Elsevier

Graphical Models

Volume 111, September 2020, 101081
Graphical Models

A review on crowd simulation and modeling

https://doi.org/10.1016/j.gmod.2020.101081Get rights and content

Abstract

Crowd simulation has emerged in the last decade as a widely used method of visual effects, computer games, and urban planning, etc. The improvement of hardware performance and the urgent need of special effect lead to an unprecedented wave of crowd simulation studies. This paper reports a review of crowd simulation models from traditional methods to recent methods (e.g. group simulation, emotion contagion). Traditional models can simulate general crowd dynamics which have the advantages of both microscopic and macroscopic models. The recent studies of crowd simulation from group simulation to social psychology crowds are possible to simulate realistic crowds. The purpose of this review is to introduce commonly used crowd simulation methods for newcomers to this field by making a systematic literature review, discussions and analysis of different models. The results reveal the traditional models can simulate most of the normal crowds, but lack expressiveness for special groups which needs to be solved urgently in the current applications, particularly on visual effects and urban planning. Group simulation and emotion contagion could improve the simulation realism, but it also needs to be improved in computation cost and model optimization. Also, future research directions are suggested aiming to develop new applications focused on more realistic, natural and efficient crowd simulation.

Introduction

Crowd simulation is a technique to simulate the motion dynamics of virtual individuals. The earliest crowd simulation system “Boids” was proposed by Reynolds [1], [2] in 1987, which was an artificial life project to simulate the flocking behavior of birds. Since then, crowd simulation has been studied by many researches and has been widely used in the fields of visual effects [3], [4], [5], [6], computer games [7], [8], and urban planning [9], [10], etc. Although this field has gained a lot of research progress and is developing rapidly, the influence of locomotion, sensory abilities and a series of psychological factors make individual behavior become complex in different situations. Due to high computational complexity of such heterogeneous crowds, there exist many different challenges which limit the realism in crowd simulation. With the development of computer equipment, understanding and controlling human behavior has become a hot topic. It is of great significance to study how to simulate realistic people to improve the authenticity of visual effects, enhance the immersion of virtual reality, and ameliorate the rationality of urban planning and the efficiency of emergency evacuation.

In [11], a crowd was defined as “a large group of individuals share information in the same environment alone or in a group”. The key criteria of a crowd included crowd size, crowd density, specific time, crowd collectivity and crowd novelty [12].

In [13], Ijaz et al. gave a survey on hybrid crowd simulations, and the real-time crowd rendering techniques were reviewed in [14], [15], [16], [17]. An overview of crowd simulation and its application was provided in [18], which mainly introduced the related works of CAD & CG laboratory in Zhejiang University. The state-of-the-art models were presented and compared in [19]. Surveys of example-based and position-based crowd simulation approaches can be found in [20] and [21]. The introduction of different crowd simulation methods was shown in [22]. Efficient algorithms to animate, control, and author human-like agents were described in [23] to help researchers to design realistic digital humans and their interactions with environments. A deep understanding of state-of-art methods to simulate realistic crowds were provided in [24].

During the past 20 years, researchers have proposed various techniques for crowd simulation, e.g. path planning [25], [26], [27], [28], [29], [30], behavior modeling [31], [32], [33], [34], navigation graphs [35], emergency evacuation [36], [37], [38], [39], hybrid modeling [40], [41], [42], [43] and biomechanical models [44], [45]. The urgent need for simulating a variety of crowds is limited by the great need for manpower and hardware resources [46]. The performance of the algorithm, such as rendering strategy and group generation scheme, needs to be improved. At the same time, it is necessary to consider the reality and accuracy of virtual groups.

In this paper, we track the current widely used models which are representative or frontier. Section 2 briefly introduces the research methodology used in this paper. Then we review mainstream traditional crowd simulation methods include microscopic models in Section 3 and macroscopic models in Section 4. We switch focus to recent crowd simulations in Section 5, which includes dynamic group structure, adaptive group formation, and social psychology crowds. Section 6 provides the comparisons and analysis of the major approaches. Conclusions and development tendency are provided in Section 7.

Section snippets

Methodology

Microscopic models

Microscopic models are also called “Bottom-Up” methods, which focus on low-level behavioral details and individual features [13]. In this category, individuals are considered as discrete objects whose motion is influenced by their neighbors and the obstacles. On the other hand, the collision avoidance is the local interactions with the surrounding environment, and the combination of these local behaviors produces individual final movement.

Macroscopic models

While simulating large-scale and dense scenario, crowds are considered as a unified and continuous entity, and its movement is governed by potential fields or fluid dynamics, etc. Crowd path planning and collision avoidance are both governed by the global problem solver. These models do not focus on the underlying details, so that each virtual agent does not consider the individual level interactions between others and the environment.

Mesoscopic crowd simulation models

The macro model can simulate thousands of people, while the micro model is based on individuals, and the simulation of large-scale crowd movement is less efficient. Recently, researchers begin to focus on mesoscopic models (e.g. group simulations) which can be divided into (1) dynamic group behaviors such as the social relationships among individuals in dynamic groups, (2) interactive group formation in the RTS games and manipulating schemes, and (3) social-psychological crowds in emergency

Comparison of different methods

In order to evaluate the performance of different methods, it is desirable to establish uniform evaluation criteria. Berseth et al. [159] investigated the effect of the model parameters on steer algorithm’s performance to optimize an algorithm’s parameters for a range of objectives. However, it is quite difficult to establish uniform evaluation criteria for different categories of models. We collect and analyze the information provided by the papers describing the respective model and focus on

Conclusion

This paper reviewed the models used in crowd simulation. We compared the models in different categories and provided a broad overview of the current literature on crowd simulation models of the last decades. The simulation results generated by the traditional models (e.g. microscopic and macroscopic models) sometimes are not as realistic as the real-world human behavior. In general, people’s movements are unpredictable and effected by their own internal personality traits. Therefore, even in

CRediT authorship contribution statement

Shanwen Yang: Writing - original draft, Methodology, Software. Tianrui Li: Conceptualization, Supervision, Writing - review & editing. Xun Gong: Resources, Investigation. Bo Peng: Resources, Investigation. Jie Hu: Investigation, Writing - review & editing.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgment

This work is supported by the National Key R&D Program of China (No. 2019YFB2101802) and the National Natural Science Foundation of China (Nos. 61773324, 61876158).

References (167)

  • KanyukP. et al.

    Headstrong, hairy, and heavily clothed: Animating crowds of scotsmen

  • KanyukP. et al.

    Brain springs: Fast physics for large crowds on WALLe

  • UlicnyB. et al.

    Towards interactive real-time crowd behavior simulation

    Comput. Graph. Forum

    (2002)
  • WolinskiD. et al.

    Parameter estimation and comparative evaluation of crowd simulations

    Comput. Graph. Forum

    (2014)
  • FengT. et al.

    Crowd-driven mid-scale layout design

    ACM Trans. Graph.

    (2016)
  • MathewC.D.T. et al.

    Urban walkability design using virtual population simulation

    Comput. Graph. Forum

    (2019)
  • ShepherdR. et al.

    Understanding Crowd Behaviours, Volume 2: Supporting Theory and Evidence

    (2010)
  • ZhouB. et al.

    Measuring crowd collectiveness

    IEEE Trans. Pattern Anal. Mach. Intell.

    (2014)
  • IjazK. et al.

    A survey of latest approaches for crowd simulation and modeling using hybrid techniques

  • BeaccoA. et al.

    A survey of real-time crowd rendering

    Comput. Graph. Forum

    (2016)
  • RyderG. et al.

    Survey of real-time rendering techniques for crowds

    Comput. Graph. Forum

    (2005)
  • RyderG. et al.

    Survey of techniques for rendering real-time virtual humans

  • AzaharM.A.B.M. et al.

    Survey on real-time crowds simulation

  • XuM.-L. et al.

    Crowd simulation and its applications: Recent advances

    J. Comput. Sci. Tech.

    (2014)
  • PejsaT. et al.

    State of the art in example-based motion synthesis for virtual characters in interactive applications

    Comput. Graph. Forum

    (2010)
  • BenderJ. et al.

    A survey on position-based simulation methods in computer graphics

    Comput. Graph. Forum

    (2014)
  • ThalmannD.

    Crowd simulation

  • KapadiaM. et al.

    Virtual crowds: Steps toward behavioral realism

  • KapadiaM. et al.

    Simulating Heterogeneous Crowds with Interactive Behaviors

    (2016)
  • SudA. et al.

    Real-time path planning for virtual agents in dynamic environments

  • SudA. et al.

    Real-time path planning in dynamic virtual environments using multiagent navigation graphs

    IEEE Trans. Vis. Comput. Graphics

    (2008)
  • ParisS. et al.

    Environmental abstraction and path planning techniques for realistic crowd simulation

    Comput. Animat. Virtual Worlds

    (2006)
  • LiuB. et al.

    Real-time path planning in emergency using non-uniform safety fields

  • GayleR. et al.

    Interactive navigation of heterogeneous agents using adaptive roadmaps

    IEEE Trans. Vis. Comput. Graphics

    (2009)
  • KapadiaM. et al.

    Egocentric affordance fields in pedestrian steering

  • LemercierS. et al.

    Realistic following behaviors for crowd simulation

    Comput. Graph. Forum

    (2012)
  • GuyS.J. et al.

    A statistical similarity measure for aggregate crowd dynamics

    ACM Trans. Graph.

    (2012)
  • SungM. et al.

    Scalable behaviors for crowd simulation

    Comput. Graph. Forum

    (2004)
  • LernerA. et al.

    Fitting behaviors to pedestrian simulations

  • LaiY.-C. et al.

    Group motion graphs

  • ChuM.L. et al.

    Modeling social behaviors in an evacuation simulator

    Comput. Animat. Virtual Worlds

    (2014)
  • BersethG. et al.

    Environment optimization for crowd evacuation

    Comput. Animat. Virtual Worlds

    (2015)
  • HaworthB. et al.

    On density–flow relationships during crowd evacuation

    Comput. Animat. Virtual Worlds

    (2017)
  • HaworthB. et al.

    CODE: Crowd-optimized design of environments

    Comput. Animat. Virtual Worlds

    (2017)
  • GolasA. et al.

    Hybrid long-range collision avoidance for crowd simulation

    IEEE Trans. Vis. Comput. Graphics

    (2014)
  • SewallJ. et al.

    Interactive hybrid simulation of large-scale traffic

    ACM Trans. Graph.

    (2011)
  • SinghS. et al.

    A modular framework for adaptive agent-based steering

  • SinghS. et al.

    Footstep navigation for dynamic crowds

    Comput. Animat. Virtual Worlds

    (2011)
  • BersethG. et al.

    Robust space-time footsteps for agent-based steering

    Comput. Animat. Virtual Worlds

    (2015)
  • D. Thalmann, H. Grillon, J. Maim, B. Yersin, Challenges in crowd simulation, in: 2009 International Conference on...
  • Cited by (81)

    • Modeling domino effect along the queue using an improved social force model

      2023, Physica A: Statistical Mechanics and its Applications
    View all citing articles on Scopus
    View full text