> There is a theme issue this past year on pedestrain modeling in
> Environment and Planning B and for your collective information, I
> attach the pdf files.
> This has relevance generally to the state of the art in this area
Although these papers are interesting, not one of them approaches urban
level application for obtaining pedestrian numbers. Kerridge et al.'s
paper really addresses the questions about agent simulation, and the
examples are microscale (e.g., pedestrian approaching bus-stop), Helbing
et al.'s paper is strictly microscale interactions (e.g., pedestrian
encounters pedestrian on busy street), Lake's is really more of a
socialogical approach, such as Epstein and Axtell's Sugerscape, Hakley
et al.'s is an explanation more of the modular approach to simulation,
without implementational detail. Of these, the remaining Kurose paper
does indeed look at the beginnings of an agent system which could be
applied at an urban level, but involves significant preknowledge about
the intentions of people within the environment (e.g., *if* pedestrians
make three predefined stops, then...).
I can't really criticise EPB as a place to look for agent-based
pedestrian models as I've recently submitted an agent paper myself!
But, there are other journals which might (currently) be a more
appropriate place to start for techniques for working out aggregate
pedestrian numbers at an urban level using agent-based (or other
non-space syntax) techniques. For example, Transportation Research (all
parts), contains many transporation models, at least some of which
include (or could be applied to) pedestrian modelling.
For further papers, if you can get hold of the conference proceedings
for Pedestrian and Evacuation Dynamics, although again mostly
microscale, it does also contain some innovative pedestrian modelling
techniques (only a little self-plugging going on here! :-) which might
be applicable to urban level numeric data.
Going back to the original thread, I think we have to be a little
careful on density. Dense areas will tend to have more axial lines
(although the street length in the city of London may be comparable, the
road packing is certainly higher, one only has to look at an AZ to see
the difference between the city centre (typically enlarged so all the
roads can be seen) and a surburban area. Therefore the city of London
probably has higher integration in any case, so building density isn't
strictly an independent variable as Juan implies. In addition,
centrality surely plays a role as well. Alan's response about building
height seems a good pointer, but I can't help thinking it's a
transportation model calibration entering through the back door, which
leads to...
If we are going to do this properly I think (and I suspect many others
do as well) it is time to really do some hetrogenous modelling --- from
the looks of things, Jake and Juan are starting along this route already
--- perhaps a topological / metric graph weighted by the transportation
network and land use --- or my own favourite of agents using space
syntax-based decision processes weighted by land-use and from modal
interchange surveys.
Alasdair
|