Scenic area large passenger flow-oriented surrounding road network dynamic collaborative management and control and induction method and system

Abstract

The invention provides a dynamic collaborative control and induction method and system for a surrounding road network facing large passenger flow in a scenic area, belongs to the field of intelligent traffic management, and aims to solve the problem of congestion of the surrounding road network caused by large passenger flow in the scenic area. According to the method, through multi-source data acquisition and fusion, traffic flow, parking and scenic spot passenger flow data are fused through Kalman filtering, a road network congestion index is predicted by using an ASTGCN model, and parking lot saturation is predicted by using an LSTM model. A dynamic path induction, self-adaptive signal control and connection resource scheduling collaborative strategy is generated by taking the maximum road network passing efficiency, the balanced parking lot utilization rate and the minimum tourist waiting time as targets, and the strategy is executed in a multi-channel mode and closed-loop feedback optimization is carried out. The system comprises a data sensing layer, a data processing and decision-making layer, an application execution layer and a network transmission layer. The method realizes active and accurate cooperative treatment, improves the traffic efficiency and tourist experience, and is suitable for large-passenger-flow scenes such as festivals and holidays in scenic spots.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent traffic management systems for scenic spots, and in particular to a method and system for dynamic collaborative management and guidance of surrounding road networks for large passenger flows in scenic spots. BACKGROUND

[0002] With the sustained growth of national tourism demand, famous tourist attractions are under tremendous passenger flow pressure during certain periods. A large number of self-driving vehicles arrive in a short period of time, which can easily lead to serious traffic congestion on the surrounding road network of the scenic spot, difficulty in finding a parking space, and chaotic traffic order. This not only greatly reduces the experience and satisfaction of tourists, but also poses a serious challenge to the operation and management of the scenic spot and the daily life of the surrounding residents.

[0003] Existing traffic management systems are mostly designed for urban daily commuting scenarios, and their technical architecture and functional logic are difficult to adapt to the special traffic demand caused by large passenger flows in scenic spots. Direct application has three major core defects, which seriously restrict the control effect.

[0004] First, the response mode is passive and lagging, and lacks proactive intervention capabilities. Existing systems rely on historical traffic flow data or real-time detection data for regulation, and do not include key prior information such as scenic spot ticket reservation volume and time period distribution in the decision-making system. For example, when the number of reserved visitors to the scenic spot increases during holidays, the system cannot predict the peak of self-driving vehicle arrivals in the next 1-2 hours based on reservation data, and can only respond passively by adjusting signal timing or issuing traffic information after traffic congestion has formed. This post-control mode not only makes it difficult to quickly alleviate congestion, but also increases the travel time of tourists, exacerbates travel anxiety, and reduces the travel experience.

[0005] Second, the system lacks coordination, and the subsystem decisions are disjointed or even conflicting. Traffic signal control, path guidance, parking management, and transfer scheduling core function modules are mostly operated independently, and lack a unified data interaction and decision coordination mechanism. For example, the optimal path recommended by navigation software based on real-time traffic conditions may not be synchronized with real-time saturation data for parking lots, which can guide a large number of vehicles to the nearly saturated scenic spot internal parking lot, causing vehicles to queue at the parking lot entrance road, and even causing local paralysis of the surrounding road network. At the same time, the signal control system is not linked with the transfer scheduling system, and cannot allocate priority travel time for transfer vehicles, resulting in low transfer efficiency and further increasing the waiting time of tourists.

[0006] Third, the scene adaptability is insufficient, and the general algorithm is difficult to deal with the traffic characteristics of the scenic spot. The passenger flow of the scenic spot has the characteristics of significant tidal nature and single destination, such as the morning 9-11 o'clock is the concentrated influx period of self-driving vehicles, and the afternoon 16-18 o'clock is the concentrated departure period, and all vehicles are ultimately directed to the scenic spot or the departure place. However, the existing general traffic algorithm is designed for urban multi-destination and multi-period balanced traffic, and cannot accurately capture the pulse fluctuation law of the scenic spot traffic. For example, the signal timing scheme of the algorithm in the morning peak period cannot match the explosive growth of traffic flow on the road network around the scenic spot in a short time, which easily leads to the continuous increase of the queue length at the intersection, and it is difficult to fundamentally solve the periodic congestion problem of the scenic spot, and also causes serious interference to the daily travel of the surrounding residents.

[0007] Therefore, there is an urgent need in the art for an intelligent solution that can deeply integrate scenic operation data and integrally and cooperatively manage and induce various elements of the surrounding road network. SUMMARY

[0008] The primary purpose of the present application is to overcome the shortcomings of the prior art and provide a surrounding road network dynamic cooperative management and induction method and system for large passenger flow in scenic spots, so as to realize active, accurate and cooperative management of the traffic around the scenic spot, effectively alleviate congestion and improve the travel experience of tourists. The present application provides a surrounding road network dynamic cooperative management and induction method for large passenger flow in scenic spots, which comprises the following steps: S1. Multi-source data acquisition and fusion processing (1) Traffic flow data acquisition: The traffic flow, speed, occupancy rate and queue length data of the key road sections are collected by microwave detectors, geomagnetic sensors, video vehicle detectors, etc., and the sampling frequency is 1-5 minutes / time.

[0009] (2) Parking data acquisition: The real-time empty parking space data of each parking lot (including the scenic internal parking lot, the surrounding public parking lot and the commercial supporting parking lot) are obtained through the intelligent management system of the parking lot, including the total number of parking spaces, the number of remaining parking spaces and the entry rate.

[0010] (3) Scenic passenger flow data acquisition: The number of reserved tourists and the time period distribution data are obtained through the API interface of the scenic ticket system, and the real-time number of people in the park is obtained through the gate counting system.

[0011] (4) Data fusion processing: The data fusion algorithm based on Kalman filtering is adopted to perform time and space alignment and consistency processing on the multi-source heterogeneous data, and a unified traffic state data set is generated.

[0012] S2. Traffic state prediction and congestion warning Based on the fused historical and real-time data in S1, machine learning algorithms are used to predict the congestion index of key nodes in the road network and the saturation of each parking lot in a specific period in the future. The prediction uses an attention mechanism-based spatio-temporal graph convolution network (ASTGCN) model for short-term traffic flow prediction and a long short-term memory network (LSTM) model for parking lot saturation prediction.

[0013] (1) Short-term traffic flow prediction: an attention mechanism-based spatio-temporal graph convolution network (ASTGCN) model is used to predict the traffic state of the road network in the next 30 minutes, 1 hour, and 2 hours, with historical traffic flow data, real-time traffic flow data, and scenic reservation data as input; (2) Parking lot saturation prediction: a parking lot saturation prediction model based on a long short-term memory network (LSTM) is established, with historical parking data, real-time parking data, and scenic reservation data as input, and the predicted value of parking lot saturation in the future time period as output; S3. Multi-objective collaborative decision-making The multi-objective optimization function is constructed as follows: ; The represents the total travel time of the road network, representing the total time consumed by all vehicles in the road network in a certain time period, and the expression is: ; where a is the road segment index, is the traffic flow (vehicles / hour) on the road segment t in the time window a , is the travel time function (hours) of road segment a, calculated using the Federal Highway Administration function (BPR function): ; where is the free flow time (hours), is the road capacity (vehicles / hour), and are model parameters (both are dimensionless constants, usually , ); The represents the variance of parking lot utilization, and the expression is: ; where is the predicted utilization rate (%) of parking lot j at time t , Average utilization of all parking lots (dimensionless, %) The utilization formula of the parking lot is: ; Wherein, is the time t The number of occupied parking spaces (vehicles) of the parking lot j , the total number of parking spaces (vehicles) of the parking lot j , the 100 times of the ratio, in percentage (%), as a numerical value in subsequent formulas; The represents the total delay time of users (person-hours), which represents the total additional time consumed by all tourists, including travel time and transfer waiting time, and the expression is: ; Wherein, is the decision variable (vehicles), is the travel time (hours), is the number of people in line at the transfer point, is the transfer vehicle service rate (person / hour); , , , is a dynamic weight coefficient, which is adjusted according to the real-time management effect through reinforcement learning algorithm.

[0014] The collaborative control strategy includes three sub-strategies: (1) Dynamic path induction strategy: based on the prediction results, the optimal parking lot and path for vehicles from different directions are calculated through Dijkstra algorithm combined with constraint optimization, considering factors including: real-time traffic, parking lot saturation, path length, and estimated travel time.

[0015] (2) Adaptive signal control strategy: using reinforcement learning method, the signal timing scheme of key intersections is dynamically adjusted to improve the efficiency of the road network, especially for the main path leading to the parking lot with capacity.

[0016] (3) Transfer resource scheduling strategy: according to the prediction of passenger flow induced to the remote parking lot, the number of transfer vehicles and the interval between vehicles are calculated using a queuing model to generate a transfer vehicle scheduling scheme.

[0017] S4. Multi-channel strategy implementation and feedback optimization The induction strategy is published through variable information signs, navigation APP, and scenic spot applets, the signal control strategy is sent to the intersection signal machine, the implementation effect of the strategy is monitored, and the feedback data is input into S2 to realize closed-loop dynamic optimization.

[0018] (1) Publish dynamic induction information through the following channels: a. Roadside variable message signs (VMS): Display recommended parking lots, remaining parking spaces, and estimated travel time; b. Navigation platform interface: Integrate with navigation platforms such as Gaode and Baidu through RESTful API to push recommended routes; c. Official platform of scenic spots: Push personalized driving suggestions to pre-booked tourists through mini programs and APPs; (2) Signal control execution: Downlink the optimized signal timing scheme to the intersection signal controller through the standard NTCIP protocol; (3) Connection scheduling execution: Send scheduling instructions to connection vehicle drivers through the scheduling system; (4) Effect monitoring and feedback: Real-time monitoring of traffic state changes after strategy execution, comparing the deviation between predicted and actual values, and dynamically adjusting model parameters using gradient descent-based algorithms to achieve closed-loop optimization.

[0019] The present application provides a system for implementing the above method, comprising the following levels and modules: 1. Data perception layer Including traffic flow detectors deployed at the roadside, parking lot space monitoring devices, and data interface modules connected to scenic spot reservation systems: Roadside traffic detection unit: including microwave traffic detectors, video vehicle detectors, geomagnetic sensors, etc. Parking monitoring unit: including parking lot space detectors, license plate recognition systems, and parking management system data interfaces; Scenic spot passenger flow monitoring unit: including ticketing system data interfaces, gate counting systems, and face recognition systems; Vehicle positioning unit: including GPS / Beidou positioning modules and vehicle-mounted OBU.

[0020] 2. Data processing and decision-making layer Used for cleaning, standardizing and fusing multi-source heterogeneous data; Data fusion processing module: realizes cleaning, alignment and fusion of multi-source data; Traffic state prediction module: built-in machine learning prediction algorithm models such as ASTGCN and LSTM for predicting congestion and parking lot saturation; Collaborative optimization decision module: used for running the multi-objective optimization model to generate collaborative control strategies and realize multi-objective optimization algorithm and reinforcement learning decision-making functions; System management module: provides user permission management, system configuration, log recording and other functions.

[0021] 3. Application execution layer The dynamic induction information publishing unit, the signal control executing unit, and the information service platform interface unit are included. The information publishing module: manages the information publishing channels such as VMS, navigation platform, mobile terminal, etc. The signal control module: realizes communication and control with the intersection signal controller. The connection scheduling module: realizes the management and scheduling of the connection vehicles. The man-machine interaction interface: provides a visual monitoring and command platform for administrators.

[0022] 4. Network transmission layer Used for connecting each layer to realize safe and reliable transmission of data. Wired communication network: adopts industrial Ethernet to connect fixed equipment. Wireless communication network: adopts 5G / V2X technology to realize communication between mobile equipment and the system platform. Data security transmission: adopts the TLS encryption protocol to guarantee the safety of data transmission.

[0023] Compared with the prior art, the beneficial effects of the present application include: (1) Prospective and initiative: from "passive response" to "source prediction". The existing scheme is mostly decongestion after congestion occurs, which has a lag. The present scheme can predict the peak of traffic flow in advance by fusing scenic spot reservation data, real-time traffic data and historical rules to build a full-link prediction model, and can push off-peak routes in advance through the navigation APP to realize "controlling before congestion occurs", avoiding resource waste and tourist retention.

[0024] (2) Deep synergy: breaking down barriers to build an "integrated closed loop". Current subsystems such as signal control, path induction, etc. are often independently operated and decision conflicts. The present scheme relies on "traffic brain" to integrate data of each subsystem in real time and output integrated instructions such as synchronous induction of traffic flow, scheduling of connection vehicles, and optimization of signal timing to form a "induction-control-parking-connection" closed loop, which has been measured to improve overall efficiency and reduce conflict rate.

[0025] (3) High adaptability to scene: matching the "tide" and "explosive" passenger flow of scenic spots. General algorithms are difficult to adapt to the characteristics of scenic spot passenger flow. The present scheme greatly shortens the prediction period, sets up a response mechanism for sudden passenger flow, and customizes the model according to the terrain and passenger flow distribution of the scenic spot, which can still guarantee the traffic when the passenger flow is overloaded.

[0026] (4) Significantly improve the experience: from "functional control" to "cultural service". Effectively alleviate congestion, reduce the blind round-trip time and anxiety of tourists looking for parking spaces, and improve the overall tour satisfaction through seamless connection of connection services. BRIEF DESCRIPTION OF DRAWINGS

[0027] The application will be further described below with reference to the drawings and examples.

[0028] Figure 1 A schematic diagram of the system architecture of the application. DETAILED DESCRIPTION

[0029] In order to better further understand the application, the following examples are provided, but the following examples do not constitute limitations on the content and protection scope of the application, and any product that is the same or similar to the application obtained by anyone under the inspiration of the application or by combining the application with other prior art features falls within the protection scope of the application.

[0030] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application; the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion.

[0031] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive or alternative embodiments. It is expressly understood that the embodiments described herein are merely examples from a whole class of comparable embodiments which those skilled in the art will readily appreciate. It is also expressly understood that the terms "comprising" and "having," and variations thereof, are intended to cover a non-exclusive inclusion.

[0032] In the description of the embodiments of the application, the term "and / or" only means a relationship between associated objects, and means that there can be three relationships, for example, A and / or B, which means that there are three cases of A alone, A and B together, and B alone.

[0033] In the description of the embodiments of the application, the term "at least one" means one or more than two (including two).

[0034] If not specifically stated, all steps of the application can be performed in sequence or randomly, and are preferably performed in sequence. For example, the method comprises steps (a) and (b), which means that the method can comprise steps (a) and (b) performed in sequence, or steps (a) and (b) performed in sequence. For example, the method can also comprise step (c), which means that step (c) can be added to the method in any order, for example, the method can comprise steps (a), (b) and (c), or steps (a), (c) and (b), or steps (c), (a) and (b), etc.

[0035] EMBODIMENT The system architecture of a large-flow-oriented scenic area surrounding road network dynamic coordination control and guidance method and system is as shown in Figure 1 The specific implementation is as follows: Taking a 5A-level landscape scenic area as an example, there are three main access channels (east, west and north entrance channels) around the core scenic area, two core parking lots (P1 and P2) and one standby parking lot (P3), wherein P1 is adjacent to the east entrance (2000 parking spaces), P2 is located at the west entrance (1500 parking spaces), and P3 is 1.5 kilometers away from the north entrance (1000 parking spaces). On the first day of the National Day holiday, the number of scenic area reservations reached the maximum carrying capacity of 80,000 people, and it is estimated that 9:00-11:00 is the peak arrival period of traffic flow, and there is a risk of entrance channel congestion and core parking lot saturation.

[0036] S1. Real-time collection of multi-source data (7:00 start) The system synchronously acquires the following real-time data through a preset data interface: (1) Traffic flow data: 1200 vehicles / hour (50% higher than the daily morning peak) for the east entrance channel, 800 vehicles / hour for the west entrance channel, and 300 vehicles / hour for the north entrance channel; (2) Parking lot data: P1 parking lot saturation 65% (700 remaining parking spaces), P2 parking lot saturation 40% (900 remaining parking spaces), and P3 parking lot saturation 20% (800 remaining parking spaces); (3) Reservation data: In the next 1 hour (7:00-8:00), the proportion of visitors entering from the east entrance is 62%, the proportion of visitors entering from the west entrance is 28%, and the proportion of visitors entering from the north entrance is 10%.

[0037] S2. Double model cooperative prediction (7:10 completed) (1) Road congestion prediction: ASTGCN (Attention Spatio-Temporal Graph Convolutional Network) model is used, combined with real-time traffic flow and historical National Day first day traffic flow rules (east entrance 9:00-10:00 congestion index peak value reaches 0.92), to predict that 1.5 hours later (8:30-9:30) the east entrance channel congestion index will break through 0.8 (serious congestion threshold), and the west entrance channel congestion index will rise to 0.6 (mild congestion); (2) Parking lot saturation prediction: based on the LSTM (Long Short-Term Memory Network) model, input P1 real-time saturation, east entrance traffic growth rate, predict that P1 will reach 100% saturation in 1 hour (8:00), and no empty parking space will be released in the next 2 hours.

[0038] S3. Multi-objective optimization strategy generation (7:15 output scheme) The collaborative decision-making module takes "minimize total delay of road network + balance parking lot utilization" as the goal (i.e. formula α=0.4, β=0.3, γ=0.3), and after 100 iterations of calculation, the optimal management strategy is generated: (1) Path induction strategy: for vehicles coming from the east entrance direction after 7:15 and not binding P1 reserved parking spaces, push the "east entrance congestion, suggest detouring west entrance → stop P2 parking lot, enjoy free transfer to core area" induction information through navigation APP (such as Gaode, Baidu Map), and display the same guidance on the VMS (variable message sign) at the east entrance 3 kilometers away; (2) Signal control strategy: adjust the signal timing of the 5 key intersections along the west entrance channel (such as west entrance highway exit-scenic west road intersection, scenic west road-P2 parking lot entrance intersection), extend the green light duration of westbound traffic from 30 seconds to 45 seconds, generate green wave band, and ensure that the induced vehicle passing efficiency is improved by 30%; (3) Transfer scheduling strategy: dispatch 6 19-seater transfer vehicles from the scenic transfer station (located near the west entrance) to arrive at P2 parking lot before 7:30, operate at a frequency of 1 class every 10 minutes (daily 1 class every 20 minutes), and the route is "P2 parking lot → west entrance visitor center → core scenic cableway station"; (4) Backup parking lot activation: push the "P3 parking lot has sufficient space, can enjoy 10 yuan / day parking fee (daily 20 yuan / day)" to the north entrance direction, and dispatch 2 transfer vehicles to P3 standby.

[0039] S4. Strategy implementation and effect feedback (7:30-12:00) (1) Implementation landing: from 7:30, the coverage rate of navigation induction information reaches 90%, the signal timing of west entrance channel is adjusted, P2 and P3 transfer vehicles are in place according to the plan, at 8:00, P1 parking lot is saturated as scheduled, the system automatically closes the navigation induction entrance of P1, and only keeps the access permission of reserved parking space vehicles; (2) Effect monitoring: by 12:00, the highest congestion index of east entrance channel is reduced to 0.65 (18.75% lower than the predicted value), the congestion index of west entrance channel is stable at 0.5 (not reaching the light congestion threshold), the saturation of P2 parking lot is improved to 88% (remaining 180 parking spaces), and the saturation of P3 parking lot is improved to 65% (remaining 350 parking spaces), the utilization rate difference of the three parking lots is narrowed to 23% (the original predicted difference is 80%); (3) User feedback: the average time for tourists to find parking spaces is shortened from 22 minutes last year to 7 minutes, the average waiting time for transfer vehicles is controlled within 8 minutes, and the real-time evaluation score of "traffic convenience" in the scenic area reaches 4.7 / 5.0 (3.2 / 5.0 last year).

[0040] This embodiment successfully avoided severe congestion at the core entrance passage on the first day of the National Day holiday through closed-loop management of data collection, prediction, decision-making, and execution, achieving balanced utilization of parking resources and significantly improving the tourist transportation experience. This verifies the effectiveness and practicality of the technical solution of this invention in scenarios with large passenger flows.

Claims

1. A method for dynamic collaborative management and guidance of the surrounding road network for large visitor flows in scenic areas, characterized in that, Includes the following steps: S1. Collect real-time traffic flow data of the road network around the scenic area, real-time parking space data of each parking lot, reservation flow data of the scenic area reservation system and real-time number of people in the park, and use a data fusion algorithm based on Kalman filtering for fusion processing; S2. Based on the fused data, predict the congestion index of key road network nodes and the saturation of each parking lot in a specific future period; S3. With the objectives of balancing road network load, maximizing parking lot utilization, and minimizing average visitor waiting time, a multi-objective optimization function is constructed to generate a collaborative control strategy. The multi-objective optimization function is as follows: ； in, Total travel time on the road network represents the total time spent by all vehicles on the road network within a specific time period. Let Variance be the parking lot utilization rate. Total delay time for users , , These are dynamic weighting coefficients; S4. Execute the aforementioned collaborative control strategy and monitor the execution effect for dynamic feedback optimization.

2. The method according to claim 1, characterized in that, In step S1, the Kalman filter data fusion algorithm predicts the current state of multi-source data through the state equation and corrects the predicted value by combining the observation values ​​of multi-source data with the observation equation, thereby realizing the spatiotemporal alignment and consistency processing of multi-source heterogeneous data and eliminating data redundancy and conflicts.

3. The method according to claim 1, characterized in that, In step S2, the prediction is performed using a spatiotemporal graph convolutional network model based on an attention mechanism for short-term traffic flow prediction, and a long short-term memory network model for parking lot saturation prediction.

4. The method according to claim 1, characterized in that, In step S3, the total travel time of the road network The calculation formula is: ； Where 'a' is the road segment index. Traffic flow (vehicles / hour) on road segment a within time window t. The travel time (in hours) for road segment a is calculated using the Federal Highway Administration (BPR) function: ； in, Free-flow time (hours) This refers to the road segment's traffic capacity (vehicles per hour). and The model parameters (all are dimensionless constants, usually taken as...) , ); The The variance of parking lot utilization is expressed as: ； in, For parking lot j In time t The predicted utilization rate (%). Average utilization rate (dimensionless, %) The utilization rate formula for the parking lot is: ； in, For time t PARKING LOT j The number of parking spaces (vehicles) already occupied. Let j be the total number of parking spaces (vehicles). It is 100 times the ratio, expressed as a percentage (%), and will be used as a numerical value in subsequent formulas. The The total delay time (person-hours) represents the total extra time consumed by all tourists, including travel time and waiting time for connections. The expression is: ； in, For decision variables (vehicles). Travel time (in hours) The number of people queuing at the shuttle point. Shuttle bus service rate (people / hour).

5. The method according to claim 1, characterized in that, In step S3, the collaborative control strategy includes a dynamic path guidance strategy, an adaptive signal control strategy, and a connection resource scheduling strategy.

6. The method according to claim 4, characterized in that, The dynamic route guidance strategy specifically includes: allocating optimal parking lots and travel routes to vehicles based on their origin direction, real-time traffic conditions, and parking lot saturation, and publishing this information through variable message signs, navigation software, and the scenic area's official platform.

7. A system for implementing the method as described in any one of claims 1-5, characterized in that, include: The data perception layer is used to collect multi-source traffic and scenic area operation data; The data perception layer includes a roadside traffic detection unit, a parking monitoring unit, a scenic area visitor flow monitoring unit, and a vehicle positioning unit. The roadside traffic detection unit includes a microwave traffic detector, a video vehicle detector, and a geomagnetic sensor. The parking monitoring unit includes a parking space detector, a license plate recognition system, and a parking management system data interface. The scenic area visitor flow monitoring unit includes a ticketing system data interface, a gate counting system, and a facial recognition system. The vehicle positioning unit includes a GPS / BeiDou positioning module and an on-board unit (OBU). The data processing and decision-making layer is used for data fusion, traffic forecasting, and collaborative decision-making. The data processing and decision-making layer includes a data fusion processing module, a traffic state prediction module, a collaborative optimization decision-making module, and a system management module. The data fusion processing module performs cleaning, alignment, and fusion of multi-source data, and generates a unified traffic state dataset using a Kalman filter-based algorithm. The traffic state prediction module incorporates an attention-based spatiotemporal graph convolutional network (ASTGCN) model and a long short-term memory network (LSTM) model to predict road network congestion index and parking lot saturation. The collaborative optimization decision-making module runs a multi-objective optimization model and combines reinforcement learning and queuing theory algorithms to generate collaborative control strategies. The system management module provides user permission management, system configuration, and log recording functions. The application execution layer is used to execute collaborative control strategies and publish information. The application execution layer includes an information publishing module, a signal control module, a shuttle dispatch module, and a human-machine interface. The information publishing module includes information publishing channels such as managing roadside variable message signs (VMS), navigation platforms, and scenic area mini-programs / apps. The signal control module realizes communication and control with the intersection signal controller. The shuttle dispatch module realizes the management of shuttle vehicles, the issuance of dispatch instructions, and the monitoring of shuttle status. The human-machine interface provides administrators with an operation platform for visual monitoring of traffic status, strategy adjustment, and system operation and maintenance. The network communication layer is used to connect the above layers and realize data transmission; The network communication layer includes a wired communication network, a wireless communication network, and a data security transmission unit. The wired communication network uses industrial Ethernet to connect fixed equipment such as roadside detectors and signal controllers. The wireless communication network uses 5G / V2X technology to enable communication between mobile devices such as vehicle-mounted OBUs and shuttle bus dispatch terminals and the system platform. The data security transmission unit uses the TLS encryption protocol to ensure the security of data transmission and prevent data leakage and tampering.

8. The system according to claim 7, characterized in that, The parking space detectors in the data perception layer use ultrasonic sensors or video recognition sensors to collect parking space occupancy status in real time, with a sampling frequency of 10-30 seconds per sampling to ensure the real-time availability of parking space data.

9. The system according to claim 7, characterized in that, The traffic state prediction module of the data processing and decision-making layer has a built-in model update unit that periodically retrains the ASTGCN model and LSTM model based on newly collected historical data to adapt to the long-term changing characteristics of traffic flow in the scenic area.

10. An electronic device and a computer-readable storage medium, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the dynamic collaborative management and guidance method for the surrounding road network for large passenger flows in scenic areas as described in any one of claims 1-6.

Images