WebCrowds An Authoring Tool for Crowd Simulation Gabriel Fonseca Silva

2025-05-06 0 0 659.45KB 7 页 10玖币

侵权投诉

WebCrowds: An Authoring Tool for Crowd

Simulation

Gabriel Fonseca Silva

Virtual Humans Lab

PUCRS

Porto Alegre, Brazil

gabriel.fonseca94@edu.pucrs.br

Paulo Knob

Virtual Humans Lab

PUCRS

Porto Alegre, Brazil

paulo.knob@edu.pucrs.br

Rubens Montanha

Virtual Humans Lab

PUCRS

Porto Alegre, Brazil

rubens.montanha@edu.pucrs.br

Soraia Musse

Virtual Humans Lab

PUCRS

Porto Alegre, Brazil

soraia.musse@pucrs.br

Abstract—Crowd simulation is an area of research largely used

in the game industry. From the movement of a single NPC to the

movement of an entire army, crowd simulation methods can be

used to move agents through the environment while avoiding

collisions with obstacles and between each other. Thus, it is

important that game developers have access to crowd simulation

tools that are both powerful and easy to use. In this paper,

we present WebCrowds, an authoring tool for crowd simulation

which can be used by anyone to build environments and simulate

the movement of agents. The results achieved by our research

suggest that WebCrowds is easy to use, delivers trustworthy

simulation results, and can be used as an authoring tool for

game developers who need to simulate crowds in their games.

Index Terms—Crowd Simulation, Authoring Tool, Virtual

Agents, Framework

I. INTRODUCTION

Several contributions have been made in the area of crowd

simulation, given its application in many different ﬁelds1.

Since the pioneer work of Thalmann and Musse [1], many

other methods emerged for both microscopic [2] and macro-

scopic [3] points of view and, more recently, combining them

both [4]. Ranging from the simulation of a crowd of people or

robots [5] to the modeling of personality for simulated virtual

agents [6], [7], crowd simulation has a wide range of uses in

different ﬁelds.

Focusing on the area of games, crowd simulation is im-

portant for many types of games. Since the movement of

Non-Playable Characters (NPCs) to the management of whole

armies, crowd simulation methods can be used to move virtual

agents from one point to another while avoiding collision with

themselves or other obstacles. Suyikno et al. [8] proposed

a model to simulate the hiding behavior of NPCs, while

Silva et al. [9] proposed a method to move agents forward

in time and space ﬁt to be used in ”Fog of War” systems.

Given the importance of crowd simulation in games, it is

essential that game developers have a tool where they can

build environments and simulate crowds.

This work presents an authoring tool for crowd simulation.

Named as WebCrowds, our tool is divided into two parts:

the user Editor, which is accessed through a web Browser,

and the crowd simulator, which is served as a web service

1Draft version made for arXiv: https://arxiv.org/

and is hosted inside a Server. WebCrowds allows users to

build environments placing agents, obstacles, and goals in

the scene. Users can also load presets from an OBJ ﬁle,

and even save/load their built scenes using JavaScript Object

Notation (JSON) ﬁles, to be used for other simulations. Since

the simulation is executed on a Server, users can access

WebCrowds even with lower-performance computers.

II. RELATED WORK

This section discusses some work related to authoring tools

and techniques for crowd simulation scenarios. Ulicny et

al. [10] presented a way of drawing a crowd simulation where

the user could use different types of brushes to create the

crowds and change their appearance, behavior, and direction

to create a scene to simulate. Chen et al. [11] presented a way

to compose crowd scenarios using natural language processing

techniques, where different words and expressions represent

the attributes and behavior of each agent. Some of these

expressions may contain emotions, such as ”walk happily” and

”walk sadly”. These agents are created using data provided by

a user, and this data is used to extract character keywords. This

model was also employed by Liu et al. [12], where keywords

such as quantity, character type, behavioral verb, location, and

object, are taken into account. The data structure of the crowd

scene is divided into three components: crowd, behavior, and

destination. Each of these components has its keywords.

Colas et al. [13] presented a model for authoring crowd

simulations using sketches. The user draws curves in the

interface, which are used to compose an interaction ﬁeld

grid, inﬂuencing the steering behavior of agents. Other agents

and obstacles in the scene are also taken into account when

composing the grid, which allows agents to avoid collision

with one another or move behind obstacles. Kim and Sung [14]

presented a model using sketching in combination with aug-

mented reality, where users can use dry-erase markers to

compose scenes. The authors use computer vision techniques

to identify shapes and colors in the sketches to deﬁne which

type of object (e.g. goals, paths, and trees) should be placed

in that position.

The work found in the literature focuses on crowd anima-

tion, where users may deﬁne behaviors or directly input an

arXiv:2210.04624v3 [cs.HC] 20 Dec 2022

animation for virtual characters; or explore new user inter-

action techniques, such as sketching, augmented reality, and

natural language. Differing from them, WebCrowds provides

information on the constructed environment and the simulation

run, allowing users to evaluate trajectories and densities of

virtual agents and make any adjustments desired.

III. PROPOSED MODEL

In this section, we present WebCrowds, an authoring tool

for crowd simulation which can be used by anyone to build

environments and simulate the movement of virtual agents.

Next, we present the main elements of our model.

A. System Architecture

Figure 1 presents an overview of our proposed model, which

is available online and can be accessed through a Browser2.

Inside WebCrowds, the users can create an environment by

adding, positioning, and editing objects. In the bottom right

of Figure 2(a), highlighted in purple, it is possible to see the

Objects that can be used, namely: Agents, Goals, Obstacles,

and Presets. More details about these objects are going to be

provided in Section III-B.

Following the pipeline presented in Figure 1, when the user

opens WebCrowds in their Browser, they have access to the

Editor, where the Scene can be edited to be then simulated.

Once the agents, goals, and obstacles are included, the user

can run the simulation. When the user runs the simulation,

the Request Controller gathers information about the Objects

in the Scene (e.g. their positions and properties) and sends

a request to the Server (Send Request). In its turn, when the

Server receives a new simulation, it triggers its own Routine. In

short, the Server executes the simulation with the parameters

sent from the Browser and, once it is ﬁnished, generates

the results. More details about the Server are provided in

Section III-C.

The results are generated and sent back to the Request

Controller (Receive Response), which will display them in

the browser (Simulation Results). These results are comprised

of three metrics: a Density Map (which shows the density of

agents during the simulation), the Agent’s Trajectories (which

shows the trajectories of each agent), and the Simulation Time.

Finally, In Figure 1, it is possible to see that the Server (on the

right) also connects with a Database. It was done to overcome

request timeout issues, which would prevent the simulation

from ﬁnishing.

B. Editor

In this section, we describe the three main elements of We-

bCrowds: Objects, Object Actions, and General Actions. All

three can be seen in Figure 2(a), highlighted in purple (bottom

right), yellow (bottom left), and blue (top), respectively.

2WebCrowds is available at: https://vhlab.com.br/projects/webcrowds/.

1) Objects: In WebCrowds, the users can deal with four

types of Objects: Agents, Goals, Obstacles, and Presets. All

of them can be seen in the bottom right of Figure 2(a),

highlighted in purple. These objects can be positioned within

the scene’s environment and are detailed next:

•Agents: Allows the creation of agents spawn areas, repre-

sented as blue squares in the central region of Figure 2(a);

•Goals: Allows the creation of goals for the agents to

achieve, represented as green circles in Figure 2(a);

•Obstacles: Allows the creation of obstacles in the envi-

ronment that should be avoided by the agents, represented

as red squares in Figure 2(a); and

•Presets: Allows the user to load built-in scenarios into

the environment. Presets behave as a set of Obstacles

that can’t be modiﬁed. We modeled ﬁve possible presets,

as presented in Figure 2(b).

2) Object Actions: In WebCrowds Object Actions are sim-

ply the actions that users can perform with Objects. The users

can control four types of Objects Actions: Create, Remove,

Move and Edit. All of them can be seen in the bottom left of

Figure 2(a), highlighted in yellow, and are detailed as follows:

•Create: Allows the creation of any type of Object (i.e.

Agent, Obstacle, Goal, or Preset);

•Remove: Allows the removal of Objects present in the

environment;

•Move: Allows the user to move any Object present in the

environment, positioning it where he/she sees ﬁt; and

•Edit: Allows the user to edit any Object present in the

environment. The editable options depend on the type of

Object. When editing an Agents object, users can change

the number of virtual agents to be spawned in that area,

as well as the Goal that these agents want to reach. When

editing an Obstacle, the user can change its size and

rotation. Goals and Presets have no editing options.

3) General Actions: In WebCrowds we have ﬁve General

Actions: Save Scene, Load Scene, Run Scene, Camera Control

and Reset Scene, as next detailed. All of them can be seen at

the top of Figure 2, highlighted in blue, where the Recycle

Bin icon represents the Reset action.

•Save Scene: Allows the user to save the current envi-

ronment into her/his computer/device. All the Objects

present in the environment are saved into a JSON ﬁle

and can be used in later simulations;

•Load Scene: Allows the loading of a JSON ﬁle from the

user’s computer/device. Ideally, this JSON ﬁle is the same

one that was generated with the Save Scene action, but

any JSON ﬁle can be used as long as it follows the same

information pattern;

•Run Scene: Sends a request to the Server to run a

simulation with the environment built;

•Camera Control: Allows the user to perform some actions

concerning the camera, such as Movement (with WASD

keyboard keys), changing the Camera Speed (as shown

in Figure 2(a)) and Zoom in/Zoom out (using the mouse

wheel); and

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

WebCrowds:AnAuthoringToolforCrowdSimulationGabrielFonsecaSilvaVirtualHumansLabPUCRSPortoAlegre,Brazilgabriel.fonseca94@edu.pucrs.brPauloKnobVirtualHumansLabPUCRSPortoAlegre,Brazilpaulo.knob@edu.pucrs.brRubensMontanhaVirtualHumansLabPUCRSPortoAlegre,Brazilrubens.montanha@edu.pucrs.brSoraiaMusseVirtua...

展开>> 收起<<

WebCrowds An Authoring Tool for Crowd Simulation Gabriel Fonseca Silva.pdf

共7页,预览2页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

WebCrowds An Authoring Tool for Crowd Simulation Gabriel Fonseca Silva

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: