Improved social force model parameter tuning target method, device, equipment and medium

Abstract

The application provides an improved social force model parameter tuning method, device and equipment and a storage medium, and relates to the technical field of virtual group simulation. The method comprises the following steps: obtaining a virtual group simulation target scene and evaluation indexes determined according to the target scene; setting multi-objective particle swarm optimization algorithm parameters by taking interaction force parameters A, n, n', gamma and lambda in the improved social force model as decision variables and taking the evaluation indexes as a target function; setting virtual group simulation experiment parameters for the target scene; and completing evaluation of particles in the multi-objective particle swarm optimization algorithm through virtual group simulation experiment of the target scene to obtain calibrated values of the A, n, n', gamma and lambda under the optimized target scene. The method can quickly obtain optimal model parameters under different simulation scenes, and virtual group simulation using the optimized model parameters can improve the motion interaction effect of the virtual group.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of virtual crowd simulation, and in particular to an improved social force model parameter tuning method and device. BACKGROUND

[0002] The model parameters of a virtual crowd simulation model have a certain influence on the motion interaction effect of the virtual crowd. The parameters of the improved social force model can be divided into two parts according to whether they have actual physical meaning. The radius, desired speed, and neighbor range of the agent are parameters that have actual physical meaning and can reflect the physical and social attributes of the agent itself. The interaction force related parameters do not have actual physical meaning, and there is no real reference for the value range. Table 1 shows the original parameter settings of the improved social force model. This set of parameters is obtained by fitting the model with real crowd data through experiments by et al. Due to the single experimental scene, these parameters cannot fully reflect the virtual crowd motion interaction, and when the scene changes, the model using the same set of parameters may not achieve the best simulation effect in the new scene.

[0003] In the application process, the values of this part of the parameters are mostly set based on experience. When the simulation scene changes, the original parameter set will no longer be suitable for the new scene, and it is necessary to re-tune the appropriate parameter configuration through multiple experiments. This manual tuning of the model parameters is very inefficient and it is difficult to ensure that the new parameters selected can maximize the performance of the algorithm. When there are multiple such parameters in the model, the effective range and interaction of these parameters form a high-dimensional space, and blind search of such a high-dimensional space manually will consume a lot of time and resources, and often cannot obtain a satisfactory parameter combination in a short time. SUMMARY

[0004] The present application provides an improved social force model parameter tuning method and device, aiming to realize flexible and efficient virtual crowd simulation model parameter tuning to solve the technical problems existing in the background art.

[0005] In a first aspect, the present application provides an improved social force model parameter tuning method, comprising:

[0006] obtaining a virtual crowd simulation target scene and an evaluation index determined according to the target scene;

[0007] setting the interaction force parameters in the improved social force model as decision variables, and setting the evaluation index as the objective function, and setting the multi-objective particle swarm optimization algorithm parameters; A, n, n', γ, λ

[0008] setting virtual crowd simulation experiment parameters for the target scene;​

[0009] Through virtual swarm simulation experiments on the target scenario, the particles in the multi-objective particle swarm optimization algorithm are evaluated, and the optimized particle swarm optimization algorithm for the target scenario is obtained. A, n, n', γ, λ The calibration value.

[0010] Furthermore, the target scenario is a virtual group movement interaction scenario within a virtual simulation scenario.

[0011] Furthermore, the evaluation index uses the ability to avoid collisions during the virtual group's motion interaction as an optimization index for the improved social force model. It includes evaluation indexes for motion interaction between virtual groups and global motion interaction, specifically three evaluation indexes: average number of collisions per person (AACC), average collision depth (AACD), and global collision score (CS).

[0012] More specifically, the decision variables are represented by an array Param = { A, n, n', γ, λ} represents the calibration value of the interaction force parameter in the target scenario to be solved; for all particles in the multi-objective particle swarm optimization algorithm, at time t, the particle i Location in solution space x i ( t () represents a specific set of Param values ​​for particles. i After the current position is evaluated by the objective function, the particle is obtained. i The infeasibility and target value at the current location are given, with the target value represented by the array cost. i ( t )={ n 1 , n 2 , n 3}, of which n 1 , n 2 , n 3 corresponds to the results of particle position evaluation by three evaluation indicators: average collision count (AACC), average collision depth (AACD), and global collision score (CS). Based on the decision variables and target values, the parameters in the multi-objective particle swarm optimization algorithm are determined and set, including the decision space dimension, target space dimension, search range of the decision space, particle swarm size, maximum warehouse size, and maximum number of iterations.

[0013] Furthermore, the virtual swarm simulation experiment uses an improved social force model, and the experimental parameters include agent radius, perceived neighbor range, agent desired rate, and simulation limit duration.

[0014] More specifically, the process of evaluating particles in the multi-objective particle swarm optimization algorithm yields the optimized target scenario. A, n, n', γ, λ The specific method for calibrating the values ​​is as follows:

[0015] initializing a particle swarm, a warehouse and an iteration number using the multi-objective particle swarm optimization algorithm parameters;

[0016] a position of a particle in the particle swarm x i t , an improved social force model with the position of the particle as an interaction force parameter input, and virtual crowd simulation experiment parameters set for the target scene, performing a virtual crowd simulation experiment on the target scene, returning an evaluation result of the target function on the current position of the particle through the simulation experiment, the evaluation result including a target value array cost i t = { n 1 , n 2 , n 3} and infeasibility;

[0017] updating a particle speed, a position, a dominance attribute, an individual optimum pBest, a global optimum gBest in the multi-objective particle swarm optimization algorithm, and a non-dominated solution in the warehouse according to the target value array and the infeasibility of the particle; completing iteration updating of the particle swarm and evaluation of the particles in the particle swarm according to a preset iteration number, and outputting an optimal non-dominated solution set;

[0018] selecting a solution from the optimal non-dominated solution set to calibrate A, n, n', γ, λ parameters under the target scene.

[0019] Further, the infeasibility of the particle refers to interrupting a simulation program and assigning a maximum value to the infeasibility of the particle when the simulation experiment exceeds the simulation limit duration and is not completed, indicating that the parameter value represented by the position of the particle is unusable. A, n, n', γ, λ

[0020] In a second aspect, the present application provides an improved social force model parameter tuning and calibration device, characterized in that it comprises:

[0021] an input module configured to obtain a virtual crowd simulation target scene and an evaluation index determined according to the target scene;

[0022] a parameter setting module configured to set multi-objective particle swarm optimization algorithm parameters with an interaction force parameter in the improved social force model as a decision variable and the evaluation index as a target function, and set virtual crowd simulation experiment parameters for the target scene; A, n, n', γ, λ

[0023] a tuning and calibration module configured to complete evaluation of particles in the multi-objective particle swarm optimization algorithm through a virtual crowd simulation experiment on the target scene, and obtain the A, n, n', γ, λ ​​​​a calibration value of the target.

[0024] In a third aspect, the present application also provides an electronic device, comprising: a processor; and a memory storing a program executable by the processor, wherein when the program is executed, the processor executes the method of any one of the preceding first aspect.

[0025] In a fourth aspect, the present application also provides a storage medium storing a program executable by a processor, wherein when the program is executed, the processor executes the method of any one of the preceding first aspect.

[0026] By adopting the technical scheme, the present application has the following beneficial effects:

[0027] The parameters of the virtual group simulation model determine the simulation effect. Using intelligent optimization algorithm to optimize the parameters of the virtual group simulation model can not only find the best parameter configuration in the limited parameter range by the global search ability of the optimization algorithm, improve the simulation accuracy of the model, but also reduce the waste of resources by the parallel processing ability of the algorithm, and quickly obtain the optimal model parameters corresponding to different simulation scenes.

[0028] The present application takes improving the collision avoidance ability in the virtual group motion interaction process as the optimization target of improving the social force model, and comprehensively evaluates the optimization target from multiple dimensions. The multi-objective particle swarm optimization algorithm is selected, and the selection, combination and switching of multiple parameters and the comprehensive evaluation of different parameters in each dimension can be efficiently realized by the algorithm. It can be seen from the experimental results that the optimal model parameters under different simulation scenes can be quickly obtained by using the method of the present application, and the motion interaction effect of the virtual group can be improved by using the optimized model parameters for virtual group simulation. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the specific embodiments of the present application or the technical solutions in the prior art, the drawings needed in the specific embodiments or prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0030] Figure 1 The improved social force model parameter optimization calibration method flow chart provided for the embodiments of the present application is shown in the figure.

[0031] Figure 2 The Agent-Agent collision schematic diagram in the virtual group simulation model evaluation method provided for the embodiments of the present application is shown in the figure.

[0032] Figure 3A flow chart of a multi-objective particle swarm optimization algorithm provided for an embodiment of the present application is shown in FIG. 1.

[0033] Figure 4 A flow chart of a particle evaluation process provided for an embodiment of the present application is shown in FIG. 2.

[0034] Figure 5 A diagram showing virtual group distribution in different scenarios of a simulation experiment scenario provided for an embodiment of the present application is shown in FIG. 3.

[0035] Figure 6 A diagram showing a comparison of virtual group motion trajectories before and after optimization of model parameters provided for an embodiment of the present application is shown in FIG. 4.

[0036] Figure 7 A diagram showing a comparison of collision conditions in different scenarios of an improved social force model before and after optimization of model parameters provided for an embodiment of the present application is shown in FIG. 5. DETAILED DESCRIPTION

[0037] The technical solutions of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0038] The present application will be further explained and described below with reference to specific embodiments.

[0039] Before the embodiments of the present application are described in detail, an improved social force model and a multi-objective particle swarm optimization algorithm are first introduced to facilitate understanding of the embodiments of the present application.

[0040] Improved Social Force Model

[0041] The improved social force model proposed by Li et al. verifies the pedestrian interaction behavior in the classical social force model and extracts the behavior rules, making the model more consistent with the pedestrian motion rules in the actual scenario. Currently, this model has been widely used in virtual group simulation research and has become a basic and effective tool for analyzing and predicting complex group dynamic behavior.

[0042] The core principle of the improved social force model algorithm is the same as that of the classical social force model, which considers that the motion of a pedestrian is affected by three social forces, i.e., self-driving force, repulsive force of other pedestrians, and repulsive force of wall obstacles. The motion speed of a pedestrian is affected by the social force as shown in formula (1).

[0043] (1)

[0044] : The internal acceleration term reflects the pedestrian's motivation to move towards their destination at a specific speed. This part of the definition follows the classical social force model's definition of self-driving force:

[0045] (2)

[0046] Indicates the relaxation time Inside, make Current speed To the desired rate and target direction Approximately. Through experimental fitting, the desired rate was obtained. Relaxation time .

[0047] :wall The repulsive force, and vertical distance to the wall Inversely proportional, which can be expressed as Related functions:

[0048] (3)

[0049] in, , All are parameters, and are set. =3, =0.1.

[0050] : and The interaction effect produced by the interaction between them, according to the The influence of motion state can transform interaction forces The function is broken down into along and Interaction direction Speed ​​change (deceleration in collision avoidance mode), and along Normal direction The direction of change. Represented as:

[0051] (4)

[0052] and They described along deceleration and along The direction of change.

[0053] Interaction direction yes and The relative motion direction in combination with the relative position between the two is denoted as:

[0054] (5)

[0055] where, are model parameters, , denote the velocities of and respectively, denotes the unit vector pointing from the position of to the position of , and is denoted as:

[0056] (6)

[0057] The distance between and is defined as The angle between the vector and the vector can be expressed as a function of and :

[0058] (7)

[0059] (8)

[0060] where, denotes the sign of the angle ; ; A, n, n', γ, λ Figure 2 are model parameters, The parameter values obtained by fitting the experimental data by et al. are shown in Table 1.

[0061]

[0062] Substituting equations (7) and (8) into equation (4) and expanding, the interaction force between and can be expressed as a function of the distance and the angle :

[0063] (9)

[0064] ​​Adjusting the model parameters in the formula can realize virtual groups with different motion interaction characteristics. Based on the formula, the improved social force model is optimized, which will improve the Agent-Agent motion interaction effect of the improved social force.

[0065] Multi-objective particle swarm optimization algorithm

[0066] Particle Swarm Optimization (PSO) is a swarm intelligence optimization algorithm, which is inspired by the foraging behavior of bird flocks in nature. The algorithm simulates the group cooperation process of bird flocks in foraging, and finds the optimal solution of the problem through information sharing and cooperative search between particles. Each particle in PSO represents a potential solution and has its own speed and position in the solution space. At the beginning, the positions and speeds of all particles are randomly generated to complete the initialization of the particle swarm. In each round of evolutionary iteration, first, the fitness function of all particles is evaluated, and the individual optimal solution of each particle and the global optimal solution of the particle swarm are updated according to the evaluation results, and the speed and position of the particle are adjusted to make the whole particle swarm evolve towards the global optimal direction until the termination condition is met.

[0067] Particle Swarm Optimization is simple to implement and has fast convergence speed, and is particularly suitable for solving continuous optimization problems and application scenarios with high requirements for algorithm execution time. However, since the particle swarm optimization algorithm represents the value of the optimization objective function by a single fitness function during the solving process, it cannot well balance multiple objective functions. In actual problems, there are often multiple optimization objectives, and there may be conflicts between objective functions, so the Multi-Objective Particle Swarm Optimization (MOPSO) algorithm is proposed, which is based on the particle swarm optimization algorithm and solves the multi-objective optimization problem, that is, to find a set of optimal solutions that can balance multiple optimization objectives.

[0068] Due to the above advantages of the multi-objective particle optimization algorithm, the multi-objective particle swarm optimization algorithm is selected in the method of the present application to optimize the interaction force parameters of the improved social force model. In order to facilitate the understanding of the method of the present application, important concepts and definitions involved in the algorithm are introduced and explained.

[0069] (1) Pareto solution and Pareto front

[0070] Pareto solution (Pareto solution): In a multi-objective optimization problem, given two solutions: solution A and solution B, when solution A is better than solution B (or solution A is not worse than solution B) for all objectives, solution A dominates solution B, and solution A is called a non-dominated solution, also called a Pareto solution.

[0071] Pareto front: A set of non-dominated solutions in the solution space, which represents the optimal solutions of a multi-objective optimization problem. Any solution in the Pareto front cannot be dominated by other solutions in the front.

[0072] (2) Swarm and Repository

[0073] Swarm: A group of particles, each representing a potential solution in the solution space. The swarm is dynamically explored, representing the real-time dynamics of the search process. Each particle has its own speed and position, constantly moving and updating its position and speed in the solution space to find the optimal solution of the problem.

[0074] Repository: An external archive of MOPSO, which stores and updates the Pareto front. The repository stores all non-dominated solutions to guide the swarm search for optimal solutions in future iterations. MOPSO uses an adaptive grid method to maintain the repository and introduces a diversity preservation mechanism to avoid the solution set from being too clustered, ensuring that the stored solutions cover different areas of the objective space.

[0075] (3) Individual best and global best

[0076] Individual best pBest: The best position of each particle in the historical search process. At initialization, the individual best of each particle is the current position of the particle; in each round of iteration, the dominance relationship between the new position of the particle and pBest is judged, and if the new position of the particle is better than pBest, the pBest of the particle is updated to the current position of the particle.

[0077] Global best gBest: The best position of all particles in the swarm in the historical search process. It is selected from the Pareto front and used to guide the search direction of the swarm. After each iteration, an adaptive grid method and a roulette wheel strategy are used to select a particle from the repository to update gBest.

[0078] Based on the above concepts and definitions, the core principles of the MOPSO algorithm can be summarized as follows: In the iteration process, each particle updates its speed and position based on the individual best and global best, exploring the multi-dimensional solution space; stores the non-dominated solutions found in the repository, constantly updating and maintaining the Pareto front, and uses specific algorithm strategies to handle constraints and maintain the diversity of solutions, effectively balancing between multiple objectives.

[0079] The specific algorithm flow of MOPSO is as follows:

[0080] (1) Initialization: generate particle swarm, randomly assign position to each particle, set initial velocity to zero, set pBest as the initial position of the particle; create and fill the warehouse; set the number of iterations t = 1.

[0081] (2) Select gBest: select particles as gBest in the warehouse grid according to the roulette strategy in the warehouse.

[0082] (3) Update particle swarm: for each particle in the particle swarm, update the particle's speed according to gBest and its pBest according to formula (10), update the particle's position according to formula (11) to get the next generation of particles. In the formula is the inertia weight, is the individual learning factor, is the group learning factor, is a random number.

[0083] (10)

[0084] (11)

[0085] (4) Evaluate the particle swarm: evaluate the position of each particle in the particle swarm, calculate the value of all objective functions corresponding to each particle; if the current position of the particle is better than pBest, update pBest to the current position of the particle. Non-dominated sorting is performed on the new generation of particle swarm, and all non-dominated solutions are stored in the warehouse.

[0086] (5) Update the warehouse: update the dominance attribute of the solution stored in the warehouse, delete the dominated solution in the warehouse. Determine whether the non-dominated solution stored in the warehouse exceeds the upper limit of the warehouse size, if it exceeds the upper limit, clear the warehouse according to the adaptive grid method and the roulette strategy.

[0087] (6) Termination condition: determine whether the termination condition is met, if it is met, return the optimal solution and end the program. Otherwise, the number of iterations t = t + 1, jump to step (2) and continue iteration. The algorithm flow is shown in the flow chart as Figure 1 .

[0088] The specific implementation of the scheme will be described below in conjunction with the drawings.

[0089] The core idea of the method is to optimize the interaction force parameters in the social force model under different simulation scenarios, so as to efficiently improve the simulation effect of the model in diversified application scenarios. How to efficiently and accurately find the optimal parameter configuration of the model under different scenarios is the key technical problem to be solved by the present application. A, n, n', γ, λ As shown in the figure, the improved social force model parameter optimization method proposed by the present application comprises:

[0090] Obtain the target scenario for virtual group simulation and the evaluation indicators determined based on the target scenario;

[0091] To improve the interaction force parameters in the social force model A, n, n', γ, λ Set the parameters of the multi-objective particle swarm optimization algorithm with the evaluation index as the objective function and the decision variable as the decision variable;

[0092] For the target scenario, set the parameters for the virtual group simulation experiment;

[0093] Through virtual swarm simulation experiments on the target scenario, the particles in the multi-objective particle swarm optimization algorithm are evaluated, and the optimized particle swarm optimization algorithm for the target scenario is obtained. Figure 3 The calibration value.

[0094] The above method uses a multi-objective particle swarm optimization algorithm to fine-tune and calibrate the interaction force parameters of the improved social force model, which can greatly improve the efficiency of parameter tuning. This allows for rapid fine-tuning and calibration of model parameters for different virtual group simulation scenarios and multi-dimensional evaluation indicators, thereby improving the motion interaction effect of the model simulation and enhancing the user experience.

[0095] Collision avoidance behavior in virtual groups enhances their realism and vividness, significantly improving user experience. In virtual group simulation models, force-based models, such as the social force model, effectively reflect individual characteristics and are suitable for simulating the dynamics of groups with social attributes and relationships. However, they are less effective at handling collision avoidance during virtual group interactions. Therefore, in this embodiment, a virtual group motion interaction scenario within a virtual simulation environment is selected as the target scenario, with improving collision avoidance capabilities during virtual group motion interaction serving as the optimization metric for the improved social force model.

[0096] Currently, there is no unified and comprehensive crowd simulation evaluation system for assessing collision avoidance behavior during virtual motion interaction. This invention focuses on the collision avoidance capabilities of virtual groups during motion interaction, primarily considering evaluation indicators related to virtual group collision avoidance, including evaluation indicators for interaction between virtual groups and global motion interaction, specifically three evaluation indicators: collision count, collision depth, and collision score.

[0097] The number of collisions is a direct reflection of the collision avoidance capability of a virtual swarm simulation model and serves as an evaluation metric for interactions between virtual swarms. The Agent is defined as a disk on a two-dimensional plane. and When the boundaries overlap, it is considered that and A collision occurs, such as Figure 3 As shown. Among them. , respectively, with the radius of , respectively, with the position. Since the collision has a certain duration in time, in order to more scientifically count the experimental data, in the present application, from the beginning of the overlap of the boundaries of the two colliding parties to the complete departure of the two parties from each other's boundary is defined as one collision.

[0098] In the present embodiment, the collision times are specifically the average agent-agent collision count (AACC), which represents the number of times of agent-agent collisions per agent on average, and can reflect the collision frequency in the agent-agent motion interaction process, with the unit of times, and the specific definition is shown in formula (12).

[0099] (12)

[0100] wherein, represents the size of the virtual group, represents the number of collisions between and in the entire simulation process.

[0101] The collision depth is one of the commonly used indicators for quantifying the severity of the collision, and is an evaluation index for reflecting the interaction between the virtual groups. The collision depth between and is defined as the length of the overlapping part of the two parties along the line connecting the centers of and , as shown in , and the specific calculation formula is as follows: A, n, n', γ, λ

[0102] (13) Without special instructions, the collision depth is defined as the maximum collision depth in the process of one collision.

[0103] In the present embodiment, the collision depth is specifically the average agent-agent collision depth (AACD), which represents the average maximum collision depth of each agent in each agent-agent collision, and can reflect the severity of the agent-agent collision in the spatial dimension, with the unit of meters, and the specific calculation formula is as follows:

[0104]

[0105] ​(14)

[0106] wherein denotes and the maximum collision depth at the kth collision, The calculation method refers to formula (13).

[0107] The collision score is an evaluation index that combines the number of collisions and the collision depth, and reflects the global motion interaction performance. It can more comprehensively evaluate the realism of virtual group simulation.

[0108] In this embodiment, the collision score (CS) is defined as the sum of the collision depth of all ongoing collisions in the entire scene at each time step in the entire simulation process, and then the total time round and the number of agents in the group are averaged. The collision score can reflect the overall collision avoidance effect of the model in the time and space dimensions. The lower the collision score, the better the overall collision avoidance performance of the model, and the higher the credibility of global motion interaction. The specific definition of CS is shown in formula (16). Wherein, denotes the time step required for the entire simulation process, denotes the collision depth of the kth collision at the time step, and

[0109] (16)

[0110] In this embodiment, it should be particularly pointed out that the collision depth used in the calculation of the collision score is different from the maximum collision depth used in the calculation of other evaluation indexes in the foregoing. The meaning of the collision depth here is to record the collision depth corresponding to each time step of a complete collision process, i.e., from the beginning of the collision to the end of the collision, and does not include the maximum value. Therefore, the collision score not only considers the severity of the collision in the spatial dimension, but also considers the influence of the duration of each collision on the global motion interaction from the time dimension.

[0111] In summary, in this embodiment, the target scene is a virtual group motion interaction scene in a virtual simulation scene, and the determined related evaluation indexes are three evaluation indexes of AACC, AACD and CS. Based on the target scene and the evaluation indexes, the multi-objective particle swarm optimization algorithm and the virtual simulation model are initialized and configured.

[0112] ​More specifically, in the experiment and result analysis of the present embodiment, three representative classic crowd simulation scenarios, namely Circle, crossroads and venue evacuation, are selected, and different crowd densities are set as the experimental scenarios for the optimization and comparative experiment of the virtual crowd model in the present embodiment.

[0113] In the present embodiment, after determining the above target scenarios and the multiple evaluation indexes for the target scenarios, the technical problem to be solved is how to use the intelligent optimization algorithm to optimize the interaction force parameters in the improved social force model A, n, n', γ, λ A, n, n', γ, λ to quickly determine the optimal parameter combination, so as to optimize the multi-dimensional simulation performance of the virtual crowd in the above target scenarios and improve the fitness of the simulation model.

[0114] The method of the present embodiment optimizes the model parameters based on the improved social force model, so as to improve the specific performance of the simulation model in the specific scenario. Since the simulation model has fewer parameters, the parameter range to be searched is smaller, the corresponding model parameter optimization problem belongs to a continuous optimization problem, and there are multiple optimization objectives, so the multi-objective particle swarm optimization algorithm can well meet the needs of the method of the present embodiment.

[0115] According to the principle and implementation process of the multi-objective particle swarm optimization algorithm, the interaction force parameters in the improved social force model A, n, n', γ, λ are used as the decision variables (i.e. the positions of the particles) in the multi-objective particle swarm, the quantitative indexes of the collision frequency, collision depth and collision score of the virtual crowd are used as the objective function of the multi-objective particle swarm optimization algorithm to evaluate the fitness of the particles, and the individual optimal solution of each particle and the global optimal solution of the particle swarm are updated according to the evaluation results, and the speed and position of the particle are adjusted.

[0116] In the present embodiment, the decision variable is represented by an array as Param = { , γ}, which represents the calibrated value of the interaction force parameter in the target scenario to be solved; for all particles in the multi-objective particle swarm optimization algorithm, at time t, the position of the particle i in the solution space x i ( t ) represents a specific group of Param values, and the current position of the particle i is evaluated by the objective function to obtain the infeasibility and target value of the particle i at the current position, and the target value is represented by an array as cost i ( t )={ n 1 , n 2 , n 3} where n 1, n 2 , n 3 corresponds to the results of particle position evaluation by three evaluation indicators: average collision count (AACC), average collision depth (AACD), and global collision score (CS). Based on the decision variables and target values, the parameters in the multi-objective particle swarm optimization algorithm are determined and set, including the decision space dimension, target space dimension, search range of the decision space, particle swarm size, maximum warehouse size, and maximum number of iterations.

[0117] In this embodiment, the interaction force module in the improved social force model is optimized, and the corresponding parameters are used as decision variables in the MOPSO algorithm, represented by an array as Param = { , λ} represents the problem to be solved, namely, the calibration values ​​of the interaction force parameters in the target scenario; the solution space has a dimension of 5, where each dimension corresponds to all possible values ​​of a decision variable. That is, for all particles in the particle swarm optimization algorithm, at time t, the particle... i Location in solution space x i ( t ) is a 5-dimensional array representing the values ​​of a specific set of decision variables, with each dimension representing the interaction force parameter of the corresponding dimension.

[0118] More specifically, the size of the solution space is determined by the number of decision variables (i.e., the dimension of the solution space) and the search range of the decision variables. Each decision variable in the solution space has a finite search range, corresponding to the numerical range of each interaction force parameter in the improved social force model. The search range of the solution space is defined by two 5-dimensional arrays, specifically the upper search bound. = {A max , n max , n′ max , γ max , λ max} and search lower bound = { A min , n min , n′ min A, n, n', γ, λ min Figure 4 min During the optimization process, all particles search within a given range of each dimension of the solution space to ensure that the optimal solution obtained is within the parameter range.

[0119] In this embodiment, the interaction force parameters of the improved social force model are optimized to enhance the motion interaction performance during virtual group simulation. The simulation model's performance is evaluated based on three aspects: collision frequency, collision severity, and overall collision situation during motion interaction. These correspond to the three evaluation metrics: average number of collisions per person (AACC), average collision depth (AACD), and global collision score (CS). To directly and effectively optimize the model's performance and ensure that the optimization objective aligns with the desired model performance improvement, the three model evaluation metrics are set as objective functions in the optimization method, resulting in a target space with a dimension of 3, where each dimension represents the value of the corresponding evaluation metric. For particles... i Current location x i ( t The position of the particle is evaluated using an objective function with corresponding evaluation indices, resulting in the particle's position. i The evaluation results at the current location include infeasibility and the target value, which is represented by an array called cost. i ( t )={ n 1 , n 2 , n 3}, of which n 1 , n 2 , n 3 represents the results of particle position evaluation based on three assessment metrics: Average Collision Count per Person (AACC), Average Collision Depth (AACD), and Global Collision Score (CS). The infeasibility of a particle refers to the situation where, during particle evaluation, the simulation experiment exceeds the specified time limit and is not completed; in this case, the simulation program is interrupted, and the particle's infeasibility is assigned a maximum value, indicating that the particle position represents... Figure 4 The parameter value is unavailable.

[0120] In addition to setting the aforementioned parameters for the decision space dimension, target space dimension, and search range of the decision space, it is also necessary to determine and set key parameters such as the particle swarm size, maximum warehouse size, and maximum number of iterations for the multi-objective particle swarm optimization algorithm based on the decision variables and target values. In this embodiment, the particle swarm size is set to 100, the maximum warehouse size is set to 20, and the maximum number of iterations is set to 50.

[0121] In this embodiment, the virtual group simulation experiment uses an improved social force model, and the experimental parameters include agent radius, perceived neighbor range, agent expected rate, and simulation limit duration.

[0122] In this embodiment, the improved social force model is parameterized by MOPSO, and the two are connected by particle evaluation. The decision variables of MOPSO (i.e. the positions of particles) are used as the interaction force parameters of the input social force model, and the improved social force model is used for virtual group simulation. The indexes obtained by simulation are fed back to the particles as target values.

[0123] More specifically, as shown in the particle evaluation process A, n, n', γ, λ , the positions of the particles in the particle swarm x i ( t ) are used as the interaction force parameters of the improved social force model, and the virtual group simulation experiment parameters set for the target scene are used for virtual group simulation experiment of the target scene. The evaluation results of the target function for the particles at the current positions are returned by the simulation experiment, and the evaluation results include the target value array cost i ( t )={ n 1 , n 2 , n 3} and infeasibility.

[0124] According to the target value array and infeasibility of the particles, the particle speed, position, dominance attribute, individual optimum pBest, global optimum gBest in the multi-objective particle swarm optimization algorithm, and the non-dominated solutions in the warehouse are updated. The iteration update of the particle swarm and the evaluation of the particles in the particle swarm are completed according to the preset iteration number, and the optimal non-dominated solution set is output.

[0125] The pseudo code of the improved social force model optimization method based on the multi-objective particle swarm optimization algorithm is shown in Table 2. Among them, evaluate() is the particle evaluation method, and the overall process is consistent with Figure 5 : the particle position is used as the parameter input, and the target value array and infeasibility of the particles at the current positions are returned after model simulation. The infeasibility represents the degree to which the solution represented by the current position of the particle does not meet the constraint condition. When evaluating the dominance attribute of the particles in the group, the infeasibility of the particles needs to be considered first. For particles with too high infeasibility, their target value array is no longer evaluated, and they are directly regarded as dominated solutions and not put into the warehouse; for multiple particles with the lowest infeasibility, the non-dominated sorting of their target value array is performed, and the non-dominated solutions among them are filled into the warehouse.

[0126] A solution is selected from the optimal non-dominated solution set to calibrate the Figure 6 parameters under the target scene.

[0127]

[0128] By using the multi-objective particle swarm optimization algorithm, selection, combination and switching of multiple parameters within respective predetermined parameter ranges can be quickly realized for specific target scenarios and optimization targets, multi-dimensional evaluation of each parameter combination is completed through virtual group simulation experiments, and the target value of the evaluation result is returned to the multi-objective particle swarm optimization algorithm, thereby completing rapid parameter combination screening, determining the optimal parameter combination, and effectively reducing resource consumption required for model tuning.

[0129] In order to further improve the authenticity of model simulation and the adaptability to different crowd densities and different simulation scenes, and qualitatively and quantitatively illustrate that the method of the present application effectively improves the performance of virtual groups in motion interaction, the present embodiment takes the classic crowd simulation scene of virtual group interactive motion as an example to perform comparative experiments and result analysis.

[0130] In the present embodiment, three classic crowd simulation scenes and different crowd densities are set, the improved social force model under different scenes is optimized, and the simulation results of the improved social force model with the optimized model parameters and the original default parameters are analyzed and compared to verify the effectiveness and superiority of the method of the present application for optimizing the virtual group simulation model.

[0131] Circle, crossroads and venue evacuation are selected as three representative classic crowd simulation scenes as the experimental scenes for virtual group model optimization and comparative experiments in the present embodiment. The specific description of the scenes is as follows:

[0132] (1) Circle scene: a circular boundary with a radius of 15m is defined, and a certain number of Agents are uniformly distributed on the circumference. At the beginning, all Agents move along the diameter direction towards the other end of the circumference, and the simulation ends when all Agents reach the symmetric position of the starting point on the circumference.

[0133] (2) Crossroads scene: the crossroads range is x∈[-40m, 40m], y∈[-40m, 40m], the road width is 10m, and the scene area is 1500 square meters. At the beginning, a certain number of Agents are uniformly and randomly distributed on the upper, lower, left and right four half sides of the crossroads, and all Agents move towards the opposite road of the road they are on. When all Agents reach the end of the opposite road of the starting point, the simulation ends.

[0134] (3) Stadium evacuation scenario: The stadium is a rectangular space with the range of x∈[-15m, 15m], y∈[-15m, 15m], and the area of the stadium is 900 square meters. There is an exit with a width of 3 meters in the middle of the right side of the stadium. At the beginning, a certain number of agents are uniformly and randomly distributed in the stadium, and all the agents move forward towards the exit. When all the agents leave the stadium from the exit, the simulation ends.

[0135] For the Circle scenario, the number of agents N = {6, 8, 10, 12, 14, 16} is set respectively. For the two application scenarios of crossroads and stadium evacuation, according to the area per capita from large to small, three different congestion levels of experimental groups are set, and the experimental group settings are shown in Table 3. The distribution of agents in different scenarios at the beginning is shown in Table 3. Figure 6

[0136]

[0137] Using the virtual population simulation model optimization method provided by the application, for the Circle scenario and each scenario listed in Table 3, the improved social force model is optimized respectively. The experimental parameter settings are as follows:

[0138] MOPSO part: the dimension of the decision space is 5, the dimension of the target space is 3, the search range of the decision variable is set to [0, 5], the particle swarm size is 100, the maximum warehouse size is 20, and the maximum number of iterations is 50.

[0139] Virtual population simulation model part: the radius of the agent is 0.35m, the perception neighbor range is 2m, and the expected speed is 1m / s; since the position of the agent in the scene is random, in the optimization process, some solutions may not be able to complete the simulation within a reasonable time range. According to the scene and the size of the virtual population, the corresponding simulation time limit timeLimit is set. In each particle evaluation, when the simulation time exceeds timeLimit and is not completed, the program is interrupted, and a maximum value is assigned to the infeasibility of the particle, so as to avoid the situation that the optimization program is stuck due to the fact that a certain particle represents a solution that cannot complete the simulation in the optimization process.

[0140] The interaction force parameters of the improved social force model in each scenario obtained after optimization using the above method are shown in Tables 4-6.

[0141]

[0142]

[0143] ​

[0144] Simulation results and analysis

[0145] (1) Comparison of movement trajectories

[0146] The virtual crowd movement trajectories generated by the improved social force model before and after optimization are obtained by virtual crowd simulation on the Circle scenario, as shown in Figure 6 Figure 7 (a) shows the virtual crowd movement trajectory generated by the improved social force model without optimization, and the interaction force parameters in the model use the default parameter settings in Table 1; Figure 7 (b) shows the virtual crowd movement trajectory generated by the improved social force model after optimization, and the optimized model uses the corresponding parameter settings in Table 4 according to the number of Agents in the scenario.

[0147] By comparing the movement trajectories before and after optimization, it can be seen that compared with the unoptimized case, the virtual crowd obtained after optimization will take a turning strategy in advance when facing an impending collision, and the turning amplitude is larger, so as to avoid the occurrence of collision, and therefore the movement trajectory presents a more obvious "vortex" around the center of the circle.

[0148] Table 7 shows the simulation performance of the improved social force model before and after optimization in the Circle scenario under different Agent numbers. For different Agent numbers, the average number of collisions per Agent and the average collision depth during simulation, as well as the total time required for simulation, are counted. From the collision situation, for the Circle scenario, the number of Agent collisions and collision depth in the simulation of the unoptimized model are positively correlated with the number of Agents in the scenario, while the optimized model can effectively avoid the occurrence of collision. From the simulation efficiency, although the virtual crowd movement trajectory generated by the optimized model is longer than that generated by the unoptimized model, the optimized model does not have much additional time overhead in terms of simulation time. The reason is that the optimized model reduces the occurrence of collision, the Agent movement is more smooth, and the speed change is smaller, so although the path of the Agent after optimization is longer, the time spent to reach the end point is close to or even shorter than the unoptimized case.

[0149]

[0150] (2) Comparison of collision situations

[0151] ​Virtual crowd simulation is performed for the intersection and venue evacuation scenarios. The optimized improved social force model (parameters in Tables 5-6 are used for the corresponding scenarios) is compared with the unoptimized model (default parameters in Table 1 are used for all scenarios). The collision conditions in the virtual crowd movement process are evaluated by three indicators, i.e., the number of collisions per Agent, the average collision depth of Agents, and the global collision score, and the experimental results are shown in Figure 7

[0152] Overall, the optimized improved social force model is generally superior to the original model using default parameters in terms of virtual crowd movement interaction performance in simulation for different simulation scenarios and congestion levels.

[0153] In terms of collision frequency, it can be seen from Figure 7 (a) and Figure 7 (d) that when the crowd distribution in the scene is loose, the number of collisions per virtual crowd is significantly lower than that of the unoptimized model, and almost no collision occurs; as the crowd density in the scene increases and the area per capita decreases, the collision frequency between Agents increases, but the number of collisions per capita of the optimized model is always lower than that of the unoptimized model.

[0154] The average collision depth of Agents is used to quantitatively evaluate the severity of Agent-Agent collision. From Figure 7 (b) and Figure 7 (e) can be seen that when the virtual crowd is loose, the optimization effect is most significant, and the average collision depth of the optimized model is reduced by more than 75% compared with the original model; when the congestion level is moderate, the optimization has a smaller effect on reducing the collision depth, but still ensures that the optimized model is better than the unoptimized model. When the virtual crowd is densely distributed in the scene, the collision depth of the original model increases significantly, while the optimized model always keeps the collision depth at a low level.

[0155] The collision score is an evaluation of the virtual crowd movement interaction in the whole simulation process, and from A, n, n', γ, λ (c) and A, n, n', γ, λ (f) can be seen that for the two scenarios and three congestion levels, the optimized model is significantly improved compared with the unoptimized original model, and the collision score is significantly lower than that of the unoptimized model.

[0156] ​Based on the above analysis, it can be concluded that after optimizing the interaction force parameters of the improved social force model using the method of the present application, the virtual group motion interaction effect can be effectively improved: by taking collision avoidance measures in advance and increasing the turning amplitude when the Agent avoids collision, the frequency of Agent collision is reduced, the depth of Agent collision is reduced, the impact of collision events on virtual group motion is reduced in both time and space dimensions during the simulation process, thereby improving the overall fluency of virtual group motion and the authenticity of motion interaction.

[0157] In other scenarios involving the use of virtual group simulation based on the improved social force model, the above-mentioned embodiments can also be used to optimize and calibrate the interaction force parameters in the improved social force model, achieving the purpose of improving the virtual group simulation effect. In another embodiment, a user motion-aware virtual group motion control method calculates the interaction force between the user avatar and the virtual Agent based on the improved social force model to optimize the collision avoidance strategy of the Agent. The interaction force parameters are optimized using the method of the present application, so that the virtual group presents a motion interaction mode between the virtual Agent and the user avatar that is more consistent with the motion interaction mode set by the interaction force.

[0158] The present application also proposes an improved social force model parameter optimization and calibration device, comprising:

[0159] An input module is configured to obtain a virtual group simulation target scenario and an evaluation index determined according to the target scenario;

[0160] A parameter setting module is configured to set the interaction force parameters in the improved social force model as decision variables and set the evaluation index as the objective function, and set the multi-objective particle swarm optimization algorithm parameters and the virtual group simulation experiment parameters for the virtual group simulation target scenario; ​

[0161] An optimization and calibration module is configured to complete the evaluation of the particles in the multi-objective particle swarm optimization algorithm through virtual group simulation experiments for the virtual group simulation target scenario, and obtain the calibration value of the interaction force parameters in the improved social force model under the optimized virtual group simulation target scenario. ​ ​

[0162] In another embodiment, the method can be implemented by a device, which can include corresponding modules for performing one or more steps in the above embodiments. Therefore, each step or several steps of the above embodiments can be performed by the corresponding modules, and the electronic device can include one or more of these modules. The modules can be one or more hardware modules specially configured to perform the corresponding steps, or implemented by a processor configured to perform the corresponding steps, or stored in a computer readable medium for implementation by a processor, or implemented by some combination.​​

[0163] In the above embodiments, the description of each embodiment is focused on, and the part not described in detail in a certain embodiment can be referred to the relevant description of other embodiments. Each technical feature of the above embodiments can be combined arbitrarily. In order to make the description simple, each technical feature in the above embodiments is not described in all possible combinations, however, as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the description.

[0164] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. An improved social force model parameter tuning calibration method, characterized in that, The method comprises the following steps: obtaining a virtual group simulation target scene and evaluation indexes determined according to the target scene; Improving interaction force parameter in social force model A, n, n', γ, λ For the decision variable, the evaluation index is the objective function, and the parameters of the multi-objective particle swarm optimization algorithm are set. setting virtual group simulation experiment parameters for the target scene; The evaluation of the particles in the multi-objective particle swarm optimization algorithm is completed through the virtual group simulation experiment of the target scene, and the calibrated value of the target scene under the optimization is obtained A, n, n', γ, λ . The evaluation indexes take the improvement of the collision avoidance capability in the virtual group motion interaction process as the optimization indexes of the improved social force model, and comprise virtual group motion interaction evaluation indexes and global motion interaction evaluation indexes, specifically three evaluation indexes of average collision times per capita AACC, average collision depth AACD and global collision score CS; Definition: Agent is a circle in two-dimensional plane, when the boundary of has some overlap, then it is considered that and have a collision; the overlap from the boundary of both sides until both sides completely leave the boundary of the other side is defined as a collision; the definition of and between the collision depth is the length of the overlap of the main body of both sides along the line connecting the centers of and , which is specifically defined as: ； wherein , are the positions of and the radii of , are the positions of and the radii of The average collision times per capita AACC represent the number of Agent-Agent collisions per capita, reflect the collision frequency in the Agent-Agent motion interaction process, and the unit is times, and the specific definition is represented as: ； wherein, representing the size of the virtual group, representing the number of collisions between and occurring during the entire simulation process; The collision depth is defined as the maximum collision depth in one collision process, the average collision depth AACD represents the average maximum collision depth of each Agent-Agent collision per capita, and reflects the severity of Agent-Agent collision in the spatial dimension, and the unit is meters, and the specific definition is represented as: ； wherein represents with the maximum collision depth at the occurrence of the kth collision; The collision score is defined as the sum of the collision depths of all ongoing collisions in the entire scene at each time step in the entire simulation process, and then the total time wheel number and the number of Agents in the group are averaged, and the specific definition of the global collision score CS is represented as: ； wherein, denotes the time step required for the entire simulation process, denotes the collision depth of the time step with the other object; The decision variables are represented by an array Param = { A, n, n', γ, λ } represents the calibration value of the interaction force parameter in the target scenario to be solved; for all particles in the multi-objective particle swarm optimization algorithm, at time t, the particle i Location in solution space x i ( t () represents a specific set of Param values ​​for particles. i After the current position is evaluated by the objective function, the particle is obtained. i The infeasibility and target value at the current location are given, with the target value represented by the array cost. i ( t )={ n 1 , n 2 , n 3}, of which n 1 , n 2 , n 3 corresponds to the results of particle position evaluation by three evaluation indicators: average collision count per person (AACC), average collision depth (AACD), and global collision score (CS); based on the decision variables and target values, determine and set the parameters of the multi-objective particle swarm optimization algorithm, including the decision space dimension, target space dimension, search range of the decision space, particle swarm size, maximum warehouse size, and maximum number of iterations; The virtual group simulation experiment uses the improved social force model, and the experiment parameters comprise an agent radius, a perception neighbor range, an agent expected speed and a simulation limit time length; The evaluation of the particles in the multi-objective particle swarm optimization algorithm is completed, and the optimized target scene under the target scene is obtained A, n, n', γ, λ The specific method of the calibration value of the target scene is: The multi-objective particle swarm optimization algorithm is used for parameter initialization of a particle swarm, a warehouse and an iteration number; a position of a particle in the particle swarm x i ( t ) an improved social force model with the position of a particle in the particle swarm as an interaction force parameter input and virtual crowd simulation experiment parameters set for the target scene, performing a virtual crowd simulation experiment on the target scene, returning an evaluation result of the target function on the current position of the particle through the simulation experiment, the evaluation result including the target value array cost i ( t )={ n 1 , n 2 , n 3} and infeasibility; According to the target value array of the particle and the infeasibility, the particle speed, the position, the domination attribute, the individual optimal pBest, the global optimal gBest in the multi-objective particle swarm optimization algorithm and the non-dominated solution in the warehouse are updated; the iteration update of the particle swarm and the evaluation of the particles in the particle swarm are completed according to the preset iteration number, and the optimal non-dominated solution set is output; selecting one solution from the optimal non-dominated solution set for the target scene under consideration A, n, n', γ, λ parameters.

2. The method of claim 1, wherein, The target scene is a virtual group motion interaction scene in a virtual simulation scene.

3. The method of claim 1, wherein, The unfeasibility of the particle means that when the simulation experiment exceeds the simulation limit time length and is not completed in the particle evaluation, the simulation program is interrupted and the unfeasibility of the particle is assigned a maximum value, indicating that the position represented by the particle A, n, n', γ, λ Parameter group is not available.

4. An improved social force model parameter tuning targeting device, characterized in that, The method comprises the following steps: An input module is used for obtaining a virtual group simulation target scene and evaluation indexes determined according to the target scene; A parameter setting module is configured to improve an interaction force parameter in a social force model A, n, n', γ, λ For the decision variable, the evaluation index is set as a target function, and a multi-objective particle swarm optimization algorithm parameter is set; and for the target scene, a virtual group simulation experiment parameter is set. The tuning and calibration module is configured to complete the evaluation of the particles in the multi-objective particle swarm optimization algorithm through a virtual group simulation experiment of the target scene, and obtain the calibrated value of the target scene under the optimization A, n, n', γ, λ . The evaluation indexes take the improvement of the collision avoidance capability in the virtual group motion interaction process as the optimization indexes of the improved social force model, and comprise virtual group motion interaction evaluation indexes and global motion interaction evaluation indexes, specifically three evaluation indexes of average collision times per capita AACC, average collision depth AACD and global collision score CS; Define Agent as a disk on a two-dimensional plane, when and When the boundaries overlap, it is considered that and A collision is defined as the process from the initial overlap of the boundaries of the colliding objects until they completely move away from each other's boundaries. and Collision depth between For the overlapping part of both main bodies along and The length of the line connecting the centers of the circles is specifically defined as follows: ； wherein , are respectively and the radius of , are respectively and the position of AACC, representing the average number of Agent-Agent collisions per Agent, reflects the collision frequency in the process of Agent-Agent motion interaction, and the unit is times, and the specific definition is represented as: ； wherein, representing the size of the virtual group, representing the number of collisions between and occurring during the entire simulation process; The collision depth is defined as the maximum collision depth in the process of collision, and the average collision depth AACD represents the average maximum collision depth of each Agent in each Agent-Agent collision, which reflects the severity of Agent-Agent collision in the spatial dimension, and the unit is meter, and the specific definition is represented as: ； wherein represents with maximum depth of collision at the occurrence of the kth collision; The collision score is defined as the sum of the collision depth of all ongoing collisions in the entire scene at each time step in the entire simulation process, and then the total time wheel number and the number of Agent in the group are averaged, and the specific calculation definition of the global collision score CS is represented as: ； wherein, denotes the time step required for the entire simulation process, denotes the collision depth at the time step and the time step. The decision variables are represented by an array Param = { A, n, n′, γ, λ } represents the calibration value of the interaction force parameter in the target scenario to be solved; for all particles in the multi-objective particle swarm optimization algorithm, at time t, the particle i Location in solution space x i ( t () represents a specific set of Param values ​​for particles. i After the current position is evaluated by the objective function, the particle is obtained. i The infeasibility and target value at the current location are given, with the target value represented by the array cost. i ( t )={ n 1 , n 2 , n 3}, of which n 1 , n 2 , n 3 corresponds to the results of particle position evaluation by three evaluation indicators: average collision count per person (AACC), average collision depth (AACD), and global collision score (CS); based on the decision variables and target values, determine and set the parameters of the multi-objective particle swarm optimization algorithm, including the decision space dimension, target space dimension, search range of the decision space, particle swarm size, maximum warehouse size, and maximum number of iterations; The virtual group simulation experiment uses an improved social force model, and the experimental parameters include the radius of the agent, the perception neighbor range, the desired speed of the agent, and the simulation limit time length; The evaluation of the particles in the multi-objective particle swarm optimization algorithm is completed, and the optimized target scene under the target scene is obtained A, n, n′, γ, λ The specific method of the calibration value of the target scene is: The multi-objective particle swarm optimization algorithm is used to initialize the particle swarm, the warehouse, and the iteration number; a position of a particle in the particle swarm x i ( t ) an improved social force model with the position of a particle in the particle swarm as an interaction force parameter input and virtual crowd simulation experiment parameters set for the target scene, performing a virtual crowd simulation experiment on the target scene, returning an evaluation result of the target function on the current position of the particle through the simulation experiment, the evaluation result including the target value array cost i ( t )={ n 1 , n 2 , n 3} and infeasibility; According to the target value array and the infeasibility of the particle, the particle velocity, the position, the dominance attribute, the individual optimal pBest, the global optimal gBest in the multi-objective particle swarm optimization algorithm, and the non-dominated solution in the warehouse are updated; the iteration number is completed according to the preset iteration number, the iteration update of the particle swarm and the evaluation of the particles in the particle swarm are completed, and the optimal non-dominated solution set is output; selecting one solution from the optimal non-dominated solution set for the target scene under consideration A, n, n′, γ, λ parameters.

5. An electronic device, comprising: It includes: A processor; A memory storing programs executable by the processor, when the programs are executed, the processor executes the method of any one of claims 1-4.

6. A storage medium, characterized by A memory storing programs executable by the processor, when the programs are executed, the processor executes the method of any one of claims 1-4.

Images