I was very lucky in the last academic year to work with two very bright students on a couple of agent-based computing research projects. One of these was Oliver Don, and our work on modelling ant colonies is reported below.
The other student was Andrew Wood, and we decided to investigate the evacuation of the new Airbus A380 (the first ever double-decker passenger aircraft). Where this aircraft differs from those that have gone before is that the upper-deck exits are significantly higher than those on standard aircraft, and we wondered if "door delay" (ie. hesitation induced by the realisation of just how high the exit is) would have a significant impact on the time taken to evacuate the aircraft. In order to investigate this, we built an individual-based model of the aircraft (basically, model each individual passenger, each with their own particular attributes, and then "let it run"), and I think we came up with some interesting findings. The most significant conclusion we drew was that if average door delay exceeds just over 1s, then the aircraft will fail its certification trial (which requires it to be fully evacuated of passengers and crew within 90s).
Of course, this is just a prediction based on a simplistic model of the aircraft, but previous small-scale studies with real people have suggested that door delay will play a significant role. We await with interest the results of the full certification trial (which has yet to take place).
An article based on this work is currently in submission to a journal, but the project web page points to a pre-print version of it. There's also a version of the simulation that you can run in your browser and play with. In the mean time, I'll post an update when I hear anything about the progress of the paper, but this post is also meant to flag a forthcoming documentary, which I believe is showing in the UK this Friday (25th November) on Discovery Wings.