CN-121997768-A - Dynamic crowd evacuation simulation method based on generation type artificial intelligence and application thereof

CN121997768ACN 121997768 ACN121997768 ACN 121997768ACN-121997768-A

Abstract

The application discloses a dynamic crowd evacuation simulation method based on a generated artificial intelligence and application thereof, wherein the method comprises the steps of constructing an evacuation scene model; the method comprises the steps of initializing a behavior state model, dynamically generating individual behavior parameters based on the state information of pedestrians and a pre-trained generation type artificial intelligent model, carrying out evacuation simulation in the evacuation scene model, planning an evacuation path based on the individual behavior parameters and the state information in each time step of simulation, simulating the motion behaviors of the pedestrians based on the state information of the pedestrians, the individual behavior parameters and the evacuation path, and updating the behavior state model until the simulation is finished. According to the application, the individual behavior parameters of the environmental awareness are generated based on the generated artificial intelligent model, and meanwhile, the route generation, the route selection and the route re-planning are performed based on the conflict awareness, so that the stability, the authenticity and the application range of the crowd evacuation model in a complex scene are effectively improved.

Inventors

CAO KECAI
XIA XINYU
GU JUPING
WANG ZELIN
WANG XIAOLIANG
Cheng Hanjing

Assignees

苏州科技大学

Dates

Publication Date: 20260508
Application Date: 20260211

Claims (13)

1. The dynamic crowd evacuation simulation method based on the generated artificial intelligence is characterized by comprising the following steps of: Constructing an evacuation scene model, wherein the evacuation scene is used for describing the space range of an evacuation area, the position of an exit area and the position of an obstacle; initializing a behavior state model, wherein the behavior state model is used for describing state information of each pedestrian in the evacuation crowd at any moment; Dynamically generating individual behavior parameters based on the state information of the pedestrians and a pre-trained generation type artificial intelligent model; And carrying out evacuation simulation in the evacuation scene model, planning an evacuation path based on the individual behavior parameters and the state information in each time step of the simulation, simulating the motion behavior of the pedestrian based on the state information of the pedestrian, the individual behavior parameters and the evacuation path, and updating the behavior state model until the simulation is finished.
2. The dynamic crowd evacuation simulation method of claim 1, wherein the individual behavior parameters include a desired speed parameter, a reaction time parameter, an inter-pedestrian avoidance weight, an obstacle avoidance weight, and a collision avoidance intensity parameter.
3. The dynamic crowd evacuation simulation method of claim 2, wherein the status information includes location, speed, and target egress; The step of dynamically generating individual behavior parameters based on the pedestrian state information and the pre-trained generated artificial intelligent model comprises the following steps: Based on the state information, acquiring environment semantic information of a target pedestrian and evacuation task constraint, wherein the environment semantic information comprises crowd density in a local vicinity where the target pedestrian is located, a distance from the target pedestrian to a target exit and a distance between the target pedestrian and a nearest barrier; And inputting the state information, the environment semantic information, the evacuation task constraint and the random hidden variable into the generated artificial intelligent model to obtain individual behavior parameters of the target pedestrian.
4. A dynamic crowd evacuation simulation method according to claim 3, wherein the step of planning an evacuation path based on the individual behavior parameters and the state information comprises: constructing a local conflict neighborhood of the target pedestrian based on the individual behavior parameters; Inputting the state information, the local conflict neighborhood, the environment semantic information, the individual behavior parameters and the random hidden variables into a generated path strategy model to obtain a plurality of candidate paths; calculating the speed conflict risk of the target pedestrian and other pedestrians in the local conflict neighborhood in each candidate path to obtain conflict cost; calculating the space conflict risk of the target pedestrian and the obstacle in each candidate path to obtain obstacle cost; Calculating the path length cost of each candidate path; And obtaining the comprehensive cost of each candidate path based on the conflict cost, the obstacle cost and the path length cost, and selecting the candidate path with the optimal comprehensive cost as an evacuation path.
5. The dynamic crowd evacuation simulation method of claim 4, wherein the formula of the local conflict neighborhood is as follows: ; in the collision sense radius , wherein, As a parameter of the weight-bearing element, In order to provide the desired speed parameter, Is the reaction time parameter, and/or, The calculation formula of the conflict cost is as follows: , in the formula, For candidate paths, s represents a path point in the candidate path, For a local conflict neighborhood of a target pedestrian, i represents the target pedestrian, j represents other pedestrians located in the local conflict neighborhood, Representing the risk of speed collision at the path point s, , For the position of the time step t, At a speed of t time steps, and/or, The calculation formula of the obstacle cost is as follows: , in the formula, For candidate paths, s represents a path point in the candidate path, Representing the risk of spatial collision at the path point s, , For the collision avoidance intensity parameter, Distance of the target pedestrian from the nearest obstacle at time step t, and/or, The calculation formula of the path length cost is as follows: 。
6. The method of claim 2, wherein the status information includes a location, a speed, and a target exit, and wherein simulating the movement behavior of the pedestrian based on the status information, the individual behavior parameters, and the evacuation path comprises: Determining an expected movement direction of a target pedestrian based on the target exit and the evacuation path; Obtaining a desired driving force of the target pedestrian based on the desired movement direction, the desired speed parameter, the reaction time parameter and the speed of the target pedestrian; Based on the continuous social force model, determining pedestrian repulsive force and obstacle repulsive force of the target pedestrian; based on the pedestrian repulsive force and the inter-pedestrian avoidance weight, obtaining the pedestrian driving force of the target pedestrian; Obtaining an obstacle driving force of a target pedestrian based on the obstacle repulsive force and the obstacle avoidance weight; Based on the expected driving force, the pedestrian driving force, and the obstacle driving force, an instantaneous acceleration of the target pedestrian is determined, and based on the instantaneous acceleration, the state information of the target pedestrian is updated.
7. The dynamic crowd evacuation simulation method of claim 6, wherein the step of determining the expected direction of movement of the target pedestrian based on the target exit and the evacuation path is specified by the following formula: ; in the formula, Representing the geometric direction in which the target pedestrian is pointing toward its target exit, Represents a path direction determined based on the evacuation path, A response coefficient modified for the target pedestrian strategy, and/or, The calculation formula of the instantaneous acceleration is as follows: ; Where i denotes a target pedestrian, j denotes other pedestrians, k denotes an obstacle, Indicating the direction of the desired movement in question, Which is indicative of the desired speed parameter, The parameter of the reaction time is indicated, Representing the weight of avoidance between the pedestrians, Representing the repulsive force of the pedestrian, Represents the avoidance weight of the obstacle, Representing the obstacle repulsive force.
8. The dynamic crowd evacuation simulation method of claim 1, further comprising: After the simulation is finished, constructing a global feedback index based on the average speed, the average conflict occurrence rate and the evacuation success ratio in the evacuation process; based on the global feedback index, a performance evaluation function is constructed, wherein the performance evaluation function has the following specific formula: ; in the formula, As the weight coefficient of the light-emitting diode, Indicating the average speed of the evacuation process, Representing the average collision occurrence rate per unit time, Indicating the proportion of individuals who successfully reach the evacuation outlet within a prescribed time.
9. The dynamic crowd evacuation simulation method of claim 8, wherein the training method of the generated artificial intelligence model comprises: performing multiple evacuation simulation in a target evacuation scene based on the generated artificial intelligent model, and constructing the performance evaluation function after each simulation is finished; Based on the performance evaluation function, constructing a loss function, wherein the loss function specifically has the following formula: In which, in the process, As a function of the evaluation of the performance, Representing a smooth regularization term, -For the individual behavioral parameters in the kth simulation; Based on the loss function, updating parameters of the generated artificial intelligence model along the gradient descending direction until the loss function converges.
10. The dynamic crowd evacuation simulation method of claim 9, further comprising: Acquiring performance evaluation functions of a generated artificial intelligent model trained in different evacuation scenes and individual behavior parameters generated in simulation; based on the performance evaluation function and the individual behavior parameters in different evacuation scenes, constructing an objective function, wherein the objective function specifically has the following formula: In which, in the process, Represents the expected value of the performance evaluation function in the mth evacuation scenario, Represents the cross-scene consistency weight coefficient, Representing the individual behavioral parameters in the mth evacuation scenario, Euclidean distance for expressing the individual behavior parameters of the mth and nth evacuation scenarios; Updating parameters of the generated artificial intelligence model in a direction that minimizes the objective function.
11. A dynamic crowd evacuation simulation device based on generated artificial intelligence, comprising: A scene construction module for constructing an evacuation scene model, wherein the evacuation scene is used for describing the space range of an evacuation area, the position of an exit area and the position of an obstacle; the system comprises a state model construction module, a state model generation module and a state model generation module, wherein the state model construction module is used for initializing a behavior state model, and the behavior state model is used for describing state information of each pedestrian in the evacuation crowd at any moment; The individual parameter generation module is used for dynamically generating individual behavior parameters based on the state information of the pedestrians and a pre-trained generation type artificial intelligent model; The simulation module is used for carrying out evacuation simulation in the evacuation scene model, planning an evacuation path based on the individual behavior parameters and the state information in each time step of the simulation, simulating the motion behavior of the pedestrian based on the state information of the pedestrian, the individual behavior parameters and the evacuation path, and updating the behavior state model until the simulation is finished.
12. An electronic device, comprising: At least one processor, and A memory storing instructions that, when executed by the at least one processor, cause the at least one processor to perform the generated artificial intelligence based dynamic crowd evacuation simulation method of any one of claims 1 to 11.
13. A machine-readable storage medium storing executable instructions that when executed cause the machine to perform the generated artificial intelligence based dynamic crowd evacuation simulation method of any of claims 1 to 11.

Description

Dynamic crowd evacuation simulation method based on generation type artificial intelligence and application thereof Technical Field The application belongs to the technical field of crowd evacuation modeling and simulation, and particularly relates to a dynamic crowd evacuation simulation method based on generation type artificial intelligence and application thereof. Background With the continuous promotion of the urban process, the personnel density of large public buildings, transportation hubs, underground spaces and complex places is continuously increased, and the efficient and reliable crowd evacuation capability becomes one of the key technical problems in the fields of public security engineering and urban emergency management under the emergency conditions of fire, earthquake, explosion and the like. In order to evaluate evacuation safety in a planning and design stage and assist decision analysis in emergency management, the evacuation modeling and simulation method capable of truly describing crowd movement and behavior evolution features is constructed, and has important engineering application value. The existing crowd evacuation modeling method can be divided into a macroscopic model, a microscopic model and a macro-micro mixed model. The microscopic model uses an individual as a modeling object, describes the motion and interaction process of pedestrians in space through a rule driving or dynamic equation, can reflect phenomena such as local crowding, conflict, path selection and the like more truly, and is widely applied to evacuation simulation and safety evaluation. In a plurality of microscopic crowd modeling methods, a social force model (Social Force Model, SFM) is one of the most representative basic models in current crowd evacuation research and engineering application due to the advantages of clear physical meaning, simple model structure, easiness in combination with a numerical simulation platform and the like. The model constructs a continuous time kinetic equation by introducing the terms of expected force, pedestrian-pedestrian acting force, pedestrian-obstacle acting force and the like, thereby describing the motion behavior evolution of an individual in the evacuation process. The crowd simulation method based on the social force model is integrated into various evacuation analysis software and research platforms, and has higher engineering maturity. However, with the continuous increase of the complexity of the evacuation scene, the traditional social force model gradually exposes a series of technical defects which are difficult to avoid in practical application. Firstly, behavioral parameters (such as expected speed, social strength, psychological distance and the like) in the model usually depend on manual experience setting or static calibration, and dynamic self-adaptive adjustment is difficult to realize according to crowd density change, environmental constraint or sudden disturbance, so that the applicability of the model in multi-density, multi-outlet and complex obstacle environments is limited. Secondly, the traditional social force model has limited capability of describing individual heterogeneity, behavior differences among different individuals are often approximated through a small amount of parameter disturbance, and diversified decision and response behaviors of people in an emergency state are difficult to truly reflect. In addition, in a high-density evacuation or bottleneck scene, traditional social force models are easy to generate non-physical phenomena such as local congestion, oscillation amplification, conflict accumulation and the like, so that evacuation efficiency evaluation results are deviated. Some existing improvement methods mainly release the problems by introducing additional constraint rules, local heuristic strategies or parameter adjustment aiming at specific scenes, but the whole method still belongs to passive correction, lacks uniform modeling of a pedestrian behavior strategy generation mechanism, and is difficult to realize migration and generalization among different scenes. Disclosure of Invention The application aims to provide a dynamic crowd evacuation simulation method based on a generated artificial intelligence and application thereof, so as to solve the problems that in the prior art, the behavior parameters of crowd evacuation modeling through a traditional social force model are difficult to adaptively adjust and individual behavior differences are difficult to truly reflect. In order to achieve the above object, a first aspect of the present application provides a dynamic crowd evacuation simulation method based on generated artificial intelligence, including: Constructing an evacuation scene model, wherein the evacuation scene is used for describing the space range of an evacuation area, the position of an exit area and the position of an obstacle; initializing a behavior state model, wherein the behavior state model