• Crowd simulation for emergency response using BDI agent based on virtual reality - (2006) - Simulation of an event and crowd panic/evacuation aided by police agents
  
• Defense and homeland security applications of multi-agent simulations - (2007) - 2D simulation involving many agent types reacting to an emergency. *Probably closest to our ideas*
  
• EMERGENCY FIRST RESPONSE TO A CRISIS EVENT: A MULTI-AGENT SIMULATION APPROACH - (2006) - Master's Thesis of paper immediately above
  
• Group information foraging in emergency response: an illustration incorporating discrete-event simulation - (2006) - simulation, but not in animation sense
  
• Synthetic humans in emergency response drills - (2006) - research poster, seems to have a system to simulate crowd emergency situation based on a floorplan, viewable in 2 or 3D
  
• Responding to terrorist attacks and natural disasters: a case study using simulation - (2005) - simulation, but not in animation sense
  
• Simulation for emergency response: homeland defense center network - capitalizing on simulation, modeling and visualization for emergency preparedness, response and mitigation - (2003) - background information
  
• Simulation for emergency response: a framework for modeling and simulation for emergency response - (2003) - background information
  
• Simulation of large crowds in emergency situations including gaseous phenomena - (2005) - mixing crowds and fluid animation
  
• Modeling Crowd and Trained Leader Behavior during Building Evacuation - (2006) - crowd evacuation with some group interaction, communication makes exiting quicker
  
• Implementation of Crowd system in Maya - (2006) - useful if approaching implementation as Maya plug-in

* a couple sentence description of the project

The goal of this project is to develop Massive-like software. MASSIVE agents act according to their environment as opposed to an internal model. Design principles (as of 2004) were based on idea that interesting effects can be achieved with carefully designed rules rather than complex A.I. algorithms.

* a summary of the state of the art refer to 2 or 3 of the most relevant papers you've found so far

A collection of course notes from SIGGRAPH 2004, including some from MASSIVE's chief developer
Crowd and Group Animation

This paper covers some basics on crowd animation, with focus on work in pathing and visualization issues
Real-Time Navigating Crowds: Scalable Simulation and Rendering

There are a few open source projects that may be of interest:
Crowd Simulation Framework
crowd movement simulation
AICrowd
SiRC IA

MASSIVE may be the only crowd simulation software that can handle individual interaction in real-time scenes of over ~50,000 agents. Most academic work is biased towards advances in behavior modeling, A.I., pathing, or rendering involved.

* what's the motivation for doing the project?

A demo system and future research, in particular the agent framework and (time and ideas permitting) a particular instance of agents, models, and animations in an interesting domain for a proposal.

* what would be the major challenges in doing the project? what are the unknowns?

The major challenge in this project is the design of the system so that it is powerful, flexible, and easy to use. For a good looking animation, a lot of artistic content is needed. The application domain is still needed, and this determines some of the possible agent subtypes and agent interactions.

* what could be done this quarter? what's the plan? what are the major tasks that would have to be done. can something useful be done by 1 person? how many people working on this would make sense?

• Base agent class representing an agent in 2/3D space
• Subroutines for specific actions like moving, following another agent, panic, etc.
• environment specification

One person could make progress this quarter, but it can be split so that 2-4 people could work on it at once.