system introduces a multi-objective optimization model, can dynamically allocate parking spaces for each vehicle, and plan an optimal path for comprehensive static and real-time traffic cost and prediction of congestion influence, so that the locating time is greatly shortened. More importantly, the system is provided with a dynamic re-planning mechanism, and can be recalculated in real time when the congestion of a path or the abnormality of a parking space is detected, so that the robustness and the reliability of a guiding strategy are ensured.

Drawings

FIG. 1 is a flow chart of an embodiment of the method of the present invention;

FIG. 2 is a flow chart of dynamic parking space allocation decision in an embodiment of the invention;

FIG. 3 is a flow chart of real-time path planning according to an embodiment of the present invention;

FIG. 4 is a flow chart of a monitoring and dynamic re-planning mechanism according to an embodiment of the present invention;

fig. 5 is a schematic diagram of a system module architecture and a data flow according to an embodiment of the invention.

Detailed Description

In order to make the objects and technical solutions of the present invention more clear, the following detailed description of the specific embodiments of the present invention will be given with reference to the accompanying drawings and examples.

As shown in fig. 1, the intelligent garage management method with collaborative dynamic guidance includes the following seven steps:

S1, receiving a parking request of a user side, and acquiring information of a plurality of surrounding candidate parking lots based on destination information in the request, wherein the user side is further preferably mainly applied to a mobile intelligent terminal, such as a special application program (APP) of a smart phone and a tablet computer, or a vehicle-mounted application program integrated in a vehicle-mounted information entertainment system (central control screen). For intelligent networked automobiles with vehicle-to-road (V2X) communication capability, the customer premise functions may also be integrated in an on-board unit (OBU).

Further preferably, the parking request is a structured data packet comprising at least:

1. vehicle identification, license plate number.

2. Vehicle attributes including brand, vehicle type (such as sedan, SUV), size (length, width, height), power type (fuel oil vehicle, pure electric vehicle, plug-in hybrid).

3. Destination information, namely a target place name input by a user or longitude and latitude coordinates acquired through map point selection.

4. User preference, optional preference tags such as "need to charge stake", "need for unobstructed parking space", "get preferentially closer to elevator/exit", "expect lowest cost", etc.

5. The estimated time of arrival is manually entered by the user or estimated time by the system based on real-time location and road conditions.

S2, determining and reserving a target parking lot for a user based on user preference information and information of the candidate parking lots, wherein the user preference information is a personalized image established for the user by a system and further comprises the following steps:

1. the static preference is that users clearly specify long-term requirements in registration or setting, such as 'new energy vehicle owners' (default needs to charge piles) 'mobility impaired people' (default needs to have barrier-free parking spaces).

2. Dynamic preferences-temporary requirements specified in a single request, such as "near mall number 3 gate" for the present shopping hope.

3. Historical behavior preferences the system mines implicit preferences by analyzing the user's historical parking records, such as frequently selecting certain dining floors to park on weekends.

Further preferably, the real-time multidimensional state data is continuously collected through an internet of things sensor network deployed in a parking lot, and mainly comprises:

1. Parking space level data, namely the occupation state (idle/occupied/reserved locking), the type of the parking space (common, charged, accessible and large-sized vehicles) and the physical size of each parking space.

2. Traffic flow data, namely real-time vehicle passing speed, traffic flow and queuing length of each lane and intersection.

3. Event data, such as abnormal parking, retrograde, congestion, accidents and other alarm events.

4. Facility status data, occupancy/idle/fault status of the charging pile, current running status of the elevator/escalator and waiting population (estimation).

5. And environment data, namely a video monitoring stream of a key position, which is used for visual analysis and verification.

S4, responding to the vehicle entering the target parking lot, and dynamically distributing the optimal parking space for the vehicle based on multi-target decision by combining the in-field dynamic situation information, the vehicle attribute and the user preference (contained in a data packet of a parking request);

The in-field dynamic situation information is fusion analysis output is structured prediction data, and mainly comprises:

1. The parking space occupation probability field is a matrix covering all the parking spaces, and each element value is the probability that the parking space is occupied in a specific future time period (such as the next 10 minutes).

2. And a traffic time prediction graph, namely a weighted directed graph covering all main roads and intersections, wherein the weight of each side is the predicted future average traffic time.

3. And (3) a congestion risk index chart, which identifies areas and levels where congestion is likely to occur in the future.

S7, updating the parking space state and completing the charging operation.

Based on the previous steps S1-7, preferably, in step S3, the fusion analysis is implemented by a "spatiotemporal fusion prediction model". The space-time fusion prediction model comprises the following contents:

1. Space modeling, namely abstracting a parking lot into a topological graph, wherein nodes represent parking spaces, lane intersections and facility points, edges represent connection paths and are given weight (such as length and width). Spatial dependence is captured using a Graph Neural Network (GNN) (e.g., a regional congestion may spread to adjacent channels).

2. And (3) time sequence analysis, namely training the historical data through deep learning (such as LSTM, transformer model), and learning the parking space occupation, the periodicity of traffic flow and the trend rule.

3. And (3) merging, namely inputting the multidimensional data acquired in real time into a trained space-time model, predicting the occupancy probability of each parking space and the passing time of each path in the future of 5-15 minutes in real time, and generating a dynamically updated situation map.

According to the optimal scheme, the prediction of the future state in the parking lot is realized by establishing a space topology model of the parking lot and fusing time sequence data analysis. The discrete sensor data are associated, so that the system not only knows the vacancy information, but also can pre-judge the occurrence possibility of the vacancy and the occurrence probability of congestion, thereby carrying out parking space allocation and path planning in advance and fundamentally avoiding guide failure and secondary congestion caused by information lag.

Based on the foregoing steps S1-7, as shown in fig. 2, preferably, dynamically allocating an optimal parking space for the vehicle based on the multi-objective decision in the step S4 includes screening candidate parking spaces conforming to the physical constraint of the vehicle from the current idle parking spaces;

Further preferably, the dynamic allocation is an online real-time optimization process:

1. triggering when the vehicle enters the entrance of the parking lot.

2. And (3) screening, namely firstly filtering out the parking spaces which do not accord with the physical size and the basic type of the vehicle (such as a non-charging vehicle cannot occupy a charging potential) from the idle parking spaces in the whole field to form an initial candidate set.

3. And scoring, namely calling a multi-objective decision model to calculate the comprehensive score for each candidate parking space. The model input comprises real-time vehicle position, dynamic situation information, user effective preference and parking space static attribute. A composite score is calculated by weighted summation (weight configurable).

4. And deciding, namely selecting the parking space with the highest score, immediately updating the state of the parking space into locked state in the system, and issuing an allocation result through a network.

The optimal scheme has the technical effects that global optimization is realized by comprehensively considering traffic efficiency (time), user convenience (walking distance), resource utilization (future idle probability) and personalized requirements (preference matching). This ensures that the allocated parking space is not only physically available, but is optimally selected according to the overall cost and user experience.

Based on the foregoing steps S1-7, as shown in fig. 3, preferably, in the step S5, when the driving path is dynamically planned for the vehicle, the path cost evaluation integrates the static physical length of the road section, the traffic time cost acquired based on the real-time sensor data, the predicted congestion influence based on the on-site dynamic situation information, and the traffic difficulty coefficient based on the vehicle attribute.

The optimal scheme has the technical effects that the planned path is not the shortest geometric path on the map, but the space-time optimal path with the lowest current and recent comprehensive passing cost. The method can actively avoid real-time congestion points and predicted congestion areas, and avoid narrow road sections for special vehicles such as large vehicles. The vehicle has the advantages that invalid running and waiting time in the field are reduced, traffic flow in the field is effectively dredged, accident risk is reduced, and overall passing efficiency is improved.

As shown in fig. 4, in the step S6, preferably, the dynamically adjusting the guiding strategy according to the environmental change includes continuously monitoring traffic flow and target parking space state in the field during the running process of the vehicle, and triggering a re-planning process if the congestion degree of the planned path is detected to exceed a specified threshold or the parking space state is abnormal, and re-executing the parking space allocation and path planning by taking the current position of the vehicle as a new starting point and combining the latest environmental data.

The guiding is realized through multi-mode man-machine interaction:

1. The mobile terminal APP guides to display a dynamically updated map in a field on the mobile phone APP of the user, and a planned path is displayed by using a highlight line and is assisted by text and voice navigation instructions (such as 'forward intersection left turn, forward B region').

2. And guiding facilities in the field, wherein the system controls indicator lights (such as green light flashing of reserved parking spaces) on the parking spaces, and displays the direction of the target parking space area and the quantity of the residual parking spaces on a shunting LED screen of the key intersection.

3. And the vehicle-mounted terminal guides that the path information can be directly issued to the vehicle navigation system for the vehicle with the vehicle-to-vehicle function.

The core of dynamic adjustment is continuous monitoring and triggered re-planning.

The monitoring is that the system tracks the position of the vehicle in real time (through the Bluetooth beacon triangulation or camera recognition) and monitors the traffic flow on the planned path and the state of the target parking space.

Triggering conditions are that when the front path congestion index exceeds a threshold value or (b) the target parking space is accidentally occupied by other vehicles (robbery position) is detected, the vehicle is immediately triggered.

And (3) re-planning the operation, namely interrupting the current guiding instruction by the system, taking the latest real-time position of the vehicle as a starting point, combining the latest dynamic situation information, re-executing the step S4 (possibly distributing a nearby new vehicle position) and the step S5 (planning a new path), pushing the new vehicle position and the path information to the user side in real time, and updating the guiding instruction.

As shown in fig. 5, the collaborative dynamic guiding intelligent garage management system for implementing the collaborative dynamic guiding intelligent garage management method includes a request processing and recommending module deployed at a cloud end for executing steps S1 and S2;

Preferably, the request processing and recommending module comprises a user interface unit for interacting with a user terminal and receiving a parking request, and a recommending engine unit for calculating a recommending score of the parking lot based on multi-source information, wherein the user interface unit is a bridge for data exchange between the system and various user terminals. It provides services in the form of Application Programming Interfaces (APIs). For the personal user, the carrier is mainly an intelligent mobile phone application program (APP) or an applet integrated with WeChat and payment treasures, and for the vehicle-mounted scene, interaction is carried out through an embedded client of a vehicle-mounted information entertainment system or a vehicle-mounted unit conforming to a vehicle-road cooperation (V2X) communication protocol. The unit receives structured parking request data packets, which specifically include vehicle identification (such as license plate number), vehicle attributes (brand, vehicle type, length, width, height, size of power type such as fuel/pure electricity/plug-in mix), destination information (destination location name or longitude and latitude coordinates), user preferences (such as "need charge stake", "preferentially approach elevator", "lowest cost expected" and other labels), and expected arrival time.

Recommendation engine unit, which is the core decision unit of the module. It accesses multisource external data services including geocoding services for map facilitators (e.g., germany, hundred degrees), and city level parking platform data. The system maintains a parking lot static information database inside, and stores basic attributes (position, total number of car positions, floor structure, facility distribution and rate rules) of each networking parking lot. After receiving the request, the engine executes the following processes of firstly searching a plurality of candidate parking lots in the surrounding database according to destination information, secondly obtaining information such as real-time vehicle number, estimated queuing time of an entrance and the like of each parking lot through a real-time data interface of each parking lot, and finally comprehensively evaluating the matching degree of each candidate parking lot and the user request based on a set of configurable multi-objective scoring model. Scoring factors include static facility match (whether there is a charging peg), space accessibility (distance traveled versus time), destination convenience (distance walked from parking lot to destination), economy (rate), and real-time load conditions. The highest scoring parking lot is determined as a target parking lot, the unit then calls a reservation interface of a parking lot system to lock the parking space qualification, and reservation success information and navigation initial guide information are returned to the user side through a user interface unit.

The in-field sensing modules are distributed in each parking lot and are responsible for comprehensively and real-timely collecting multidimensional state data of the physical world in the field. The in-field sensing module comprises a sensor network unit and a data aggregation and preprocessing unit.

The sensor network unit consists of heterogeneous sensor nodes and data acquisition hardware all over the parking lot. The method specifically comprises the following steps:

And the parking space state sensing subunit is used for deploying video parking space cameras or geomagnetic sensors on each parking space and detecting the idle or occupied or reserved locking state of the parking space in real time. The video camera may also perform license plate recognition to correlate to a particular vehicle.

The traffic flow sensing subunit is used for deploying wide-angle monitoring cameras at key positions such as main roads, intersections, slopes and the like, internally arranging a video analysis algorithm, calculating traffic flow, average speed and vehicle queuing length in real time, and complementarily deploying millimeter wave radars in areas with complex visual conditions to provide more stable target detection and speed measurement.

The environment and event sensing subunit is used for deploying a passenger flow statistical camera in an elevator hall and a stairwell, integrating an intelligent ammeter and a communication module on a charging pile, reporting the working state (idle/charging/fault), and automatically detecting safety events such as abnormal parking, retrograde running, pedestrian intrusion and the like by analyzing a global video stream.

Vehicle locating beacon subunit, which regularly deploys Bluetooth beacons (iBeacon/Eddystone) or Ultra Wideband (UWB) locating base stations on the ceilings of parking lots to provide real-time position information with meter-level precision for entering vehicles.

The data aggregation and preprocessing unit typically runs on an edge computing server local to the parking lot. It receives the raw data stream from all sensor subunits over a wired (ethernet) or wireless (industrial Wi-Fi) network. The method has the core functions of data cleaning (noise and abnormal value filtering), timestamp synchronization and format standardization, and locally caching the processed structured state data (for example, the state of a parking space of a B region 023 is occupied, the license plate is A12345 and the time is T), and uploading the structured state data to a data fusion and situation prediction module of a cloud in real time through a 4G/5GCPE or an optical fiber private line.

The data fusion and situation prediction module comprises a sensor network unit for acquiring parking space occupation, vehicle traffic and facility use state data in real time and a prediction analysis unit for carrying out fusion analysis on the data and generating in-field dynamic situation information, and is deployed on a cloud high-performance computing platform and responsible for converting massive real-time and historical data into insight into future situations.

And the data storage and management unit is responsible for the persistent storage and organization management of the system data. The system adopts a hybrid database architecture, wherein a time sequence database (such as InfluxDB) is used for efficiently storing and inquiring time-stamped sequence data (such as parking space state change and traffic flow) generated by a sensor, a relational database (such as MySQL) is used for storing relational data such as parking lot structure information, user files and event logs, and an object storage service is used for storing unstructured data such as video clips and pictures. The historical data mainly comprises the time-sharing occupancy rate of the parking spaces accumulated for a long time, the traffic flow mode rule, event occurrence statistics and the like.

And the prediction analysis unit is a core algorithm unit of the module and realizes 'fusion analysis'. The interior of the model runs a space-time fusion prediction model, and the model works by the following steps:

Space topology modeling, namely abstracting a parking lot physical structure into a graph structure. Nodes represent key positions such as parking spaces, lane intersections, elevator doors and the like, edges represent connecting paths, and attributes such as length, width, direction and the like are given. This graph is used to encode spatial dependencies.

And (3) time sequence rule learning, namely training historical time sequence data by using a deep learning model such as a Long Short Term Memory (LSTM) or a Transformer, and learning the periodicity (such as the morning and evening peaks of working days), the trending and the event relevance of parking space occupation and traffic jam.

And (3) fusing and predicting in real time, namely injecting multidimensional state data (serving as observation input) uploaded by the in-field perception module in real time into the trained model. The model performs rolling prediction by combining the relevance of the space diagram and the learned time law, and outputs the dynamic situation information in the field. The information is specifically expressed by a parking space occupation probability matrix (predicting the probability of each parking space being occupied for 5-15 minutes in the future), a passage passing time prediction graph (predicting the passing time of each path segment in the future), and a congestion risk thermodynamic diagram (identifying areas and levels of possible congestion in the future).

The dynamic parking space allocation module and the real-time path planning module comprise a decision unit for executing optimal allocation of parking spaces according to the situation information, the vehicle attributes and the user preferences, and a path calculation unit for planning a driving path based on the situation information and the vehicle attributes;

Decision unit this unit is the execution core of the allocation logic. The working flow is as follows:

The candidate set is generated by rapidly screening a parking space set conforming to physical constraints of vehicles from full-field parking spaces, wherein the constraints comprise that the sizes (length, width and height) of the parking spaces are required to accommodate the outlines of the vehicles, the types of the parking spaces are matched with the requirements of the vehicles (such as the electric vehicles can be only allocated with charging parking spaces), and the current state of the parking spaces is idle.

And (3) multi-target comprehensive scoring, namely calculating a comprehensive evaluation value for each candidate parking space. The scoring function uses a weighted sum model: score=w1× f1+w2×f F1+w2 x F. Wherein:

F1 The estimated time for reaching the parking space is calculated based on the current position of the vehicle, the real-time traffic flow and the traffic time prediction diagram in the situation prediction diagram.

F2 (arrival preference facility walk distance) the shortest path distance from the stall walk to the user's preferred target facility (e.g., a particular elevator gate) is calculated.

F3 The future idle probability of the parking space is obtained from the complement number (1-occupation probability) of the probability that the parking space is occupied in the future in the situation information, and the parking space which is more likely to be kept idle in the future is preferentially allocated so as to improve the turnover rate.

F4 And (5) quantifying the fit degree of the parking space attribute (such as the power of a charging pile, whether the parking space is an unobstructed parking space or not and the quietness degree of the area) and the preference (static state and dynamic state) of the user.

W1, w2, w3, w4 are dynamically configurable weight coefficients.

And (3) optimal decision, namely selecting the parking space with the highest comprehensive evaluation value, and immediately marking the state of the parking space as 'allocated and locked' through an internal interface of the system.

And the path calculation unit is responsible for calculating the optimal path in the dynamic environment. The working flow is as follows:

And constructing a dynamic cost map, namely dynamically endowing each road section (edge) in the map with comprehensive passing cost by taking a static topological map of the parking lot as a base map and combining real-time sensor data (current passing speed) and in-field dynamic situation information (predicted passing time and congestion risk). The cost calculation integrates static physical length, real-time transit time cost, prediction of congestion influence and transit difficulty coefficient based on vehicle attributes (such as that a large vehicle needs to avoid narrow curves).

And searching an optimal path, namely searching on the constructed dynamic cost map by adopting an A-algorithm or a dynamic version D-Lite algorithm (which are path planning algorithms) of the A-algorithm. The method takes the real-time position of the vehicle (provided by a positioning beacon) as a starting point and takes the allocated optimal parking space as an end point to find the path with the minimum comprehensive passing cost. The output result is a series of ordered lane nodes and steering instructions.

The monitoring and rescheduling module comprises an abnormal monitoring unit for continuously monitoring traffic flow and parking space states and a rescheduling triggering unit for triggering the reassignment of parking spaces or planning paths when conditions are met.

The anomaly detection unit is used for continuously monitoring two key aspects, namely whether the vehicle deviates from a planned path or abnormally stagnates or not in real time through data of the vehicle positioning beacon unit, and whether the traffic congestion index on the planned path exceeds a threshold value and whether the state of a target parking space is occupied accidentally or not is monitored through real-time data flow of the sensor network unit.

The rescheduling triggering unit is triggered immediately when the abnormality detection unit finds any one of the conditions that the front of the planned path suddenly seriously jams, the target parking space is occupied by other vehicles, and the vehicles seriously deviate from the path and cannot automatically return. After triggering, the unit sends an interrupt signal to the system core, takes the latest real-time position of the vehicle as a new starting point, carries latest environmental data, recalls a decision unit of the dynamic parking space allocation module and a path calculation unit of the real-time path planning module, generates a new parking space and path combination scheme, and delivers the new parking space and path combination scheme to the guiding execution unit.

And the charging and state updating module is responsible for the closed loop of the parking service and the synchronous maintenance of the system resource state.

Charging and status updating unit, which is the end point of the service logic. It is depth-linked with the in-field sensing module. When the vehicle is confirmed to be parked in the allocated parking space (through video recognition or geomagnetic sensor state change combined with positioning information), the parking start is marked. When the vehicle leaves (through exit gate identification or parking space state change), the stop is marked. And automatically calculating the fee according to the parking time, the type of the parking space (such as different charging parking space rates) and the preset rate rule, and completing fee deduction through integrating a third party payment gateway (such as WeChat payment and treasury payment). Meanwhile, the unit is responsible for changing the key states of parking space states (from occupied to idle), vehicle departure events and the like, synchronously updating the key states to a central database of the system, and ensuring the consistency of the global states.

The application case (refer to figures 1-5) is that before the mr. King starts, the intelligent parking APP on the mobile phone is opened, and a destination 'city center market' is input. The APP (user interface unit) automatically acquires his vehicle information (license plate number, vehicle model SUV, new energy vehicle) and general preference ("need for charging peg" "" preferentially approaching elevator ") after obtaining the authorization of the user. He clicks "find parking spot". The APP sends this parking request (including destination, vehicle attributes, user preferences) to the cloud.

After receiving the request, the cloud recommendation engine unit searches all networking parking lots within 1 km around according to destination coordinates to obtain candidate parking lot information of the networking parking lots, wherein the candidate parking lot information comprises positions, total car number, current empty car number, whether charging piles exist or not, charging standards and estimated walking distances from all entrances of a mall to a target mall area. The engine then performs a composite score based on the mr. King's preferences, with a parking lot having sufficient free charge piles and the shortest walking distance from the garage elevator to the mall core area being the highest score. The system determines the parking lot as a target parking lot, reserves a 'charging parking space qualification' for the target parking lot, and pushes reservation information and a navigation route to the APP of the mr. Wang.

In-field situation awareness and prediction (corresponding S3)

While driving forward, the sensor network unit of the target parking lot continuously works, wherein a video parking space camera recognizes the occupied state of each parking space, a wide-angle camera analyzes the traffic flow speed of each main road, and a charging pile manager reports the real-time state of each pile. These real-time multidimensional status data (parking space status, traffic flow, charging pile status) are aggregated to a local edge server. And uploading the processed data to the cloud end in real time by the edge server. And a prediction analysis unit in the cloud data fusion and situation prediction module starts to work. The method calls historical data of the parking lot for the past month (such as parking space occupation rules of various areas in different time intervals every day), and runs a space-time fusion prediction model by combining the real-time data just uploaded. The model abstracts the parking lot structure into a topological graph, utilizes a Graph Neural Network (GNN) to analyze spatial relationship (such as B area congestion will be transmitted to C area), and simultaneously uses an LSTM time sequence prediction model to analyze time rules. After a few seconds, the method generates dynamic situation information in the field of 10 minutes in the future, wherein one part is a parking space occupation probability thermodynamic diagram which shows which parking spaces are likely to be occupied soon, and the other part is a passage transit time prediction diagram which shows which paths are about to be jammed.

Dynamic parking space allocation and real-time path planning (corresponding to S4 and S5)

Mr. King's car enters the entrance of the parking lot. The system immediately triggers the dynamic parking space allocation module. The decision unit firstly screens out large-size parking spaces which can be parked in the SUV from all idle charging parking spaces to form a candidate parking space list. Then, the comprehensive evaluation value is calculated for each candidate parking space, the estimated passing time for reaching each parking space is calculated according to the real-time position and the estimated passing time in situation information of the Mr. king, the walking distance from each parking space to a mall elevator opening is calculated according to a parking lot map, the parking space occupation probability is referred to, the parking spaces which are more likely to be kept idle for a period of time in the future are selected, and finally, the parking space attribute (quick charging pile) is completely matched with the preference of Mr. king for 'needing charging piles'. The system selects one parking space with the highest comprehensive score as the optimal parking space through weighted calculation, and locks the state of the optimal parking space immediately. The path computation element of the real-time path planning module is then started. The method takes the current position of the vehicle as a starting point and the optimal parking space as an end point, and plans a path on a digital map of a parking lot. The distance is not only seen during planning, but also the actual length of a road section, the current passing time obtained according to real-time camera data, the predicted congestion influence obtained according to situation information and the increased passing difficulty coefficient for avoiding some narrow curves for SUV vehicle types are comprehensively considered. Finally, it plans a driving route with the lowest current comprehensive passing cost.

Guiding and dynamic re-planning (corresponding S6)

The optimal parking space (such as 'D area charging position C12') and the planned path are immediately issued to the APP of Mr. king. The APP interface displays a dynamic map, and the user is guided to run by using a blue highlight line, and simultaneously a voice prompt of 'please go straight and turn left after 50 meters'. Meanwhile, a guide screen in the parking lot and an indicator light on a C12 parking space in the D area are lighted to indicate directions. During the boot process, the anomaly monitoring unit of the condition monitoring and rescheduling module is always operating. It tracks the position of mr. King vehicles through bluetooth beacons and monitors traffic flow in front of the planned path. Suddenly, due to a transient fault of a vehicle in front, the congestion index of an intersection on the planning path of Mr. king rises and exceeds a threshold value. The rescheduling triggering unit is triggered immediately. The system takes the latest position of the mr. King vehicle as a starting point, combines the latest in-field situation, instantaneously re-executes once parking space allocation and path planning, allocates an idle charging parking space (C15) with the same condition nearby for the mr. King vehicle, and plans a new path bypassing congestion. The new boot instructions are immediately updated to mr. Wang APP.

And (5) the mr. King successfully parks the vehicle into the newly allocated parking space C15 after the completion and charging (corresponding to S7) of the vehicle. The video parking space camera above the parking space recognizes that the vehicle is parked, and the system confirms the parking state. After shopping of the Mr. king is finished, clicking one-key off-site on the APP. The system automatically calculates the parking time and the charging cost, and the deduction is completed through the non-inductive payment. The barrier automatically lifts up, and at the same time, the state of the parking space C15 is updated to be idle, and the next user is waited for service.

Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered by the scope of the claims of the present invention.

Claims

1. A collaboratively dynamically guided intelligent parking garage management method, characterized by comprising the following steps:

S1. Receive a parking request from the user's terminal, and based on the destination information in the request, obtain information on multiple candidate parking lots in the vicinity;

S2. Based on user preference information and the information of the candidate parking lots, determine and reserve a target parking lot for the user;

S3. Obtain real-time multi-dimensional status data within the target parking lot, and perform fusion analysis based on the data and historical data to generate and continuously update dynamic status information within the parking lot that characterizes future parking space occupancy and traffic conditions;

S4. In response to a vehicle entering the target parking lot, the optimal parking space is dynamically allocated to the vehicle based on multi-objective decision-making, taking into account the dynamic situation information in the parking lot, vehicle attributes, and user preferences.

S5. Based on the dynamic situation information in the field and vehicle attributes, dynamically plan the driving path for the vehicle from its current position to the optimal parking space;

S6. Send the optimal parking space information and the driving route to the user terminal for guidance, and dynamically adjust the guidance strategy according to environmental changes during the guidance process;

S7. Update parking space status and complete billing operation.

2. The intelligent parking garage management method with collaborative dynamic guidance according to claim 1, characterized in that: in step S3, the fusion analysis is achieved by modeling the spatial topology of the parking lot to capture spatial dependencies, and by combining historical and real-time time series data analysis to capture temporal patterns.

3. The intelligent parking garage management method with collaborative dynamic guidance according to claim 1, wherein step S4, which dynamically allocates the optimal parking space to the vehicle based on multi-objective decision-making, includes: selecting candidate parking spaces that meet the physical constraints of the vehicle from the currently available parking spaces;

A comprehensive evaluation value is calculated for each candidate parking space by calculating a weighted average. The comprehensive evaluation value includes at least the estimated travel time based on real-time and predicted traffic flow, the walking distance to the user's preferred facility, the future vacancy probability of the parking space based on the dynamic situation information in the site, and the matching degree between the parking space attributes and the user's preferences.

The parking space with the highest overall evaluation value is assigned as the optimal parking space.

4. The intelligent parking garage management method with collaborative dynamic guidance according to claim 1, characterized in that, in step S5, when dynamically planning the driving path for the vehicle, the path cost evaluation comprehensively considers the static physical length of the road segment, the travel time cost based on real-time sensor data, the predicted congestion impact based on the dynamic situation information in the parking garage, and the travel difficulty coefficient based on vehicle attributes.

5. The intelligent parking garage management method with collaborative dynamic guidance according to claim 1, characterized in that, in step S6, dynamically adjusting the guidance strategy according to environmental changes includes: continuously monitoring the traffic flow and target parking space status during vehicle travel; if the congestion level of the planned path exceeds a specified threshold or the parking space status is abnormal, a replanning process is triggered, taking the current position of the vehicle as a new starting point, and re-executing parking space allocation and path planning in combination with the latest environmental data.

6. A collaborative dynamic guidance intelligent parking garage management system for implementing the method of any one of claims 1 to 5, characterized in that it comprises:

The request processing and recommendation module, deployed in the cloud, is used to execute steps S1 and S2;

The in-park sensing module is deployed in a distributed manner inside the parking lot to execute step S3;

The data fusion and situation prediction module, deployed in the cloud, is used to execute step S3;

The dynamic parking space allocation module, deployed in the cloud, is used to execute step S4;

The real-time path planning module, deployed in the cloud, is used to execute step S5;

The status monitoring and replanning module, deployed in the cloud, is used to execute step S6;

The billing and status update module, deployed in the cloud, is used to execute step S7;

The request processing and recommendation module, the on-site perception module, the data fusion and situation prediction module, the dynamic parking space allocation module, the real-time route planning module, the status monitoring and replanning module, and the billing and status update module interact with each other and transmit instructions through the network to collaboratively complete the dynamic guidance and management of the entire process from receiving user requests to completing billing.

7. The collaborative dynamic guidance intelligent parking garage management system according to claim 6, characterized in that: the request processing and recommendation module includes a user interface unit and a recommendation engine unit;

The user interface unit is used to interact with the user terminal to receive structured parking request data packets and transmit the data packets to the recommendation engine unit;

The recommendation engine unit is used to calculate and determine the target parking lot based on the destination information, user preferences and multi-source external data in the data packet, and send the reservation result to the user terminal through the user interface unit;

The data fusion and situation prediction module includes a sensor network unit and a predictive analysis unit;

The sensor network units are distributed and deployed in the parking lot to collect multi-dimensional status data in real time;

The predictive analysis unit is used to receive and fuse real-time data and historical data uploaded by the sensor network unit, and generate dynamic situation information in the field through a spatiotemporal fusion predictive model.

The dynamic parking space allocation module and the real-time route planning module include a decision-making unit and a route calculation unit;

The decision-making unit is used to receive the dynamic situation information in the field, vehicle attributes and user preferences, and output the optimal parking space allocation result based on the multi-objective decision-making model.

The path calculation unit is used to receive the optimal parking space information, the real-time location of the vehicle and the dynamic situation information in the field, and calculate the driving path based on the dynamic cost map.

The status monitoring and replanning module includes an anomaly detection unit and a replanning trigger unit;

The anomaly detection unit is used to continuously receive feedback from the sensor network and vehicle positioning data to monitor traffic flow and parking space status.

The replanning triggering unit is used to trigger the decision unit and the path calculation unit to re-execute the calculation when the anomaly detection unit determines that the replanning conditions are met.

8. The collaborative dynamic guidance intelligent parking garage management system according to claim 7, wherein the sensor network unit in the in-park sensing module comprises:

The parking space status sensing subunit is deployed in each parking space to detect the occupancy status of the parking space and perform license plate recognition;

Traffic flow sensing subunits are deployed on key routes to collect data on traffic volume, vehicle speed, and queue length.

The environment and event perception subunit is used to collect facility status and safety event data;

The vehicle positioning beacon subunit is deployed on-site to provide real-time positioning data for vehicles.

9. The collaborative dynamic guidance intelligent parking garage management system according to claim 7, characterized in that the real-time path planning module plans the driving path in the following manner:

A dynamic cost map is constructed, in which the cost of each road segment integrates static physical length, travel time based on real-time sensor data, predicted congestion impact based on the dynamic situation information in the field, and travel difficulty coefficient based on vehicle attributes.

The D*Lite algorithm is used to search for the optimal path from the vehicle's current position to the optimal parking space on the dynamic cost map.