Pedestrian and street simulation

Jamie's simulation seems very appealing. Jamie's work is related to the study of social relationships found in urban areas, as well as the design of proper system to distribute content within a local community (the distibution will be based on collocation events). To meet this end, he built a computer simulation of to provide the artificial collocation data upon which an information sharing application can be developed.
This is an agent-based simulation where each agent represents a pedestrian moving about the city. The spatial representation is basically a graph: a "semi-realistic street grid upon which the agents will move". Using a grid could take advantage of the space syntax methodology. The complete methodology is described in his 9th month phd report. Since Jamie is dubious about whether or not he should make this document public, I don't describe how he did this. Why do I blog this? The simulation architecture is really smart. I am interested in this sort of things to simulate the catchbob activity (or a simpler version of this collaborative treasure hunt). Don't know whether it is possible or not. The point would be to have this kind of simulation and to add a collaborative feature on it. I mean, implementing a mutual modeling feature that modify the agents' behavior depending on various components like: each agent's position, common knowledge... (personal traits?)... This said, it's just random thought here, I absolutely don't know how to code this sort of stuff. I'm just dreaming here...
Anyway, this is very stimulating, what do you think about this jamie? :)