From One Virtual Actor to Virtual Crowds: Requirements and Constraints Soraia Raupp Musse
Daniel Thalmann
Computer Graphics Lab.
Computer Graphics Lab.
Swiss Federal Institute of Technology EPFL, DI-LIG, CH 1015 Lausanne, Switzerland http://ligwww.epfl.ch
Swiss Federal Institute of Technology EPFL, DI-LIG, CH 1015 Lausanne, Switzerland http://ligwww.epfl.ch
[email protected]
[email protected]
ABSTRACT
2. RELATED WORKS
A crowd is not only a group of many individuals. Crowd modelling involves problems that arise only when we focus on crowds. For instance, avoiding collision problems related to a big number of individuals in the same place require different strategies in comparison with the methods used to avoid collision between individuals. This paper is concerned with a discussion about the requirements to model behaviours of crowds of virtual actors.
Reynolds [8] also studied behavioural animation in order to model aggregated motions. He introduced flocking systems, a hierarchical structure where the flock is composed by a group of behaviours that are collision avoidance, velocity matching and flock centring. Tu et al [10] used flocking systems to model complex group behaviours existent in aquatic ecosystems. Brogan [4] and Bouvier [3] presented groups and crowd simulations using particles systems and significant dynamic. In recent work, a crowd model has been introduced using different abstractions of behaviours, like the term guided crowd [6].
Keywords Multi-agent systems, synthetic agents and crowds model.
1. INTRODUCTION The mass behaviours and motion of crowds have been simulated in computers with different purposes. To be applied in entertainment areas in order to simulate a realistic motion of lots of virtual people that can be directed as easily as one only individual, e.g. films production and games. To supply simulation of crowd motion in order to evaluate constrained and complex environments, e.g. simulate the flow of people in a football stadium. To populate immersive virtual environments in order to improve the sense of presence, e.g. collaborative virtual environments. And, finally the behavioural simulation of crowds in generic environment describing social and sociological behaviours, the leadership, grouping and membership existing in crowds. This paper is concerned with issues related to point out the requirements to deal with crowds of virtual human agents. Also, ViCrowd system [7] is briefly presented in order to exemplify a behavioural tool to simulate crowd behaviours.
3. REQUIREMENTS FOR CROWDS MODELING Some constraints arise when we deal with crowds of virtual actors different from the modelling of virtual individuals. Indeed, there are different types of constraints depending on the different applications. First of all, the requirements for each crowd application are dependent on the nature of application. We propose a classification of crowd application based on different purposes. a) Entertainment applications include film productions and games. In this case, the groups have to be easily controlled and directed as well as individuals. Some examples are the recent films AntZ [1] and Bugs Life [2]. The challenges include the simulation of lots of people as well as the automatic generation of individualities that are important in the visual aspects. In addition, the easy control of groups represents a requirement of this kind of application. b) Crowd motion simulation aims at evaluating the motion of lots of people in a constrained environment. It normally involves a simple visualisation (embodiment of crowd) and a strong compromise with the numerical results in order to evaluate the environment. Some commercial softwares, e.g. SIMULEX [9], are examples of this kind of crowd application. The main challenges in this case include a realistic method of collision avoidance, a strong connection with the environment and the compromise with the numerical and statistical results. c) In order to populate collaborative virtual environments, e.g. in a virtual reality system, the crowd requirements implies in
providing a real-time simulation, as well as many aspects of interactivity in order to guide and direct crowds during the simulation. Yet, the facility to program crowds behaviour represents an important challenge in collaborative virtual environments, where participants can create and interact with crowds. Some experiments have been made in the context of COVEN project [6]. d) The last kind of crowd application concerns the behavioural modelling of crowds. In this case, the goal is to provide autonomous or semi-autonomous behaviours that can be applied by self-animating agents, which form the crowd. In order to deal with the constraints of having lots of actors (hundreds or thousands), this kind of application presents the following challenges: description of sociological and social aspects which arise in crowds, dealing with directed and emergent behaviours and connection with the virtual environment. Yet, needed optimisations and simplifications are required in order to be able to model and simulate crowd intelligence and decision ability.
4. ViCrowd OVERVIEW ViCrowd is the name of the system and model we have introduced [5], [6] and [7] ViCrowd aims at dealing and managing with the crowd information allowing directness (crowds can be programmed as easily as one only actor), interactivity (users and participants can guide and interact with crowds during the simulation) and autonomy (behaviours can be generated automatically based on rules associated to the crowd). Then, in ViCrowd, at the same time, we can program behaviours, interact with groups of actors and describe rules to provide self-animating agents who can decide their actions and motions. Some objectives of ViCrowd are: i)
Modelling of crowd information and structure.
ii)
Dealing with basic behaviours related to crowd problems, e.g. sharing virtual space, keeping the individual space in order to avoid collisions, be near to agents of same group, keep group goals, etc.
Fig. 1: Collision avoidance method applied between groups. One group is divided in order to avoid another one.
iii) Studying and modelling of some sociological effects which can occur in crowds or groups of people, e.g. membership, leadership, goals changing, etc. iv) Providing an interface to easily program crowds as well as behavioural rules in order to describe events and reactions to be treated by agents.
Fig. 2: Virtual crowd reacting in a Museum v)
Including the required structure of crowd control in order to provide interactivity, directness and autonomy in virtual crowds.
5. FINAL REMARKS The problem to simulate crowds of virtual human actors presents some constraints and requirements, which do not exist in individual agents’ simulation. This idea has not been much explored because the crowd simulation represents a new domain of research. Then, the goal of this paper is presenting the requirements needed to simulate crowds with different purposes
6. ACKNOWLEDGMENTS The research was sponsored by the Swiss National Research Foundation, the Federal Office for Education and Science (ESPRIT eRENA project), FUNDEPE and CAPES (Brazilian office of Education and Science).
7. REFERENCES [1] AntZ (1999) http://www.antz.com. [2] Bugs Life (1999) http://www.bugslife.com. [3] Bouvier E.; Cohen E. and L. Najman. (1997) "From crowd simulation to airbag deployment: particle systems, a new paradigm of simulation". Journal of Electronic Imaging 6(1), 94107, January. [4] Brogan, D.; Hodgins, J. (1997) “Group Behaviors for Systems with Significant Dynamics”. Autonomous Robots, 4, 137-153. [5] Musse, S.R.; Thalmann, D. (1997) “A Model of Human Crowd Behavior:”. Proc Workshop of Computer Animation and Simulation of Eurographics’97, Sept. Budapest, Hungary. [6] Musse, S. R.; Babski, C.; Capin, T.; Thalmann, D. (1998) “Crowd Modelling in Collaborative Virtual Environments”. Proc. of ACM – VRST’98. November. Taipei, Taiwan. [7] Musse, S. R., Garat, F. and Thalmann, D (1999). "Guiding and Interacting with Virtual Crowds". Proceedings of Workshop Eurographics Computer Animation and Simulation. Milan, Italy. [8] Reynolds, C. (1987) “Flocks, Herds and Schools: A Distributed Behavioral Model”. Proc. SIGGRAPH’87, Computer Graphics, v.21, n.4, July. [9] Thompson, P.A., Marchant, E.W. (1995) “A Computer Model for the Evacuation of Large Building Populations”. Fire Safety Journal, n. 24, pp-131-148. [10] Tu, X.; Terzopoulos, D. (1994) “Artificial Fishes: Physics, Locomotion, Perception, Behavior”. Proc. SIGGRAPH’94, Computer Graphics, July.
