Computational Modeling of Nonadaptive Crowd Behaviors for Egress Analysis

2004-2005 and 2005-2006 CIFE Seed Project

Civil and Environmental Engineering, Stanford University

 

Principal Investigators: Professor Kincho H. Law , Professor Jean-Claude Latombe and Dr. Ken Dauber
Research Student: Xiaoshan Pan, Peng Gao

  Project Description:

The objective of this research is to study human individual and social behavior for emergency exit in buildings and facilities. There have been numerous reports regarding overcrowding and crushing incidents during emergency situation. In case of crowd disasters, one observation is that most victims were killed by the so called "nonadaptive" behaviors of the crowd, rather than the actual cause (such as fire). There has been a wide variety of computational tools for the simulation and design of exits. However, due to the scarcity of behavioral data, these tools rely heavily on the assumptions about human individual and social behaviors. Many of these assumptions have been found inconsistent or incorrect. This proposed study will aim to investigate nonadaptive crowd behaviors from the perspectives of human and social interactions and to incorporate such behavior in a dynamic computational model suitable for emergency exit and egress analysis. (more...)


  Try MASSEgress Demo Version!

A simplified version of MASSEgress is available to download for research and educational purpose. Contact us for more details.  (more...)


  CIFE Proposals:


  Other Presentations:


  Publications:

X. Pan. "Computational Modeling of Human and Social Behaviors for Emergency Egress Analysis," PhD Thesis, Civil and Environmental Engineering Dept., Stanford University, June, 2006.

X. Pan, C. S. Han, K. H. Law, and J. Latombe. "A Computational Framework for the Simulation of Human and Social Behaviors during Emergency Evacuations," submitted to Joint International Conference on Computing and Decision Making in Civil and Building Engineering, Montreal, June 14-16, 2006.

X. Pan, C. S. Han, K. Dauber and K. H. Law. "Human and Social Behavior in Computational Modeling and Analysis of Egress," Building Future Council Doctoral Program, Las Vegas, March 16 - 17, 2005. It is also published in Automation in Construction, 15(4):448-461, 2006.

X. Pan, C. S. Han and K. H. Law. "A Multi-agent Based Simulation Framework for the Study of Human and Social Behavior in Egress Analysis," The International Conference on Computing in Civil Engineering, Cancun, Mexico, July 12-15, 2005.

X. Pan, C. S. Han, K. Dauber and K. H. Law. "A Multi-agent Based Framework for the Simulation of Human and Social Behaviors during Emergency Evacuations,Social Intelligence Design, Stanford, March 24-26, 2005. An updated version of this paper is accepted for journal publication, AI and Society, 2006.

X. Pan, C. S. Han, K. H. Law, and J. Latombe. "A Computational Framework to Simulate Human and Social Behaviors during Emergency Evacuations," Blume Center News, The John A. Blume Earthquake Engineering Center, Stanford University, Issue No. 44, Spring, 2006.


  System Description:

 

We have developed a multi-agent prototype, Multi-Agent based Simulation System for Egress analysis (MASSEgress), by using Visual Lisp, C++, MFC and OpenGL, which is able to simulate some emergent human social behaviors, such as competitive behavior, queuing behavior,  herding behavior, and bi-directional flow through modeling human agents at microscopic level.

  1. System architecture
  2. Representation of an agent
  3. Sensing, decision-making (behavior selection) and motor control
  4. Statistical analysis of evacuation pattern using clustering algorithm

  Project Update (case studies): (03/07/2006)

*In order to view movie files properly, Windows Media Player V.9 or higher is required.

  Rhode Island Nightclub Fire

On the night of February 20, 2003, a fire broke out in the Station nightclub at 211 Cowesett Avenue, West Warwick, Rhode Island. The fire spread out and became fatal in matter of minutes. Evacuation from the nightclub was hampered by crowding at the main entrance to the building. About one hundred people lost their lives in the fire, and most of the deaths occurred during the evacuation process. This case study aims to (1) replicate the egress of Rhode Island Nightclub fire based on NIST data (quantitative validations using historical data), and (2) explore other scenarios that could lead to better egress assuming the same emergency.

Please click here to see some simulation screenshots. (Analysis and more movies are on the way...)

Movie: Top view;

 

 

 

Competitive behaviors are often observed when people act very individualistically in emergencies. The associated assumptions include:

1)     there are EXIT signs placed properly in a building,

2)     a human agent can detect a  EXIT sign if  there is no obstacle between his/her eyes and the EXIT sign and the distance from his/her eyes to the EXIT sign is within a visual range.

3)     human agents act individually.

Under these assumptions, human agents compete for their own chances of exiting, which often cause blockages at the exits of a building.

 

Movies: isometric view; top view;

 

 

 

Queuing behavior assumes:

1)     there are EXIT signs placed properly in a building,

2)     a human agent can detect a  EXIT sign if (A) there is no obstacle between his/her eyes and the EXIT sign and (B) the distance from his/her eyes to the EXIT sign is within a visual range.

3)     human agents intend to self-organize into exiting lines.

Under these assumptions, human agents take into considerations of visualized EXIT signs and the presences of other agents, and then they formulate exiting queues at the places where they cannot exit at the same time.

Movies: isometric view; top view;

 

 

Herding behavior assumes

1)     there are EXIT signs placed properly in a building,

2)     a human agent can detect a  EXIT sign if  there is no obstacle between his/her eyes and the EXIT sign and the distance from his/her eyes to the EXIT sign is within a visual range.

3)     a human agent decides which way to exit by following others.

Under these assumptions, when a human agent detects more than one exits, he/she would choose the one that has the most crowds.

 

Movies: isometric view; top view;

 

 

Bi-directional flow assumes

1)     a crowd is composed of individuals who have different goals,

2)     individuals constantly steer toward the goals,

3)     individuals anticipate potential collisions and take actions to avoid them in advance.

 

Movie: isometric view;

 

Following a leader (movie)

 

Subway station: simple behavior; complex behavior;

 

20% agents competitive; 80% agents queuing; (movie)

80% agents are queuing; 20% agents competitive; (movie)

Queuing v.s. Competitive - mixed population


If you have any comments or suggestions, please send them to xpan@stanfordalumni.org