There is not really a route choice assumption about simplest routes in
space syntax, although we sometimes talk as though there were. One reason
there is not, is that there is an obvious counter example: moving
diagonally accros a  grid. The more you try to approximate a straight line
between where you are and where you think your destination is, the more
turns you will tend to make. It is more interesting to see linearity as
being involved in planning pedestrian routes, which are relatively small
scale, in two ways: first we think of the relation between where we are and
where we want to go as a an 'as the crow flies' line; then we try to use
line structure of the intervening grid to approximate that. Ruth Conroy
makes a very intersting suggestions about this: that we try initially to
minimise the deflection from the imaginary line to the real line we choose,
and it may be that we continually correct this on this way. This seemed to
be a key factor in my own consistent navigational errors I report in 'The
architectures of seeing and going' at the last Symposium. Mark Major also
made an interesting suggestion about 'diagonalising' grids: that by
diagnalising road segment we reduce angle change  and accumulate small
distance savings compared the often simpler route around the edges. With
larger scale vehicular movement it is often quite different, and time will
be much more involved. I used to go to Tunbridge Well from central London
via the M25 - a substantial detour from the imagined line, but which halved
the journey time.

On the more general issues, I am surprised by the competitive tone that
seems to be coming in. Syntax models in no way seek to emulate what
transport models already do very efficiently, and there is no competition
about who has the best correlations. Space syntax originated in
architecture, with an initial focus on the micro scale, and architectural
questions like: what will happen if we do it this way ? Design is an
activity which is highly informed by theoretical ideas and assumptions, and
what space syntax has shown is that in and of itself spatial configuration
shapes movement patterns, and that by understanding this we can understand
how the well known macro-scale linkages between accessibility, movement and
landuse, can be shown also to operate at the micro scale, and play a key
role in creating the intricate patterns of local movement and land use
differentiation that that we find in cities, which help make cities as
pleasureable as they are. If we wanted to compete with traffic models, then
of course we can add all the other factors into the model and improve the
correlations, and indeed in some circumstances - usually where the natural
relations between accessibility, movement and land use have been ruptured
by insensitive design and planning - we do exactly this through the
Walkability Index. But the fundamental point that syntax is making is a
theoretical one: that spatial configuration shapes movement, and this in
turn affects the life patterns in cities, and so needs to be part of a
theory of the city. The fact that so much can be done with such a simple
representation ought also be a matter of theoretical interest to all those
who are trying to understand why cities are as they are and how they work.


- Bill









>Dear readers,
>
>what follows is something of a stream of consciousness on transport
>modelling. Those wishing to discuss only space syntax, and not its
>relationship to transport modelling, will be happy to know that they
>can hit the delete key now and miss nothing of consequence to them.
>
>On 9 May 2004 at 17:56, Alain Chiaradia wrote:
>
>> What Space Syntax research show is that network morphology itself -
>> network characteristic - is strongly "inducing / generating" traffic or
>> "in relationship to relative" traffic volume level
>
>How did we get so quickly from observed correlation to established
>strong cause and effect? Possible causality I'll accept. Let's go on.
>
>As an aside: remember, what transport surveys show is that *trips*
>are not generated or induced by the network. This is borne out by the
>oft (mis-)quoted SACTRA report on trunk roads and traffic generation.
>
>Network configuration and many other factors do, however, affect
>choices of origin, destination, mode and route, and hence, flows on
>particular links by specific modes.
>
>One systematic problem that many traffic models (included such golden
>oldies as TRIPS, EMME2 and SATURN) usually have is in car route
>choice, and in particular modelling the unusually low flows on minor
>roads compared to main roads. Drivers often do not take the quickest
>route across large, complex networks. As I understand it, there is
>implicitly a route-choice algorithm in space syntax: that is, the
>route which involves fewest changes of direction. Now, there are
>other competitors to the (often failing) quickest-route algorithm -
>see for example Daly et al's hierarchical road algorithm: which
>identifies that a car's route typically goes: local road -> local
>distributor -> trunk road -> local distributor -> local road. I
>expect that in many cases both of these do better than the old
>standard. I know that Daly's algorithm does in some specific cases
>I've seen.
>
>Now, obviously, in terms of resulting routes chosen, there will
>generally be lots of overlap between these different algorithms. That
>is, the quickest route will feature this arching pattern across the
>road hierarchy, and involve traversing fewest axial lines.
>
>NB we get good correlations between a generalised cost based on
>travel time & distance, and mode and destination choices: not only
>that, but it's easy to understand the causal mechanism there, and it
>provides scope to model the effects of policy levers.
>
>Imagine a network with two routes - one involving fewest changes of
>direction, two hours long; another involving more changes, but only
>30 minutes long. Do you believe that time and distance would not
>affect route choice? I expect that you can find examples of similar
>choices in much of the public transport network.
>
>Many models using these principles have been proven by experience.
>Others have failed, and by looking at the successes and failures, I
>have come to my own personal conclusion that the principles are
>sound, but implementations vary in quality. Your mileage may vary.
>
>> After all, most transportation modeller whatever they say, most of the
>> times go by what available software that is recognised as accepted
>> standard practice from the paymaster even if the modeller would
>> customize it to the latest flavour of the day.
>
>I'm interested to know - experience of which countries' transport
>modelling leads you to this conclusion? I can think of quite a few
>counter-examples in most countries where I have a little experience.
>[Though perhaps that's more to do with my career trajectory than with
>general practice in those places. ;-) ]
>
>Andrew Smith
>(writing in a personal capacity)