crowd

An Event-Driven Approach toHuman Crowd Simulationwith Example Motions: 

An Event-Driven Approach to Human Crowd Simulation with Example Motions

OUTLINE: 

Motivation Basic Idea Related Work Overview Initializer Behavior Generation Motion Generation Results Conclusion OUTLINE

1. Motivation: 

1. Motivation Crowd Scenes Commonplace in computer games and feature films Evolving from the interactions among individual members and their interactions with the environment Highly dynamic and complex An Event-driven Approach to Human Crowd Simulation with Example Motions

Why example motions?: 

To generate realistic motions efficiently. How? Motion capture and reuse. On–line motion blending. [Park et al. 2002] Why example motions?

Why “Event – driven” ?: 

Why not employing a fixed–time advancing scheme? Well, …… Some well–known characteristics of crowds : Members clustered together very closely. Living creatures to actively interact with each other and also with the environment.  Adaptive correcting behaviors to dynamically changing situations even during a time step !!! Why “Event – driven” ?

2. Basic Idea: 

2. Basic Idea Event-driven approach Guided by the external events reflecting the scenario, the behavior of a crowd evolves as intended, while adding variations due to the internal events. Live-captured motion clips

3. Related Work: 

3. Related Work Non-human creatures : reactive behaviors Dynamic spatial partitioning [Reynolds 1987] Synthetic visions [Tu and Terzopoulos 1994] Pedestrian dynamics : Pedestrian crowd modelling Similar to gases and fluids [Henderson 1974] Individual pedestrian dynamics [Gipps and Markiso 1985; Helbing 1992, 1995, 2000]

3. Related Work (cont.): 

3. Related Work (cont.) Behavioral rules Behavioral model of groups: Border Collies and Olympic bicycle racing [Brogan et al. 1997] ViCrowd [Musse and Thalmann 2001] Improv [Perlin and Goldberg 1996] Real-world systems Based on pedestrian dynamics together with procedural rules AntZ, The Lord of the Rings Plug-ins to MayaTM and SoftimageTM

4. Overview: 

Script parser Motion Generator Event Scheduler External Event Generator Internal Event Generator Scenario Behavior Generator Example motions Event handler Initializer 4. Overview

5. Initializer: 

Script Parser Scenario behavioral rules external Events Event Handler Event Scheduler 5. Initializer External Event Generator

6. Behavior Generation: 

6. Behavior Generation Script Parser behavioral rules Event Handler Internal Event Generator Event Scheduler External Event Generator Motion Generatior External event Event Candidate Internal Event Motion specification

Event-Driven Paradigm: 

Event-Driven Paradigm External events Internal events (candidate)

Internal Event Prediction: 

Internal Event Prediction Predicting two candidate internal events for a crowd member A candidate resetting event A candidate colliding event Why candidates?

Space Subdivision: 

Space Subdivision

Time-Varying Bounds: 

Time-Varying Bounds Assumptions - moving trajectories unknown - bounded acceleration Time – varying bounds Uncertainty in future facilitating interactions r

Slide16: 

Registration of Time-varying Bounds (C(sk)) R(sk) C(sk) sk ℓ 2ℓ A time-varying bound sk is said to be registered in a subspace C(sk) if the center of sk was contained in C(sk) at its most recent event time. C(sk), R(sk), and R(sk) 3ℓ

Slide17: 

Candidate Resetting Event Generation (C(sk)) R(sk) C(sk) Sk = {all time-varying bounds registered in (R(sk)).

Slide18: 

Candidate Colliding Event Generation

Event Handling: 

Event Handling Deriving the reactions to events Updating the internal status Behavioral rules + external information

How to Handle External Events: 

How to Handle External Events Controlling the global flow of crowd simulation Drawing the reactions from behavioral rules Prescribing motion specifications Creating new candidate internal events

How to Handle Internal Events: 

How to Handle Internal Events Local interactions among crowd members Interaction with the environment Adding details to crowd behaviors A force field model of pedestrian dynamics

Slide22: 

Force Field Model avoiding force braking force driving force

Slide23: 

Driving Force global velocity control local velocity adjustment route goal

Braking Force: 

0 0 Braking Force

Avoiding Force: 

0 Avoiding Force

Resetting Event Handling: 

- resetting the time-varying bound sk to its initial size - probing the position xk(t) and velocity vk(t) to re-initialize the time-varying bound sk making registration change if needed. new location old location Resetting Event Handling

Colliding Event Handling: 

identify the neighbors and nearby objects compute the force exerted on each of colliding members reset the time-varying bounds Colliding Event Handling

7. Motion Generation: 

Example motions Parameterization Motion Blending Weight Computation Time Warping Posture Blending Motion Retargeting Motion Specifications Target Motion 7. Motion Generation

8. Results: 

8. Results

9. Conclusion: 

9. Conclusion Event-driven approach Locomotive behaviors Collision detection + Pedestrian dynamics Future work Scripting system (scenario + behavioral rules) Sophisticated behavioral rules Generalization to other types of motions