CN-122023644-A - Pedestrian behavior simulation scene construction method, system, terminal and storage medium

Abstract

The application discloses a method, a system, a terminal and a storage medium for constructing a pedestrian behavior simulation scene, and relates to the technical field of crowd simulation, wherein the method comprises the steps of obtaining a static three-dimensional virtual scene, and determining simulation environment data according to the static three-dimensional virtual scene, wherein the simulation environment data comprises a two-dimensional structured navigation map and at least one global reference track; acquiring state data corresponding to the virtual pedestrians, generating a model through trained track diffusion according to the state data and the simulation environment data, and acquiring predicted motion tracks corresponding to the virtual pedestrians; and controlling the virtual pedestrians to move in the static three-dimensional virtual scene according to the predicted motion trail. Therefore, the pedestrian motion trail closer to the real scene can be generated based on the pre-trained trail diffusion generation model, so that the motion trail of the virtual pedestrian is closer to the pedestrian motion trail in the real scene, the reality of the pedestrian behavior is improved, and the reality of the constructed pedestrian behavior simulation scene is improved.

Inventors

• WANG JIANKUN
• XIA BINGYI

Assignees

• 南方科技大学

Dates

Publication Date

20260512

Application Date

20251230

Claims (10)

1. 1. A method for constructing a pedestrian behavior simulation scene, the method comprising: Acquiring a static three-dimensional virtual scene, and determining simulation environment data according to the static three-dimensional virtual scene, wherein the simulation environment data comprises a two-dimensional structured navigation map and at least one global reference track; acquiring state data corresponding to a virtual pedestrian, and acquiring a predicted motion track corresponding to the virtual pedestrian through a trained track diffusion generation model according to the state data and the simulation environment data; and controlling the virtual pedestrians to move in the static three-dimensional virtual scene according to the predicted motion trail so as to obtain a pedestrian behavior simulation scene.
2. 2. The method for constructing a pedestrian behavior simulation scene according to claim 1, wherein the acquiring a static three-dimensional virtual scene, determining simulation environment data according to the static three-dimensional virtual scene, comprises: Importing a general scene description format file containing geometric and semantic information to construct the static three-dimensional virtual scene; carrying out space analysis on the static three-dimensional virtual scene, and generating a two-dimensional occupied grid map representing a walkable region as the two-dimensional structured navigation map; and determining the global reference track corresponding to the virtual pedestrian through a path planning algorithm according to the two-dimensional occupied grid map.
3. 3. The method for constructing a pedestrian behavior simulation scene according to claim 1, wherein the obtaining the state data corresponding to the virtual pedestrian, and according to the state data and the simulation environment data, obtaining the predicted motion trail corresponding to the virtual pedestrian through a trained trail diffusion generation model includes: Inputting the state data and the simulation environment data into a trained data-driven track diffusion generation model; And obtaining a predicted motion track output by the data-driven track diffusion generation model, wherein the predicted motion track is a future track of the virtual pedestrian predicted by the track diffusion generation model.
4. 4. The pedestrian behavior simulation scene construction method according to claim 3, wherein the state data corresponding to the virtual pedestrian includes: The current state of the virtual pedestrian, the historical track of the virtual pedestrian and the track of the neighbor pedestrian; The predicted motion trail comprises coordinates of two-dimensional trail points of the virtual pedestrian at a plurality of moments in the future, and corresponding course angles and speeds at the two-dimensional trail points.
5. 5. The pedestrian behavior simulation scene construction method according to claim 1, wherein the controlling the virtual pedestrian to move in the static three-dimensional virtual scene according to the predicted motion trajectory includes: generating a physical control signal for driving the whole body joints of the virtual pedestrians through a reinforcement learning controller based on the prior training of the antagonistic movement according to the predicted movement track and the state data corresponding to the virtual pedestrians; and controlling the virtual pedestrians to move in the static three-dimensional virtual scene according to the physical control signals.
6. 6. The pedestrian behavior simulation scene construction method according to claim 5, wherein the virtual pedestrian is a pedestrian agent configured based on a skin multi-person linear model; The pedestrian agent is provided with a hinge body capable of driving a joint; The controlling the virtual pedestrian to move in the static three-dimensional virtual scene according to the physical control signal comprises the following steps: and controlling the hinge body of the pedestrian agent according to the physical control signal so as to enable the virtual pedestrian to move along the predicted movement track in the static three-dimensional virtual scene.
7. 7. The pedestrian behavior simulation scene construction method according to any one of claims 1 to 6, characterized in that the method further comprises: Acquiring scene data corresponding to the static three-dimensional virtual scene and action change sequence data corresponding to the virtual pedestrian in the static three-dimensional virtual scene through a robot arranged in the static three-dimensional virtual scene; And constructing navigation scene data according to the scene data and the action change sequence data.
8. 8. A pedestrian behavior simulation scene construction system, the system comprising: the data processing module is used for acquiring a static three-dimensional virtual scene and determining simulation environment data according to the static three-dimensional virtual scene, wherein the simulation environment data comprises a two-dimensional structured navigation map and at least one global reference track; The track prediction module is used for acquiring state data corresponding to the virtual pedestrians, generating a model through trained track diffusion according to the state data and the simulation environment data, and acquiring predicted motion tracks corresponding to the virtual pedestrians; And the motion control module is used for controlling the virtual pedestrians to move in the static three-dimensional virtual scene according to the predicted motion trail so as to obtain a pedestrian behavior simulation scene.
9. 9. A terminal comprising a memory, a processor and a computer program stored on the memory and executable on the processor, which when executed by the processor, implements the steps of the pedestrian behavior simulation scene construction method of any one of claims 1 to 7.
10. 10. A computer-readable storage medium, characterized in that the computer-readable storage medium has stored thereon a computer program which, when executed by a processor, implements the steps of the pedestrian behavior simulation scene construction method according to any one of claims 1 to 7.

Description

Pedestrian behavior simulation scene construction method, system, terminal and storage medium Technical Field The application relates to the technical field of crowd simulation, in particular to a method, a system, a terminal and a storage medium for constructing a pedestrian behavior simulation scene. Background With the development of science and technology, the man-machine co-fusion is increasingly widely applied. In a man-machine co-fusion environment, the autonomous navigation of a robot needs to consider both safety and natural man-machine interaction, and complicated and various navigation scene simulation is a key for training a reliable navigation strategy and evaluating the performance of an algorithm. In terms of pedestrian behavior simulation scene construction, the prior art generally constructs a three-dimensional virtual scene, and then generates pedestrian behavior in the three-dimensional virtual scene according to a predefined fixed rule (e.g., avoid collisions, follow-up population). The problem in the prior art is that the generation of the pedestrian behavior based on the predefined fixed rule is unfavorable for improving the authenticity of the pedestrian behavior, thereby being unfavorable for improving the authenticity of the pedestrian behavior simulation scene. Accordingly, the related art has yet to be improved and developed. Disclosure of Invention The application mainly aims to provide a method, a system, a terminal and a storage medium for constructing a pedestrian behavior simulation scene, and aims to solve the technical problems that the generation of pedestrian behaviors based on a predefined fixed rule in the related technology is unfavorable for improving the authenticity of the pedestrian behaviors, so that the authenticity of the pedestrian behavior simulation scene is unfavorable for improving. In order to achieve the above object, a first aspect of the present application provides a method for constructing a pedestrian behavior simulation scene, where the method includes: acquiring a static three-dimensional virtual scene, and determining simulation environment data according to the static three-dimensional virtual scene, wherein the simulation environment data comprises a two-dimensional structured navigation map and at least one global reference track; Acquiring state data corresponding to a virtual pedestrian, and acquiring a predicted motion track corresponding to the virtual pedestrian through a trained track diffusion generation model according to the state data and the simulation environment data; And controlling the virtual pedestrians to move in the static three-dimensional virtual scene according to the predicted motion trail so as to obtain a pedestrian behavior simulation scene. Optionally, the acquiring the static three-dimensional virtual scene, determining the simulation environment data according to the static three-dimensional virtual scene, includes: importing a general scene description format file containing geometric and semantic information to construct the static three-dimensional virtual scene; Carrying out space analysis on the static three-dimensional virtual scene to generate a two-dimensional occupied grid map representing a walkable region as the two-dimensional structured navigation map; and determining the global reference track corresponding to the virtual pedestrian through a path planning algorithm according to the two-dimensional occupation grid map. Optionally, the obtaining the state data corresponding to the virtual pedestrian, according to the state data and the simulation environment data, through a trained track diffusion generating model, obtaining a predicted motion track corresponding to the virtual pedestrian includes: inputting the state data and the simulation environment data into a trained data-driven track diffusion generation model; and obtaining a predicted motion track output by the data-driven track diffusion generation model, wherein the predicted motion track is a future track of the virtual pedestrian predicted by the track diffusion generation model. Optionally, the state data corresponding to the virtual pedestrian includes: the current state of the virtual pedestrian, the history track of the virtual pedestrian and the track of the neighbor pedestrian; The predicted motion trail comprises coordinates of two-dimensional trail points of the virtual pedestrian at a plurality of future moments, and corresponding course angles and speeds at the two-dimensional trail points. Optionally, controlling the motion of the virtual pedestrian in the static three-dimensional virtual scene according to the predicted motion trail includes: Generating a physical control signal for driving the joints of the whole body of the virtual pedestrian through a reinforcement learning controller based on the prior training of the antagonistic movement according to the predicted movement track and the state data corresponding to the virt