A review on crowd simulation and modeling
Graphical abstract
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)
- et al.
State-of-the-art crowd motion simulation models
Transp. Res. C
(2013) - et al.
A GPU-assisted hybrid model for real-time crowd simulations
Comput. Graph.
(2013) - et al.
Position-based real-time simulation of large crowds
Comput. Graph.
(2019) A continuum theory for the flow of pedestrians
Transp. Res. B
(2002)- et al.
Continuum crowd simulation in complex environments
CAD/GRAPHICS 2009 Extended papers from the 2009 Sketch-Based Interfaces and Modeling Conference Vision, Modeling & Visualization
Comput. Graph.
(2010) - et al.
Emergency crowd simulation for outdoor environments
Comput. Graph.
(2010) Flocks, herds and schools: A distributed behavioral model
ACM SIGGRAPH Comput. Graph.
(1987)Steering behaviors for autonomous characters
- et al.
MURE: Fast agent based crowd simulation for VFX and animation
- et al.
Rivers of rodents: An animation-centric crowds pipeline for ratatouille
Headstrong, hairy, and heavily clothed: Animating crowds of scotsmen
Brain springs: Fast physics for large crowds on WALLe
Towards interactive real-time crowd behavior simulation
Comput. Graph. Forum
Parameter estimation and comparative evaluation of crowd simulations
Comput. Graph. Forum
Crowd-driven mid-scale layout design
ACM Trans. Graph.
Urban walkability design using virtual population simulation
Comput. Graph. Forum
Understanding Crowd Behaviours, Volume 2: Supporting Theory and Evidence
Measuring crowd collectiveness
IEEE Trans. Pattern Anal. Mach. Intell.
A survey of latest approaches for crowd simulation and modeling using hybrid techniques
A survey of real-time crowd rendering
Comput. Graph. Forum
Survey of real-time rendering techniques for crowds
Comput. Graph. Forum
Survey of techniques for rendering real-time virtual humans
Survey on real-time crowds simulation
Crowd simulation and its applications: Recent advances
J. Comput. Sci. Tech.
State of the art in example-based motion synthesis for virtual characters in interactive applications
Comput. Graph. Forum
A survey on position-based simulation methods in computer graphics
Comput. Graph. Forum
Crowd simulation
Virtual crowds: Steps toward behavioral realism
Simulating Heterogeneous Crowds with Interactive Behaviors
Real-time path planning for virtual agents in dynamic environments
Real-time path planning in dynamic virtual environments using multiagent navigation graphs
IEEE Trans. Vis. Comput. Graphics
Environmental abstraction and path planning techniques for realistic crowd simulation
Comput. Animat. Virtual Worlds
Real-time path planning in emergency using non-uniform safety fields
Interactive navigation of heterogeneous agents using adaptive roadmaps
IEEE Trans. Vis. Comput. Graphics
Egocentric affordance fields in pedestrian steering
Realistic following behaviors for crowd simulation
Comput. Graph. Forum
A statistical similarity measure for aggregate crowd dynamics
ACM Trans. Graph.
Scalable behaviors for crowd simulation
Comput. Graph. Forum
Fitting behaviors to pedestrian simulations
Group motion graphs
Modeling social behaviors in an evacuation simulator
Comput. Animat. Virtual Worlds
Environment optimization for crowd evacuation
Comput. Animat. Virtual Worlds
On density–flow relationships during crowd evacuation
Comput. Animat. Virtual Worlds
CODE: Crowd-optimized design of environments
Comput. Animat. Virtual Worlds
Hybrid long-range collision avoidance for crowd simulation
IEEE Trans. Vis. Comput. Graphics
Interactive hybrid simulation of large-scale traffic
ACM Trans. Graph.
A modular framework for adaptive agent-based steering
Footstep navigation for dynamic crowds
Comput. Animat. Virtual Worlds
Robust space-time footsteps for agent-based steering
Comput. Animat. Virtual Worlds
Cited by (81)
Improving safety management in railway stations through a simulation-based digital twin approach
2024, Computers and Industrial EngineeringSocial groups in pedestrian crowds as physical and cognitive entities: Extent of modeling and motion prediction
2023, Transportation Research Part A: Policy and PracticeArchitectural spatial layout planning using artificial intelligence
2023, Automation in ConstructionModeling domino effect along the queue using an improved social force model
2023, Physica A: Statistical Mechanics and its ApplicationsModeling heterogeneous behaviors with different strategies in a terrorist attack
2023, Virtual Reality and Intelligent Hardware