Project Groups 2009-2010 - John Drake
-
Evacuation Simulation Design & Validation (John Drake)
[Abstract | [Project Blog] | [Proposal] | [Final Report] | [Final Presentation]
Abstract
It is hard to evaluate the fidelity of existing crowd simulation systems, and in particular, how well they apply to evacuation scenarios. Elements of the design and implementation of existing crowd simulations affect their accuracy. For example, a simulation built on the idea of cellular automata may not reflect how real human beings behave. Real humans do not operate on a fixed grid and may not have simple formulas controlling how they act over time.
Beyond the simulations themselves, there are problems too with current methods of validating results. The results of a crowd simulation can be checked against read-life data from an evacuation drill. As a drill, this may not be representative of a real emergency. Recorded data from actual emergencies can be used to compensate for this. However, data from real emergencies cannot be generalized and cannot be gathered in large quantities.
The goal of this project is to build an experimental setup inside a virtual community robust and flexible enough to implement evacuation experiments. Second Life, a popular virtual world produced by Linden Labs, will be used, but the principles, and possibly some of the implementation could be adapted to other virtual environments. The immersive ability of such simulations will be evaluated by analyzing subjects' presence. Presence is a measure of emotional attachment to a virtual experience, as defined in papers by M. Slater (see top of references section). Finally, the experiment setup will be used to compare the effectiveness of evacuation safety features