Pedestrian cluster modeling method and system based on behavioral consistency and repulsion tension

Abstract

The application relates to the technical field of automatic driving and discloses a pedestrian cluster modeling method and system based on behavior consistency and repulsion tension, which comprises the following steps: obtaining pedestrian trajectory data and preprocessing the pedestrian trajectory data, constructing a pedestrian individual microscopic behavior model based on a social force model to obtain the acceleration of the pedestrian, time-integrating the acceleration of the pedestrian to obtain a predicted trajectory of the pedestrian, constructing a loss function in combination with the real trajectory and the predicted trajectory of the pedestrian and optimizing the parameters of the pedestrian individual microscopic behavior model, obtaining a final predicted trajectory according to the optimized pedestrian individual microscopic behavior model, clustering and analyzing the turning angle of the final predicted trajectory to obtain the dynamic behavior of a pedestrian cluster, adjusting the average width of the pedestrian cluster based on behavior consistency, adjusting the density line length of the pedestrian cluster based on repulsion tension, and obtaining a final pedestrian cluster characteristic model. The application can realize dynamic modeling of the internal behavior mode of the pedestrian cluster and improve simulation accuracy.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a pedestrian cluster modeling method and system based on behavioral consistency and repulsion tension. Background Technology

[0002] Autonomous driving systems are crucial for driverless vehicles. As the complexity of autonomous driving and automotive systems continues to increase, the safety requirements for these systems are also rising. During autonomous driving, the movement of pedestrians, especially pedestrian groups in complex traffic scenarios, is a critical process that needs to be carefully considered.

[0003] To improve the safety of autonomous driving, accurate simulation of pedestrian group movement is necessary. Current technologies typically use physics-based models (such as social force models) to reconstruct pedestrian movement, which can reveal, to some extent, the microscopic movement mechanisms of pedestrians influenced by factors such as target guidance, group exclusion, and obstacle avoidance, thus reproducing typical pedestrian group phenomena. However, this method also has some limitations:

[0004] 1. Traditional social force models have fixed parameters, static structural representations, and lack feedback on pedestrian behavior, which limits their predictive ability and affects the simulation accuracy of pedestrian cluster movement.

[0005] 2. Existing macro-modeling methods often use static linear or empirical regression to map the number of people to structural indicators such as width and length, ignoring the moderating effect of internal group behavior patterns (such as consistency and tension) on the overall structure, making it difficult to simulate the changing trends of pedestrian cluster behavior in dynamic traffic scenarios. Summary of the Invention

[0006] Therefore, the technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide a pedestrian cluster modeling method and system based on behavioral consistency and repulsion tension, which can realize dynamic modeling of the internal behavioral patterns of pedestrian clusters and improve simulation accuracy.

[0007] To address the aforementioned technical problems, this invention provides a pedestrian cluster modeling method based on behavioral consistency and repulsion tension, comprising: Pedestrian trajectory data is acquired and preprocessed. A micro-behavioral model of individual pedestrians is constructed based on a social force model to obtain the pedestrian's acceleration. The pedestrian's acceleration is then integrated over time to obtain the predicted trajectory of the pedestrian. A loss function is constructed by combining the pedestrian's actual trajectory and the predicted trajectory, and the parameters of the pedestrian individual micro-behavior model are optimized. The final predicted trajectory is obtained based on the optimized pedestrian individual micro-behavior model. Clustering and turning angle analysis are performed on the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster; The average width of the pedestrian cluster is adjusted based on behavioral consistency, and the density line length of the pedestrian cluster is adjusted based on repulsive tension, resulting in the final pedestrian cluster feature model.

[0008] Furthermore, the loss function is: , In the formula, Represents the loss function. This represents the predicted trajectory of the i-th pedestrian at time t. The parameter vector to be optimized in the pedestrian individual micro-behavior model. Let N represent the actual observed trajectory of the i-th pedestrian at time t, where N is the number of pedestrians and T is the observation time. The length of the vector; When optimizing the parameters of the pedestrian individual micro-behavior model, the optimization objective is to minimize the loss function.

[0009] Furthermore, when optimizing the parameters of the pedestrian individual micro-behavior model, the optimization process includes: An initial particle swarm is constructed in the parameter space of the pedestrian individual micro-behavior model, where each particle represents a set of candidate parameters, and the optimal position of each particle, the individual, and the global optimal position are initialized. For each particle, construct a pedestrian individual micro-behavior model and use the current corresponding loss function value as the fitness function value. When the fitness function value of the i-th particle is less than the fitness function value of the individual's optimal position, let the position of the i-th particle be the updated individual optimal position. When the fitness function value of the i-th particle is less than the fitness function value of the global optimal position, let the position of the i-th particle be the updated global optimal position. The optimization ends when the iteration termination condition is met, and the parameters obtained at this point are used as the parameters of the optimized pedestrian individual micro-behavior model.

[0010] Furthermore, the clustering and turning angle analysis of the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster includes: Set a spatial proximity threshold and a speed difference threshold. When the Euclidean distance between two pedestrians is less than or equal to the spatial proximity threshold and the speed direction difference is less than the speed difference threshold, it is determined that the two pedestrians belong to the same pedestrian cluster. For each pedestrian cluster, the principal direction vector at time t is calculated based on the time difference between the centroids. Define the time step as the time difference, and calculate the turning angle of the pedestrian cluster at the current moment as follows: , In the formula, This represents the turning angle of the k-th pedestrian group at time t. Let be the main direction vector of the k-th pedestrian cluster at time t. Let k be the time difference, k = 1, 2, ..., K, and K be the number of pedestrian groups. The length of the vector; according to Different changing trends can be used to determine the different dynamic behaviors of pedestrian clusters.

[0011] Furthermore, the aforementioned The calculation method is as follows: , In the formula, This represents the centroid position of the k-th pedestrian cluster at time t; The calculation method is as follows: , In the formula, Let be the position vector of the i-th pedestrian at time t. This represents the k-th pedestrian cluster. This represents the number of pedestrians in the k-th pedestrian cluster.

[0012] Furthermore, the average width of the pedestrian cluster adjusted based on behavioral consistency is: , In the formula, wide represents the average width of the pedestrian cluster adjusted based on behavioral consistency. , The width is the baseline fitting coefficient, and N1 is the number of pedestrians in the current pedestrian cluster. This is the adjustment factor for width compression based on behavioral consistency. This is the speed consistency coefficient.

[0013] Furthermore, the aforementioned The calculation method is as follows: , In the formula, The average speed of pedestrians in the current cluster. Let be the speed of the i-th pedestrian in the current cluster. is the length of the vector.

[0014] Furthermore, the density line length of the pedestrian cluster adjusted based on repulsive tension is: , In the formula, Long is the density line length of the pedestrian cluster based on repulsive tension adjustment. , The length baseline fitting coefficients are given, and N1 represents the number of pedestrians in the current pedestrian cluster. The coefficient for adjusting the effect of tension on length extension is used to exclude tension. This represents the average repulsive tension.

[0015] Furthermore, the aforementioned The calculation method is as follows: , In the formula, Let be the inter-individual repulsive force experienced by the i-th pedestrian.

[0016] This invention also provides a pedestrian cluster modeling system based on behavioral consistency and repulsion tension, comprising: The data acquisition module is used to acquire pedestrian trajectory data and perform preprocessing. The micro-modeling module is used to construct a micro-behavioral model of individual pedestrians based on the social force model to obtain the pedestrian's acceleration. The micro-model optimization module is used to perform time integration on the pedestrian's acceleration to obtain the pedestrian's predicted trajectory, and to construct a loss function by combining the pedestrian's actual trajectory and the predicted trajectory to optimize the parameters of the individual pedestrian micro-behavior model; The dynamic behavior analysis module is used to obtain the final predicted trajectory based on the optimized individual pedestrian micro-behavior model, and to perform clustering and turning angle analysis on the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster. The cluster scale adjustment module is used to adjust the average width of pedestrian clusters based on behavioral consistency and the density line length of pedestrian clusters based on repulsive tension, so as to obtain the final pedestrian cluster feature model.

[0017] Compared with the prior art, the above-described technical solution of the present invention has the following advantages: This invention optimizes the micro-behavioral model of individual pedestrians based on their actual behavioral trajectories. By utilizing the dynamic feedback of pedestrian behavior to optimize the parameters of the social force model, it improves the predictive ability of the micro-behavioral model of individual pedestrians. Based on the optimized trajectory, it identifies pedestrian clusters and analyzes their dynamic behavior. By combining behavioral consistency and repulsion tension to dynamically adjust the scale of pedestrian clusters, it can achieve dynamic modeling of the internal behavioral patterns of pedestrian clusters while maintaining the simulation accuracy of individual pedestrians, thereby improving the simulation accuracy. Attached Figure Description

[0018] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of a method in a preferred embodiment of the present invention. Detailed Implementation

[0019] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0020] Reference Figure 1 As shown, this invention discloses a pedestrian cluster modeling method based on behavioral consistency and repulsion tension, comprising the following steps: S1: Acquire pedestrian trajectory data and perform preprocessing.

[0021] S1-1: Acquire pedestrian trajectory data over continuous time periods in traffic scenarios using equipment such as video surveillance, LiDAR, or millimeter-wave radar. Construct a trajectory set T={(x i (t),y i (t),v i x(t)}, i=1,2,…,N,t∈[0,T]; where i represents the pedestrian number, t represents the time frame, and x i (t) represents the x-coordinate position of the i-th pedestrian at time t, and y-coordinate... i (t) represents the y-coordinate of the i-th pedestrian at time t, v i (t) represents the speed of the i-th pedestrian at time t, N is the number of pedestrians, and T is the observation time.

[0022] S1-2: To address issues such as occlusion and frame loss in real-world scenarios, a temporal interpolation method based on Bessel interpolation or Kalman filtering is used to complete the data in the trajectory set, ensuring the continuity of the trajectory in time and space, and providing high-quality input data for subsequent modeling.

[0023] S2: Based on the social force model, construct a micro-behavioral model of the pedestrian to obtain the pedestrian's acceleration, and perform time integration on the pedestrian's acceleration to obtain the pedestrian's predicted trajectory.

[0024] S2-1: For individual pedestrians, a social force model is used to characterize their behavioral decision-making process. The pedestrian's acceleration is caused by the combined action of three forces, and the total resultant force acting on the pedestrian is: , In the formula, Let the total net force on the i-th pedestrian be , Let be the self-driving force acting on the i-th pedestrian as they approach the target. Let be the inter-individual repulsive force experienced by the i-th pedestrian. Let be the repulsive force exerted by the obstacle on the i-th pedestrian.

[0025] The method for calculating the self-driving force for target approach is as follows: , In the formula, For the quality of the i-th pedestrian, Let be the expected velocity along the line connecting the current position of the i-th pedestrian to the target point. Let be the direction vector pointing from the position of the i-th pedestrian at the current moment to the target point. Let τ represent the velocity vector of the i-th pedestrian in the direction that is not connected to the target point, and let τ be the basic parameter value of the social force model (τ=0.5s in this embodiment). This indicates that the i-th pedestrian starts from the current speed. Towards its desired speed A trend of gradual adjustment.

[0026] The method for calculating the repulsive force between individuals is as follows: , In the formula, For repulsion strength parameters, To exclude the decay parameter, Let i be the minimum safe distance that is expected to be maintained between the i-th pedestrian and the j-th pedestrian. Let i be the actual distance between the i-th pedestrian and the j-th pedestrian. Let be the unit vector pointing from the j-th pedestrian to the i-th pedestrian.

[0027] The method for calculating the repulsive force of an obstacle is as follows: , In the formula, Let be the repulsion strength parameter of the k-th obstacle. Let be the repulsion attenuation coefficient of the k-th obstacle. The minimum safe distance (psychological or physical safe distance, set to 0.3m in this embodiment) is the expected distance between the i-th pedestrian and the k-th obstacle. Let i be the actual distance between the i-th pedestrian and the k-th obstacle. Let be the unit vector pointing from the k-th obstacle to the i-th pedestrian.

[0028] S2-2: The acceleration is calculated based on the total net force acting on the pedestrian: , In the formula, Let be the acceleration of the i-th pedestrian.

[0029] S2-3: Integrate the pedestrian's acceleration over time to obtain the pedestrian's predicted trajectory, denoted as... , Let Z represent the predicted trajectory of the i-th pedestrian at time t, and Z be the parameter vector to be optimized in the individual pedestrian micro-behavior model. .

[0030] S3: Construct a loss function by combining the pedestrian's actual trajectory and the predicted trajectory, optimize the parameters of the pedestrian individual micro-behavior model using the particle swarm optimization algorithm, and obtain the final predicted trajectory based on the optimized pedestrian individual micro-behavior model.

[0031] S3-1: The loss function is constructed based on the spatiotemporal distance error between the model-predicted trajectory and the actual observed trajectory, specifically as follows: , In the formula, Represents the loss function. This represents the predicted trajectory of the i-th pedestrian at time t. The parameter vector to be optimized in the pedestrian individual micro-behavior model. Let N represent the actual observed trajectory of the i-th pedestrian at time t, where N is the number of pedestrians and T is the observation time. is the length of the vector.

[0032] S3-2: When using the particle swarm optimization algorithm to optimize the parameters of the pedestrian individual micro-behavior model, the optimization objective is to minimize the loss function. The key parameters in the social force model are solved in reverse by the particle swarm optimization algorithm, so that the simulated trajectory is as close as possible to the real trajectory, thereby improving the model's scene adaptability and behavior interpretation ability.

[0033] The optimization process includes: S3-2-1: Initialize the particle swarm. Construct an initial particle swarm within the parameter space of the pedestrian individual micro-behavior model. Each particle represents a set of candidate parameters (i.e., a parameter vector). Let the number of particles be 𝑃 and the dimension be 𝐷=len(𝜃). Randomly initialize each particle, ensuring that the initial values ​​satisfy the parameter boundary range. Simultaneously initialize the particle velocity, individual optimal position, and global optimal position.

[0034] S3-2-2: Calculate the fitness function. For each particle (i.e., the parameters of each candidate group), construct a micro-behavioral model of the pedestrian individual, and use the current corresponding loss function value as the fitness function value. , Let be the fitness function value of the i-th particle.

[0035] When the fitness function value of the i-th particle is less than the fitness function value of the individual's optimal position, the position of the i-th particle is set as the updated individual's optimal position; when the fitness function value of the i-th particle is less than the fitness function value of the global optimal position, the position of the i-th particle is set as the updated global optimal position.

[0036] S3-2-3: The method for updating particle velocity and position is as follows: , ; In the formula, Let be the velocity of the i-th particle in the (n+1)th iteration. Inertial weights are used to balance local or global searches. Let be the velocity of the i-th particle in the nth iteration. , As a learning factor, , To be uniformly random numbers following the range [0,1], Let be the optimal position for the i-th particle. Let be the position of the i-th particle in the nth iteration. To be the globally optimal position The position of the i-th particle in the (n+1)-th iteration; all parameter values ​​are kept within the upper and lower limits of the preset parameters, and are truncated or bounced if they exceed the limits.

[0037] S3-2-4: Iterate until convergence. Repeat S3-2-2 and S3-2-3 until the iteration termination condition is met (the maximum number of iterations or the optimal fitness function value converges), at which point the optimization ends, and the parameters obtained at this point are used as the parameters of the optimized pedestrian individual micro-behavior model.

[0038] S3-3: The final predicted trajectory is obtained based on the parameters of the optimized pedestrian individual micro-behavior model.

[0039] S4: Cluster and turn angle analysis are performed on the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster.

[0040] S4-1: This invention uses a density-based spatial clustering algorithm (DBSCAN) to perform cluster analysis on trajectory points in each frame to identify pedestrian clusters with consistent behavior. A spatial proximity threshold and a velocity difference threshold are set. When the Euclidean distance between two pedestrians is less than or equal to the spatial proximity threshold and the velocity direction difference is less than the velocity difference threshold, the two pedestrians are determined to belong to the same pedestrian cluster.

[0041] S4-2: For each pedestrian cluster, the principal direction vector at time t is calculated based on the time difference between the centroids as follows: , In the formula, Let be the main direction vector of the k-th pedestrian cluster at time t. Let represent the centroid position of the k-th pedestrian cluster at time t.

[0042] The calculation method is as follows: , In the formula, Let be the position vector of the i-th pedestrian at time t. This represents the k-th pedestrian cluster. This represents the number of pedestrians in the k-th pedestrian cluster.

[0043] S4-3: Define the time step as the time difference, and calculate the turning angle of the pedestrian cluster at the current moment as follows: , In the formula, This represents the turning angle of the k-th pedestrian group at time t. Let k be the time difference, k = 1, 2, ..., K, and K be the number of pedestrian groups. is the length of the vector.

[0044] S4-4: According to Different trends in pedestrian behavior can be used to determine the different dynamic behaviors of pedestrian groups. For example, if... Continuous growth indicates that the kth pedestrian cluster exhibits dynamic behaviors such as "turning" or "changing direction".

[0045] S5: Adjust the average width of the pedestrian cluster based on behavioral consistency, adjust the density line length of the pedestrian cluster based on repulsive tension, and calculate the density of the pedestrian cluster to obtain the final pedestrian cluster feature model.

[0046] Unlike traditional linear fitting structural models based on the number of people, this invention proposes a macroscopic formula model enhanced with behavioral feedback. It adds two dynamic feedback factors—behavioral consistency and repulsion tension—to the traditional linear structural formula based on the number of people, thereby improving the model's responsiveness and interpretability to changes in cluster structure. First, a macroscopic dynamic parameter model based on three parameters is constructed to quantitatively describe the distribution characteristics of crowd movement: average linewidth, density linelength, and group density.

[0047] S5-1: The average width of the pedestrian cluster adjusted based on behavioral consistency is: , In the formula, wide represents the average width of the pedestrian cluster adjusted based on behavioral consistency. , The width is the baseline fitting coefficient, and N1 is the number of pedestrians in the current pedestrian cluster. This is the adjustment coefficient for width compression based on behavioral consistency. The setting rule is: when the pedestrian direction is consistent at the same height, the cluster tends to be narrower and longer, and the width is compressed; if the pedestrian direction is dispersed, the width is increased. For speed consistency coefficient, , The closer to 1, the more consistent the pedestrians' directions.

[0048] The calculation method is as follows: , In the formula, The average speed of pedestrians in the current cluster. Let be the speed of the i-th pedestrian in the current cluster. .

[0049] S5-2: The density line length of the pedestrian cluster based on repulsive tension adjustment is: , In the formula, Long is the density line length of the pedestrian cluster based on repulsive tension adjustment. , The length baseline fitting coefficients are given, and N1 represents the number of pedestrians in the current pedestrian cluster. This is the adjustment coefficient for the effect of repulsive tension on length expansion; The setting rule is: the stronger the repulsive tension, the farther apart pedestrians are, and the pedestrian cluster extends in a longer shape in the main direction. The average repulsive tension reflects the tension of pedestrian clusters.

[0050] In this embodiment, the values ​​of the width reference fitting coefficient, the length reference fitting coefficient, the adjustment coefficient of behavior consistency on width compression, and the adjustment coefficient of repulsive tension on length expansion are adjusted according to actual conditions.

[0051] The calculation method is as follows: , In the formula, Let be the inter-individual repulsive force experienced by the i-th pedestrian.

[0052] S5-3: The population density is calculated as follows: , In the formula, ρ is the group density in the current pedestrian cluster.

[0053] S5-4: The final pedestrian cluster feature model is represented as follows: F(t,k)={t,k,Long,wide, ρ, , ,N1}, In the formula, F(t,k) represents the pedestrian cluster feature model at time t for the k-th pedestrian cluster.

[0054] The output of the pedestrian cluster feature model is represented in the form of structured feature vectors, which can accurately characterize the spatial evolution of the cluster in a time series. When applied to high-level semantic recognition, it can support systematic analysis tasks such as flow trend prediction, abnormal clustering warning, and traffic conflict detection.

[0055] This invention also discloses a pedestrian cluster modeling system based on behavioral consistency and repulsion tension, comprising: The data acquisition module is used to acquire pedestrian trajectory data and perform preprocessing. The micro-modeling module is used to construct a micro-behavioral model of individual pedestrians based on the social force model to obtain the pedestrian's acceleration. The micro-model optimization module is used to perform time integration on the pedestrian's acceleration to obtain the pedestrian's predicted trajectory, and to construct a loss function by combining the pedestrian's actual trajectory and the predicted trajectory to optimize the parameters of the individual pedestrian micro-behavior model; The dynamic behavior analysis module is used to obtain the final predicted trajectory based on the optimized individual pedestrian micro-behavior model, and to perform clustering and turning angle analysis on the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster. The cluster scale adjustment module is used to adjust the average width of pedestrian clusters based on behavioral consistency and the density line length of pedestrian clusters based on repulsive tension, so as to obtain the final pedestrian cluster feature model.

[0056] The present invention also discloses a computer-readable storage medium storing a computer program that, when executed by a processor, implements a pedestrian cluster modeling method based on behavioral consistency and repulsion tension.

[0057] The present invention also discloses an apparatus including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a pedestrian cluster modeling method based on behavioral consistency and repulsion tension.

[0058] This invention enables high-precision modeling and quantitative representation of the structural state and behavioral trends of pedestrian clusters in complex traffic scenarios, possessing significant practical application value. Compared with existing technologies, the advantages of this invention are: 1. This invention optimizes the individual micro-behavioral model of pedestrians based on their actual behavioral trajectories. By utilizing the dynamic feedback of pedestrian behavior to optimize the parameters of the social force model, it improves the predictive ability of the individual micro-behavioral model of pedestrians.

[0059] 2. Based on the optimized trajectory, pedestrian clusters are identified and their dynamic behavior is analyzed (extracting the main direction and turning trend). By combining behavioral consistency and repulsion tension to dynamically adjust the scale of the pedestrian cluster (average width and density line length), dynamic modeling of the internal behavioral patterns of the pedestrian cluster, such as structure, density, direction, and trend, can be achieved while maintaining the simulation accuracy of individual pedestrians, thereby improving the simulation accuracy.

[0060] To further demonstrate the beneficial effects of this invention, a complex pedestrian-vehicle mixed traffic scenario was selected for simulation experiments in this embodiment. This scenario includes pedestrian behaviors such as crossing the street, avoiding vehicles, and turning. In the simulation experiments, static linear regression, traditional social force models, and the method of this invention were used to simulate pedestrian clusters, respectively. The average structural error (ASE) was used to measure the deviation between the predicted pedestrian cluster scale (width and length) and the actual trajectory.

[0061] When using static linear regression, the average error is 0.84 meters because it cannot be dynamically adjusted according to the degree of crowding. When using a traditional social force model, the error increases when the crowd turns due to fixed parameters, with an average error of 0.76 meters. Using the method of this invention, the average error is reduced to 0.66 meters. Therefore, compared to existing methods, this invention improves the simulation accuracy of pedestrian clusters by 21.4%, achieving a high-fidelity, dynamically reproducible simulation effect.

[0062] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0063] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0064] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0065] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0066] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A pedestrian cluster modeling method based on behavioral consistency and repulsion tension, characterized in that, include: Pedestrian trajectory data is acquired and preprocessed. A micro-behavioral model of individual pedestrians is constructed based on a social force model to obtain the pedestrian's acceleration. The pedestrian's acceleration is then integrated over time to obtain the predicted trajectory of the pedestrian. A loss function is constructed by combining the pedestrian's actual trajectory and the predicted trajectory, and the parameters of the pedestrian individual micro-behavior model are optimized. The final predicted trajectory is obtained based on the optimized pedestrian individual micro-behavior model. Clustering and turning angle analysis are performed on the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster; The average width of the pedestrian cluster is adjusted based on behavioral consistency, and the density line length of the pedestrian cluster is adjusted based on repulsive tension, resulting in the final pedestrian cluster feature model.

2. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to claim 1, characterized in that, The loss function is: ， In the formula, Represents the loss function. This represents the predicted trajectory of the i-th pedestrian at time t. The parameter vector to be optimized in the pedestrian individual micro-behavior model. Let N represent the actual observed trajectory of the i-th pedestrian at time t, where N is the number of pedestrians and T is the observation time. The length of the vector; When optimizing the parameters of the pedestrian individual micro-behavior model, the optimization objective is to minimize the loss function.

3. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to claim 2, characterized in that: When optimizing the parameters of the pedestrian individual micro-behavior model, the optimization process includes: An initial particle swarm is constructed in the parameter space of the pedestrian individual micro-behavior model, where each particle represents a set of candidate parameters, and the optimal position of each particle, the individual, and the global optimal position are initialized. For each particle, construct a pedestrian individual micro-behavior model and use the current corresponding loss function value as the fitness function value. When the fitness function value of the i-th particle is less than the fitness function value of the individual's optimal position, let the position of the i-th particle be the updated individual optimal position. When the fitness function value of the i-th particle is less than the fitness function value of the global optimal position, let the position of the i-th particle be the updated global optimal position. The optimization ends when the iteration termination condition is met, and the parameters obtained at this point are used as the parameters of the optimized pedestrian individual micro-behavior model.

4. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to claim 1, characterized in that: The process of clustering and turning angle analysis on the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster includes: Set a spatial proximity threshold and a speed difference threshold. When the Euclidean distance between two pedestrians is less than or equal to the spatial proximity threshold and the speed direction difference is less than the speed difference threshold, it is determined that the two pedestrians belong to the same pedestrian cluster. For each pedestrian cluster, the principal direction vector at time t is calculated based on the time difference between the centroids. Define the time step as the time difference, and calculate the turning angle of the pedestrian cluster at the current moment as follows: ， In the formula, This represents the turning angle of the k-th pedestrian group at time t. Let be the main direction vector of the k-th pedestrian cluster at time t. Let k be the time difference, k = 1, 2, ..., K, and K be the number of pedestrian groups. The length of the vector; according to Different changing trends can be used to determine the different dynamic behaviors of pedestrian clusters.

5. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to claim 4, characterized in that, The The calculation method is as follows: ， In the formula, This represents the centroid position of the k-th pedestrian cluster at time t; The calculation method is as follows: ， In the formula, Let be the position vector of the i-th pedestrian at time t. This represents the k-th pedestrian cluster. This represents the number of pedestrians in the k-th pedestrian cluster.

6. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to claim 1, characterized in that: The average width of the pedestrian cluster adjusted based on behavioral consistency is: ， In the formula, wide represents the average width of the pedestrian cluster adjusted based on behavioral consistency. , The width is the baseline fitting coefficient, and N1 is the number of pedestrians in the current pedestrian cluster. This is the adjustment factor for width compression based on behavioral consistency. This is the speed consistency coefficient.

7. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to claim 6, characterized in that: The The calculation method is as follows: ， In the formula, The average speed of pedestrians in the current cluster. Let be the speed of the i-th pedestrian in the current cluster. is the length of the vector.

8. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to any one of claims 1-7, characterized in that: The density line length of the pedestrian cluster based on repulsive tension adjustment is: ， In the formula, Long is the density line length of the pedestrian cluster based on repulsive tension adjustment. , The length baseline fitting coefficients are given, and N1 represents the number of pedestrians in the current pedestrian cluster. The coefficient for adjusting the effect of tension on length extension is used to exclude tension. This represents the average repulsive tension.

9. The pedestrian cluster modeling method based on behavioral consistency and repulsion tension according to claim 8, characterized in that: The The calculation method is as follows: ， In the formula, Let be the inter-individual repulsive force experienced by the i-th pedestrian.

10. A pedestrian cluster modeling system based on behavioral consistency and repulsion tension, characterized in that, include: The data acquisition module is used to acquire pedestrian trajectory data and perform preprocessing. The micro-modeling module is used to construct a micro-behavioral model of individual pedestrians based on the social force model to obtain the pedestrian's acceleration. The micro-model optimization module is used to perform time integration on the pedestrian's acceleration to obtain the pedestrian's predicted trajectory, and to construct a loss function by combining the pedestrian's actual trajectory and the predicted trajectory to optimize the parameters of the individual pedestrian micro-behavior model; The dynamic behavior analysis module is used to obtain the final predicted trajectory based on the optimized individual pedestrian micro-behavior model, and to perform clustering and turning angle analysis on the final predicted trajectory to obtain the dynamic behavior of the pedestrian cluster. The cluster scale adjustment module is used to adjust the average width of pedestrian clusters based on behavioral consistency and the density line length of pedestrian clusters based on repulsive tension, so as to obtain the final pedestrian cluster feature model.

Images