Agent-based people flow simulation model construction and simulation method

Abstract

The invention provides an Agent-based people flow simulation model construction and simulation method. The Agent-based people flow simulation model construction and simulation method comprises the steps of constructing a region model, uniformly dividing a scenic spot into a plurality of grids, setting basic attributes, bearing attributes and reachable attributes of the grids to construct grid attributes, and forming the region model of the scenic spot by all the grid attributes; constructing an entity model, and setting basic attributes, sightseeing attributes and state attributes of the tourists to form the entity model of the tourists; constructing a behavior model, and setting a scenic spot retention decision, a scenic spot selection decision, a tourist moving speed and an exit decision of the tourist based on the entity model to form a behavior model of the tourist; and people flow prediction: setting information of a plurality of tourists to generate a plurality of entity models, substituting the entity models into the behavior model and the area model, and independently predicting a moving path of each tourist in the area model so as to predict change conditions of people flows in different scenic spots. According to the invention, people flow change conditions in different scenic spots in a specific time period of the scenic area can be predicted, so that the overall tourist bearing capacity of the scenic area can be predicted in advance.

Description

Technical Field

[0001] The invention relates to the technical field of crowd flow simulation, and in particular to an Agent-based crowd flow simulation model construction and simulation method. Background Art

[0002] With the booming global tourism industry, more and more people are keen to visit various cultural heritage sites to explore the mysteries of history and experience the charm of ancient culture. However, cultural heritage sites often face the dilemma of being overwhelmed by tourists. A large number of tourists gather in a relatively small space with limited carrying capacity, which makes the environment where the cultural relics are located overwhelmed. The unconscious physical contact and frequent trampling of tourists cause direct wear and tear and damage to the physical structure of cultural relics. The stone slabs are dented and broken due to excessive trampling, and the doors, windows and wood carvings of ancient buildings are blurred due to frequent touching. These damage phenomena are common.

[0003] Existing technical means mainly focus on two key aspects: tourist carrying capacity limit and diversion strategy. Tourist carrying capacity limit is to formulate the upper limit of tourists that can be accommodated through scientific measurement, based on multi-dimensional factors such as the spatial scale of cultural heritage sites, the fragility of cultural relics, and the carrying capacity of supporting facilities. When the number of tourists approaches the threshold, ticket sales will be stopped immediately or flow control measures will be taken to prevent overcrowding. The diversion strategy is to use a variety of online and offline channels to push information about the flow of people in different time periods and areas to tourists in advance, and guide tourists to visit during off-peak hours. While ensuring that tourists have a good tour experience, the risk of damage to precious cultural relics is minimized. Summary of the invention

[0004] In view of this, the problem to be solved by the present invention is to provide an Agent-based crowd flow simulation model construction and simulation method.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for constructing a crowd flow simulation model based on an agent, comprising: constructing a regional model of a scenic spot, the scenic spot is evenly divided into a number of grids, and the basic attributes, carrying attributes and reachable attributes of the grids are set to construct grid attributes, and all grid attributes together constitute the regional model of the scenic spot;

[0006] Constructing a physical model of tourists, setting basic attributes, tour attributes and status attributes of tourists to form the physical model of tourists;

[0007] Construct a tourist behavior model, and set the tourist's attraction stay decision, attraction selection decision, tourist movement speed, and exit decision based on the entity model to form the tourist's behavior model;

[0008] Crowd flow prediction: set some tourist information to generate several entity models, bring the entity model into the behavior model and regional model, and predict the movement path of each tourist in the regional model separately to predict the changes in the flow of people in different scenic spots.

[0009] Furthermore, the basic attributes of the grid record the address information and attractiveness of the grid, the carrying capacity attribute of the grid records the maximum number of tourists that the grid can carry and the tourist movement speed corresponding to different tourist loads, and the reachability attribute of the grid records whether the grid can carry tourists.

[0010] Furthermore, the basic attributes of the tourist include the tourist's age information, location information, time of entry into the scenic area, and time of exiting the scenic area. The tourist's sightseeing attributes include the tourist's energy attribute and interest attribute. The energy attribute includes the tourist's energy for visiting the scenic area. The interest attribute includes the interest level of the scenic spot, the attenuation level of the interest level of the scenic spot, the recovery level of the interest level of the scenic spot, the minimum interest value of the scenic spot, and the maximum interest value of the scenic spot. The tourist's status attributes include information on the scenic spots that the tourist has not visited, information on the scenic spots that the tourist has visited, an indication of whether the tourist has moved to the target location, and an indication of whether the tourist has left the scenic area.

[0011] Furthermore, the decision to stay at a scenic spot is based on two conditions: first, when the tourist's interest in the current scenic spot drops below the minimum interest of the scenic spot, the tourist will tend to leave; second, if the tourist's stay at the current scenic spot exceeds a reasonable time, it will also trigger the decision to leave. For the current scenic spot, the formula for updating the interest is:

[0012] SIV i，current,t =(1-IDR i,current )(1-ADT i,current )×(SIV i，current，t-1 -MIV i，current ),

[0013] Among them, SIV i,current,t and SIV i,current,t-1 represents the interest value of the i-th tourist in the current attraction at time t and time t-1, IDR i,current represents the decay rate of the interest value of the i-th tourist in the current attraction, ADT i,current It represents the interest recovery rate of the i-th tourist in the scenic spot he is visiting, MIV i,current It represents the minimum interest value of the i-th tourist in the current attraction.

[0014] Furthermore, the attraction selection decision follows a priority rule, that is, tourists evaluate the attractions that have not been visited based on the current interest value and select the attractions with the highest interest to visit. For the unvisited attractions, the formula for updating the interest is:

[0015] SIV i，j，t =(IRR i，j –1)(SIV i，j，t-1 -MAV i，j ),

[0016] Among them, SIV i,j,t and SIV i,j,t-1 represents the interest value of the i-th tourist for the j-th attraction at time t and time t-1, IRR i,j represents the interest recovery rate of the i-th tourist in the j-th unvisited attraction, MAV i,j Indicates the maximum interest of the i-th tourist in the j-th attraction.

[0017] Furthermore, the formula for the tourist moving speed is:

[0018]

[0019] Among them, SR x,y,t Represents the speed ratio at time t in the grid corresponding to the x and y coordinate positions, SC x,y,t Represents the ratio of the total number of people in the grid corresponding to the x, y coordinate position to the maximum capacity.

[0020] Further, the exit decision includes obtaining the state attribute of the tourist, determining whether the list of unvisited attractions is empty, if yes, leaving the scenic spot, if no, obtaining the moving speed and the list of unvisited attractions, and selecting the most interesting attraction from the list of unvisited attractions for the next visit;

[0021] Determine whether the time to reach the next scenic spot exceeds the closing time of the scenic spot. If yes, leave the scenic spot. If no, determine the current energy value of the tourist through the energy calculation formula. If the current energy value is lower than the threshold, if yes, leave the scenic spot. If no, continue to play.

[0022] Furthermore, the energy calculation formula is:

[0023]

[0024] Among them, E i,t and E i,t-1 represents the energy value of the i-th tourist at time t and time t-1, EDR i represents the energy decay rate of the i-th visitor, E mini Represents the minimum energy of the i-th tourist.

[0025] Furthermore, the crowd flow prediction includes optimizing the regional model and entity model parameters through Optuna, and the Optuna optimization includes judging the optimization result through a loss function; the loss function is:

[0026]

[0027] Among them, N represents the number of scenic spots involved in the data, Y i is the actual visiting time distribution of the i-th scenic spot, is the visiting time distribution obtained by the simulation program under the i-th scenic spot, is the real average visiting time of the whole scenic spot, T is the simulated average visiting time of the whole scenic spot, and λ represents the importance of the total average visiting time.

[0028] A method for simulating crowd flow includes initializing a physical model of tourists and a regional model of scenic spots, predicting whether tourists continue to visit based on exit decisions, if not, determining the best exit, generating a navigation path for the scenic spots and moving, and if yes, determining whether tourists enter the grid for the first time;

[0029] If it is the first time to enter the grid area, the tourist's selected attractions are predicted based on the attraction selection decision, and the navigation path is generated and moved; if it is not the first time to enter the grid area, it is determined whether the tourist has reached the target attraction. If not, move along the existing navigation path. If yes, determine whether the tourist stays to visit through the attraction retention decision;

[0030] If the tourists stay to visit, the moving path of the tourists is predicted according to the moving speed of the tourists, and the navigation path is generated and the tourists move; if the tourists do not stay to visit, the next scenic spot selected by the tourists is predicted according to the scenic spot selection decision, and the navigation path is generated and the tourists move;

[0031] Update the visitor's status and capabilities, and jump to the step of predicting whether the visitor will continue the tour based on the exit decision.

[0032] The advantages and positive effects of the present invention are:

[0033] (1) By constructing regional models, entity models and behavior models respectively, and constructing the entity model of each tourist based on the entity model and tourist information, the entity model of each tourist is brought into the behavior model and regional model, and the visiting route of each tourist in the scenic area is predicted separately. The flow of people in different areas of the scenic area is predicted efficiently and accurately, so as to predict the tourist carrying capacity of the scenic area in advance and reduce the probability of damage to scenic spots in the scenic area due to excessive flow of people.

[0034] (2) By setting the loss function, the parameters of the regional model and the entity model are optimized to improve the accuracy of pedestrian flow prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0036] Figure 1It is a construction process diagram of an Agent-based crowd flow simulation model of the present invention;

[0037] Figure 2 It is an overall flow chart of the Agent-based crowd flow simulation model construction and simulation method of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0040] The present invention provides an agent-based crowd flow simulation model. The agent-based modeling (ABM) simulates the behavior of each agent with autonomous attributes and decision-making capabilities from a micro level and is widely used in the field of crowd flow simulation.

[0041] like Figure 1 As shown, it includes constructing a regional model of the scenic area, dividing the scenic area evenly into a number of grids, setting the basic attributes, carrying attributes and reachable attributes of the grids to construct grid attributes, and all grid attributes together constitute the regional model of the scenic area.

[0042] The regional model converts the complex scenes of the real world into a virtual environment for subsequent crowd flow simulation. The generation process of the regional model includes: obtaining the two-dimensional map of the scenic area and dividing it into several grids, setting the basic attributes, carrying attributes and reachable attributes of each grid, and forming the regional model of the scenic area.

[0043] The formulas for the grid properties are:

[0044] SSM=(S B ,S C ,S R ),

[0045] Where: S B Indicates the basic attributes, S C Represents the carrying capacity attribute, S RRepresents the reachability attribute.

[0046] The basic attributes record the address information and attractiveness of the grid. The address information is the coordinate range of the grid. Whether the tourist enters the grid is determined by whether the tourist enters the coordinate range. The scenic area includes several scenic spots, and each scenic spot is composed of several grids.

[0047] An embodiment of the present application is that the grids accessible to tourists include regular grids and scenic spot grids. Regular grids are grids that tourists pass through when transferring, and scenic spot grids are grids where tourists stay when visiting scenic spots. Each scenic spot is assigned a different attraction value.

[0048] The carrying capacity attribute records the extreme value of the number of tourists that the grid can carry and the tourist movement speed corresponding to different tourist carrying capacities. The more tourists there are in the grid, the slower the movement speed, which can reflect the impact of crowd density on mobility. The carrying capacity of a grid is related to the size of the grid area and the attributes of the area where the grid is located. For example, the tourist carrying capacity of indoor and outdoor grids is different. In actual settings, the tourist carrying capacity is set according to the location of the grid and the level of the cultural relics in the grid.

[0049] The accessibility attribute records whether the grid is open for sightseeing. The accessibility of unopened areas in the scenic area is 0, and the accessibility of open areas in the scenic area is 1. Recording the accessibility attributes of the grid can make each grid correspond to a fixed position in the scenic area, improving the authenticity of the regional model.

[0050] An embodiment of the present application is that the reachable attribute is adjustable data, and the management personnel can adjust the flow of personnel in time according to the reachable attribute of each grid area to reduce the probability of tourist congestion.

[0051] The regional model includes the grid attribute data of all grids. Since the grids are divided according to the real scene, the grid attributes are artificially set based on the actual connectivity of the scenic area, the opening status of scenic spots, and the relationship between grids and scenic spots, ensuring the consistency of the regional model with the actual scene.

[0052] Construct a physical model of tourists, set basic attributes, tour attributes and status attributes of tourists to form the physical model of tourists.

[0053] The solid model formula is:

[0054] TEM=(T BA , T VA , T SA ),

[0055] Among them, T BA represents the basic attributes of tourists, T VA represents the tourist's travel attributes, T SA Indicates the status attribute of the visitor.

[0056] The basic attributes include the age information, location information, time of entering the scenic spot, and time of leaving the scenic spot of the tourists. Before the crowd flow simulation process, the age information of each tourist is set manually. Different ages correspond to different initial energy (energy for visiting the scenic spot). The location information records the location of the tourists in the scenic spot, and the location of the tourists is continuously updated as the simulation process is executed. The time of entering the scenic spot is obtained by sampling the tourist entry curve. The moment when the tourist leaves the scenic spot during the crowd flow simulation is defined as the time when the tourist leaves the scenic spot.

[0057] The expression formula of tour attribute is:

[0058] T VA =(T VAE , T VAI )

[0059] Among them, T VAE Represents energy properties, T VAI Represents interest attributes. Tour attributes include the current energy value and the minimum energy value of tourists. The energy value of tourists can be obtained in advance through experiments or set based on experience. The longer the tour time, the lower the energy value. The minimum energy value can be used to control tourists to exit the scenic spot. Interest attributes can predict the probability of tourists visiting or not visiting scenic spots. Interest attributes include five attributes: scenic spot interest, scenic spot interest attenuation, scenic spot interest recovery, scenic spot minimum interest value, and scenic spot maximum interest value. The above five attributes are artificially set according to the age of tourists and the online popularity of scenic spots.

[0060] The status attribute records the list of attractions that the tourist has not visited, the list of attractions that the tourist has visited, the mark of whether the tourist has moved to the target location, and the mark of whether the tourist has left the scenic area.

[0061] Construct a tourist behavior model, and set the tourist's attraction stay decision, attraction selection decision, tourist movement speed, and exit decision based on the entity model to form the tourist's behavior model.

[0062] The attraction retention decision is used to predict whether tourists will stay at the current attraction. The relevant formula for the attraction retention decision is:

[0063] SIV i，current,t =(1-IDR i,current )(1-ADT i，current )×(SIV i，current,t-1 -MIV i，current ),

[0064] Among them, SIV i,current,t and SIV i,current,t-1 represents the interest value of the i-th tourist in the current attraction at time t and time t-1, IDR i,current represents the decay rate of the interest value of the i-th tourist in the current attraction, ADT i,currentIt represents the interest recovery rate of the i-th tourist in the scenic spot he is visiting, MIV i,current It represents the minimum interest value of the i-th tourist in the current attraction.

[0065] The attraction selection decision is used to predict the next attraction that tourists will visit. The relevant formula for the attraction selection decision is:

[0066] SIV i，j，t =(IRR i，j -1)(SIV i，j，t-1 -MAV i，j ),

[0067] Among them. SIV i,j,t and SIV i,j,t-1 represents the interest value of the i-th tourist for the j-th attraction at time t and time t-1, IRR i,j represents the interest recovery rate of the i-th tourist in the j-th unvisited attraction, MAV i,j Indicates the maximum interest of the i-th tourist in the j-th attraction.

[0068] Tourist moving speed is used to predict the moving speed of tourists in different grids. The formula of tourist moving speed is:

[0069]

[0070] Among them, SR x,y,t Represents the speed ratio at time t in the grid corresponding to the x and y coordinate positions, SC x,y,t It represents the ratio of the total number of people in the grid corresponding to the x and y coordinate positions to the maximum carrying capacity. One embodiment of the present application is: the maximum moving speed of tourists is 4m / s, and the product of the speed ratio and the maximum moving speed is the moving speed of tourists in the grid that can be reached at time t.

[0071] The exit decision is used to predict whether tourists choose to leave the scenic area. The scenic area exit decision includes obtaining the state attributes of the tourist from the tourist's entity model, and judging whether the list of unvisited attractions is empty based on the real-time updated state attributes. If so, it means that the tourist intends to visit all the attractions and will choose to leave the scenic area. If not, it means that there are attractions in the scenic area that the tourist has not visited (the list of unvisited attractions is not empty), and the attraction with the highest interest is selected from the list of unvisited attractions for the next visit.

[0072] The scenic spot exit decision includes obtaining the current grid where the tourist is located, generating a movement path for the tourist to move to the nearest exit based on the current location, and then determining the movement speed based on the density of tourist flow on the path. It is judged whether the time it takes for the tourist to move to the nearest exit exceeds the closing time of the scenic spot. If yes, leave the scenic spot, otherwise continue to play.

[0073] The scenic spot exit decision includes obtaining the total time the tourists have spent in the scenic spot, and determining whether the tourists’ current energy value is lower than the energy threshold through the energy calculation formula. If yes, leave the scenic spot, otherwise, continue to play.

[0074] Energy calculation is used to determine the changes in tourists' physical strength. The energy calculation formula is:

[0075]

[0076] Among them, E i,t and E i,t-1 represents the energy value of the i-th tourist at time t and time t-1, EDR i represents the energy decay rate of the i-th visitor, E mini Represents the minimum energy of the i-th tourist.

[0077] Crowd flow prediction: set some tourist information to generate several entity models, bring the entity model into the behavior model and regional model, and predict the movement path of each tourist in the regional model separately to predict the changes in the flow of people in different scenic spots.

[0078] When predicting the flow of people, set the parameters in the regional model and the entity model, and set some tourist information. The tourist information is brought into the entity model to generate the entity model of each tourist. The entity model is brought into the behavior model and the regional model. The movement path of each tourist in the regional model can be predicted, and the change of the flow of people at each scenic spot in the scenic area can be determined, so as to predict the maximum passenger capacity of the scenic area in advance, to avoid the scenic area receiving too many tourists, which will cause damage to the cultural relics in the scenic area.

[0079] Since the regional model and entity model include a large number of artificially set parameters, there is a large gap between the model output results and the actual results. Optuna's deep learning parameter adjustment method can be used to optimize the parameters and improve the accuracy of the prediction. Optuna parameter adjustment includes obtaining the tour path information of a small sample of real tourists and adjusting the parameters based on the statistics of the tour path information. Since the parameters are adjusted using a small sample of real tourist data, if the tourist sample collection is too limited, it will also lead to errors in the parameter adjustment results. Therefore, a loss function is set to judge the accuracy of parameter optimization based on the loss function. If the loss function value meets the threshold, it means that the reliability of this parameter adjustment is high and the parameter adjustment result can be used. Otherwise, the parameter adjustment result will not be used.

[0080] The loss function is:

[0081]

[0082] Among them, N represents the number of scenic spots involved in the data, Y i is the actual visiting time distribution of the i-th scenic spot, is the visiting time distribution obtained by the simulation program under the i-th scenic spot, is the actual average visiting time of the entire scenic area, T is the simulated average visiting time of the entire scenic area, and λ represents the importance of the total average visiting time. An embodiment of the present application is: setting λ to 2.

[0083] A method for simulating human flow, such as Figure 2 As shown, it includes initializing the entity model of tourists and the regional model of scenic spots. The initialization entity model assigns corresponding initial attribute values ​​to each tourist, and the initialization regional model calibrates the attributes of different grids to facilitate subsequent predictions.

[0084] Get the flow of people in all grids at time i and determine the movement speed of tourists in all grids.

[0085] Based on the exit decision, it is predicted whether the tourist will continue to visit the scenic spot. When the prediction result is yes, it means that the tourist wants to continue to visit the scenic spot. The grid where the tourist's current location is located is obtained to determine whether the tourist enters the grid for the first time. If so, it means that the tourist arrives at the location for the first time. The tourist's next attraction to be visited is predicted through the scenic spot selection decision, and then the tourist's navigation path is obtained through BFS, and the tourist moves along the navigation path; if not, it means that the tourist is not entering the grid for the first time. The state attribute of the tourist is obtained to determine whether the tourist has an identification of moving to the target location, so as to determine whether the tourist has arrived at the target attraction. If not, it means that the tourist is heading to the attraction to be visited along the generated navigation path. The generated navigation path is obtained, and the tourist continues to move along the navigation path. If not, it means that the tourist is not in a moving state. The scenic spot retention decision is used to predict whether the user will stay and visit the attraction. If yes, a navigation path for visiting the attraction is generated based on the characteristics of the attraction and BFS, and the tourist moves along the navigation path. If no, the tourist's next attraction to be visited is predicted through the scenic spot selection decision, and the tourist's navigation path is obtained through BFS, and the tourist moves along the navigation path. At time i+1, the status and energy of all tourists in the prediction model are updated, and after executing the exit decision, the process jumps to the step of judging whether the tourists continue to visit the scenic spot.

[0086] Based on the scenic area exit decision, predict whether the tourist will continue to visit the scenic area. When the prediction result is no, it means that the tourist is about to leave the scenic area. Get the current location information of the tourist and judge whether the tourist has completely left a certain scenic spot. If yes, jump to the step of judging whether the tourist continues to visit the scenic area. If no, continue to judge whether the tourist has determined the best exit. No means that the action of leaving the scenic area is triggered for the first time. Get the current location information of the tourist to determine the nearest best exit. Get the navigation path of the tourist through BFS. The tourist moves along the navigation path. No means that it is not the first time to trigger the action of leaving the scenic area. Get the navigation route generated when it is triggered for the first time. The tourist moves along the navigation path. Before time i+1, get the state of the tourist and judge whether the tourist has completely left the scenic area (whether the tourist has a mark of leaving the scenic area). Yes means that the tourist has left the scenic area and the tourist path prediction is ended. No means that the tourist is still in the scenic area. At time i+1, update the state and energy of the tourist and jump to the step of judging whether the tourist continues to visit the scenic area.

[0087] The prediction model traverses the behavior models of all tourists at time i to individually predict the movement paths of all tourists. At time i+1, it updates the location information, visit status information, energy and other information of tourists, traverses the behavior models of all tourists and continues to predict the movement paths of tourists until all tourists leave the scenic area.

[0088] By artificially setting the age combination of tourists and the time curve of tourists entering the scenic area, and then based on the crowd flow prediction model and prediction method, the change of tourist flow in each scenic spot in the scenic area (crowd flow change curve) and the length of tourist visit are predicted, so as to predict the tourist carrying range of the scenic area.

[0089] One embodiment of the present application is: when some scenic spots in a scenic area are closed for maintenance, the model is used to predict in advance the actual tourist carrying range of the scenic area, thereby reducing the probability that tourists buy tickets but cannot enter the scenic area or cannot enjoy the scenic area, and also avoiding the probability of damage to ancient relics due to too many tourists in the scenic area.

[0090] The embodiments of the present invention are described in detail above, but the contents described are only preferred embodiments of the present invention and cannot be considered to limit the scope of implementation of the present invention. All equivalent changes and improvements made within the scope of the present invention should still fall within the scope of this patent.

Claims

1. An Agent-based crowd flow simulation model, characterized in that: include, Construct a regional model of the scenic area. The scenic area is evenly divided into several grids. The basic attributes, carrying attributes and reachable attributes of the grids are set to construct grid attributes. All grid attributes together constitute the regional model of the scenic area. Constructing a physical model of tourists, setting basic attributes, tour attributes and status attributes of tourists to form the physical model of tourists; Construct a tourist behavior model, and set the tourist's attraction stay decision, attraction selection decision, tourist movement speed, and exit decision based on the entity model to form the tourist's behavior model; Crowd flow prediction: set some tourist information to generate several entity models, bring the entity model into the behavior model and regional model, and predict the movement path of each tourist in the regional model separately to predict the changes in the flow of people in different scenic spots.

2. The Agent-based crowd flow simulation model according to claim 1, characterized in that: The basic attributes of the grid record the address information and attractiveness of the grid, the carrying capacity attribute of the grid records the maximum number of tourists that the grid can carry and the tourist movement speed corresponding to different tourist loads, and the reachability attribute of the grid records whether the grid can carry tourists.

3. The Agent-based crowd flow simulation model according to claim 1, characterized in that: The basic attributes of the tourist include the tourist's age information, location information, time of entry into the scenic area, and time of exiting the scenic area. The tourist's sightseeing attributes include the tourist's energy attribute and interest attribute. The energy attribute includes the tourist's energy for visiting the scenic area. The interest attribute includes the interest level of the scenic spot, the interest decay level of the scenic spot, the interest recovery level of the scenic spot, the minimum interest value of the scenic spot, and the maximum interest value of the scenic spot. The tourist's status attributes include information on the scenic spots that the tourist has not visited, information on the scenic spots that the tourist has visited, an indication of whether the tourist has moved to the target location, and an indication of whether the tourist has left the scenic area.

4. The Agent-based crowd flow simulation model according to claim 1, characterized in that: The decision to stay at a scenic spot is based on two conditions: first, when the tourist's interest in the current scenic spot drops below the minimum interest of the scenic spot, the tourist will tend to leave; second, if the tourist's stay at the current scenic spot exceeds a reasonable time, it will also trigger the decision to leave. For the current scenic spot, the formula for updating the interest is: USE i，current,t =(1-IDR i,current )(1-ADT i，current )×(USE i，current,t-1 -CAT i，current ), Among them, SIV i,current,t and SIV i,current,t-1 represents the interest value of the i-th tourist in the current attraction at time t and time t-1, IDR i,current represents the decay rate of the interest value of the i-th tourist in the current attraction, ADT i,current It represents the interest recovery rate of the i-th tourist in the scenic spot he is visiting, MIV i,current It represents the minimum interest value of the i-th tourist in the current attraction.

5. The Agent-based crowd flow simulation model according to claim 1, characterized in that: The attraction selection decision follows a priority rule, that is, tourists evaluate the attractions that have not been visited based on the current interest value and choose the attractions with the highest interest to visit. For the unvisited attractions, the formula for updating the interest is: USE i，j，t =(IRR i，j -1)(USE i，j,t-1 -CAT i，j ), Among them, SIV i,j,t and SIV i,j,t-1 represents the interest value of the i-th tourist for the j-th attraction at time t and time t-1, IRR i,j represents the interest recovery rate of the i-th tourist in the j-th unvisited attraction, MAV i,j Indicates the maximum interest of the i-th tourist in the j-th attraction.

6. The Agent-based crowd flow simulation model according to claim 1, characterized in that: The formula for the tourist moving speed is: Among them, SR x,y,t Represents the speed ratio at time t in the grid corresponding to the x and y coordinate positions, SC x,y,t Represents the ratio of the total number of people in the grid corresponding to the x, y coordinate position to the maximum capacity.

7. The Agent-based crowd flow simulation model according to claim 1, characterized in that: The exit decision includes obtaining the state attribute of the tourist, judging whether the list of unvisited attractions is empty, if yes, leaving the scenic spot, if no, selecting the most interesting attraction from the list of unvisited attractions for the next visit; The exit decision includes determining whether the current energy value of the tourist is lower than the energy threshold through an energy calculation formula, if yes, leaving the scenic spot, if no, continuing to play; The exit decision includes obtaining the moving speed of the tourists, judging whether the time taken by the tourists to move to the nearest exit exceeds the closing time of the scenic spot, if yes, leaving the scenic spot, if no, continuing to play.

8. The Agent-based crowd flow simulation model according to claim 7, characterized in that: The tourist energy update calculation formula is: Among them, E i,t and E i,t-1 represents the energy value of the i-th tourist at time t and time t-1, EDR i represents the energy decay rate of the i-th visitor, E mini Represents the minimum energy of the i-th tourist.

9. The Agent-based crowd flow simulation model according to claim 1, characterized in that: The crowd flow prediction includes optimizing the regional model and entity model parameters through Optuna, and the Optuna optimization includes judging the optimization result through the loss function; the loss function is: Among them, N represents the number of scenic spots involved in the data, Y i is the actual visiting time distribution of the i-th scenic spot, is the visiting time distribution obtained by the simulation program under the i-th scenic spot, is the real average visiting time of the whole scenic spot, T is the simulated average visiting time of the whole scenic spot, and λ represents the importance of the total average visiting time.

10. A crowd flow simulation method, based on an agent-based crowd flow simulation model according to any one of claims 1 to 9, characterized in that: It includes initializing the physical model of tourists and the regional model of scenic spots, predicting whether tourists will continue to visit based on exit decisions, if not, determining the best exit, generating a navigation path for the scenic spot and moving, if yes, determining whether tourists enter the grid for the first time; It is the first time to enter the grid area, predict the attractions selected by tourists based on the attraction selection decision, generate a navigation path and move; If it is not the first time for tourists to enter the grid area, determine whether the tourists have reached the target scenic spot. If not, move along the existing navigation path. If yes, determine whether the tourists stay to visit through the scenic spot retention decision; If the tourist stays to visit, a navigation path is generated and the tourist moves according to the next visit location in the current scenic spot; if the tourist does not stay to visit, a navigation path is generated and the tourist moves according to the scenic spot selection decision to predict the tourist’s choice of visit spot; Update the visitor's status and capabilities and jump to the step of predicting whether the visitor will continue the tour based on the exit decision.

Images