Mike,

Well, yes, in Axman you can construct two networks that link to each other
at specific points. We do this all the time to represent multi storey
buildings with multiple staircase connections. The same can be carried out
for different transport networks and this is what Shinichi has done.
However, there are two issues. First, Axman as it stands uses undirected
graphs, and this means that one way streets are not represented as such in
the graph. Second, Axman uses unweighted graphs, and so differences of speed
and capacity on transport networks and their comparison to the pedestrian
mode are not directly representable. In our studies of vehicular traffic in
London the oneway issue appeared to be of relatively little concern. My
guess is that this is because traffic managers use one way systems for two
main purposes. First, to increase capacity by keeping all flows in one
direction on one street and all in the opposite direction in a parallel
street in the grid (eg. Tottenham Court Road and Gower St in London). Thus
correlations between spatial integration and flows are not greatly affected
by the one way nature of the system. Second they use one way systems to
block possible rat runs through part of grid by making a single alignment
have sections with opposing one way flow restrictions (eg. Marylebone High
St.). This can be represented in the axial map by breaking the single line
into the respective one way segments. This seems to work in that
correlations between spatial configuration measures and observed flows are
good - r^2 >.8 . For practical purposes - use in design - the simplicity of
the representation and its explanatory capacity make this a useful tool,
however, I think the representation is somewhat inelegant, and a directed
graph version would be well worth testing out.

Alan 

Does this mean that you can now model the 'Tube' in space syntax. I thought
that this was one of the limitations of it as I have never seen any maps
where anything other than streets are simulated ? except in my Working Paper
80; and presumably by extension, this means that you can model interlocking
networks like a bus network which is separate from a street network in that
only at certain spots does the bus connect with the street at stops etc ? If
you can model different networks then the one way street problem can problem
be handled by modelling two street networks on top of one another which with
an ingenious choice of which network on which to indicate direction, one can
simulate the existence of directionality (although as the street network is
not a graph per se in the relational sense, this is simply a proxy for a
directed graph.

Mike

At 17:52 06/05/04 +0100, Alan Penn wrote:

Victor, Mike et al

The way one does this in Axman is to use the unlink and superlink
tools. This allows you to model freeways that pass over the street grid,
or underground transport networks that connect to the streetgrid only
intermittently at stations. Shinichi Iidas models of London and Tokyo
including both the street level pedestrian/vehicular system and the rail
transport networks have done this on the very large scale and work
rather well. You may need to wait for Shinichi to return from Japan to
get a more detailed review of his findings.

Alan Penn

-----Original Message-----
Subject: Re: What streets to include in axman

I am afraid I don't have any advice about how to handle this problem in
Axman but it is handled consistently in my paper "Distance in Space
Syntax" where I show how you can deal with two different networks such
as railway lines that only intersect with streets at infrequent points
but that this requires a different formalism from traditional space
syntax. Where the railway crosses under or over a street, it does not
normally intersect it. Unfortunately there is no public domain software
to compute this available as yet.

But see the paper at
http://www.casa.ucl.ac.uk/working_papers/paper80.pdf

and Figure 6 et seq in that papers show pictures of where the loop
railway goes under streets in central Melbourne. I think that these
kinds of problems cannot be handled consistently in traditional space
syntax because it requires different networks to be handled and this
means that we need to move to thinking of streets and intersections as
raw data where the Euclidean coincidence of one street with another does
not automatically imply a junction.

Mike

_________________________________________________
Michael Batty, Director, CASA, University College London, 
1-19 Torrington Place London WC1E 6BT, UK
[t] 44 (0) 207 679 1781, [f] 44 (0) 207 813 2843

Personal Homepage 
http://www.casa.ucl.ac.uk/people/MikesPage.htm
New CASA Working Papers at
http://www.casa.ucl.ac.uk/publications/full_list.htm
_________________________________________________