Alasdair Turner wrote:

> On this note: I'd be interested in your expert opinion of (published)
> validated models -- are there any in particular that you would advise us
> in the space syntax field to take more note of?

Thank you Alasdair, that's very kind of you to say so.

And I find it a difficult question to answer. (So I'll try to cover up
that, with verbosity. Sorry, dear readers.)

I think there are some significant gaps in mutual understanding between
researchers in (and clients of) transport modelling and space syntax,
because those fields have historically been used for such different
needs. Architects and Land-use Planners look at spatial reconfiguation
as one of their primary tools of intervention. For Transport Planners,
that's typically a last resort, as it's the slowest and most expensive
of their available tools.

There is work being done by Legion.biz and others on agent-based dynamic
micro-simulation that uses a form of visibility analysis as one of its
three pedestrian movement determinants, which sits most interestingly in
the two fields. I mention that in particular because they've been most
market-responsive in finding out what types of interventions their
potential transport clients need/want to model.

There are some interesting route-choice algorithms out there; one that
comes immediately to mind, that I've probably mentioned on this list
previously, is one by Andrew Daly from some years ago which modelled
movement across the road hierarchy as traversing an arch (local road ->
local distributor -> major distributor -> local distributor -> local
road), which produces an assignment which starts to look like real life,
and also starts to look a bit like space-syntax, in that well-integrated
links [major distributor] appear on most journeys, and thus have the
highest flows.

> On a similar theme, Dijksta and Timmermans at Eindhoven seem to be
> getting more into validated activity-based modelling.

I've been lucky enough to work alongside one of their students, some
years ago now, and we looked at marrying the disaggregate-demand school
of Ben Akiva, Daly, Gunn et al with that activity-based approach. I
don't have any success to report there, I'm afraid, but that's not to
say that there isn't future hope there.

I believe that a hybrid approach, taking the best of agent-based
modelling, combining with disaggregate demand, and posssibly even system
dynamics, will be most fertile. Transport modelling has a deep unsolved
problem about the effect on land use of changes in the transport system,
and space-syntax-related research may well offer a valuable contribution
 here.

Looking more broadly at gaps in the field (just in case there's anyone
out there fancying a bite of the cherry - please do feel free to email
me directly if so); the role of partial information, habit-forming, and
satisficing must be built in somehow: we know that people do not behave
as utility-maximising agents possessing perfect information within an
equilibrium system; and future models will need to reflect that.

From what I've seen, space syntax offers very useful explanations of how
people interact with space, and that's something that traditional
transport models haven't addressed well, and haven't needed to address,
as spatial configuration has been someone else's field, and pedestrians
having been mostly ignored, at least until fairly recently.

Regards,
Andrew Smith
-- 
Director, London Analytics
T: [+44] (0)20 7627 8924
M: [+44] (0)791 046 0601
W: http://www.LondonAnalytics.info/