Tom Anderson wrote:
On Sun, 15 May 2005, Dave Arquati wrote:

Tom Anderson wrote:

On Sat, 14 May 2005, Dave Arquati wrote:

Tom Anderson wrote:

On Thu, 12 May 2005, Dave Arquati wrote:

A very impressive system would be to not only have accurate
Countdown information online and at stops, but also to have
dynamically-estimated journey times to destinations from that
stop, available both online and via Countdown at the stop itself.


Yes, this would be very cool, but the amount of effort it would
require to gather the traffic data, process it to produce
congestion forecasts (not entirely unlike weather forecasting -
congestion is a dynamic, nonlinear, mobile phenomenon), work out
delays to services and distribute this to every bus-stop would be
substantial.


I don't (think I) mean forecasting congestion, as such... just using
data on current traffic speeds to estimate those journey times.


Ah, but that does involve forecasting - if you want to know about
delays that a 38 at Victoria might suffer when it gets to Hackney,
you need to have some idea of what the traffic is going to be like
about 45 minutes into the future.


I don't think that's entirely necessary (although it would certainly
be impressive!). Knowing about existing delays on the route will give
vastly superior realtime information to that currently available - so
if congestion is already occurring in Hackney, it would be highlighted
at the 38 stop at Victoria, on the assumption that congestion tends to
clear slowly.

Okay. I'm dubious about this; it will warn people about problems that
won't affect them, and fail to warn them about problems which will.
However, it's better than nothing, which is what we have now.

My assumption is that the problems that will affect them will be relayed
to them as soon as those problems happen, but otherwise it's difficult
to predict such problems.

There's a case to be made that it's better than forecasting -
forecasting necessarily means telling people things you don't know to be
true, but are basically guesses. Telling them possibly irrelevant truths
is straightforward and honest, even if it does lumber them with more
working-out to do themselves.

One crucial quantitative factor is the volatility of congestion; if the
typical lifetime of congestion is longer than the length of the bus's
route, then it's worth telling peole about congestion that's occuring
now. If it's not, it isn't. And i don't know if it is.

I think it is - but I'd need to do some research to be sure (and I'm
supposed to be researching something else at the moment, so it may have
to wait!). Traffic delays seem to last for a long time after that which
caused the delay has disappeared. I found the following simulation
extremely entertaining:
http://vwisb7.vkw.tu-dresden.de/~tre.../simFrame.html

Perhaps we can say there are two types of traffic delay: (a) those due
to lack of sufficient network capacity, and (b) those due to
spontaneously-arising problems like vehicles blocking the road. The
former can be forecast by their nature, and can probably be forecast
some time in advance - to some extent, these will already be
incorporated into the bus timetable (e.g. rush hours). The latter cannot
be forecast, but once they occur, the delays they cause can be monitored
and the travelling public alerted.

On the flip side, it's quite likely TfL are already doing this
forecasting as part of their realtime congestion management work.


I wonder how they would do it. Weather forecasting involves predicting
movements of fronts and such (although I know very little about it!),
but congestion forecasting would seem more difficult as it can arise
much more spontaneously - e.g. if a lorry breaks down in an awkward
location.


True. However, congestion that's not related to accidents might be
predictable, and the degree to which an accident could generate
congestion (congestion potential, if you will) might be.

Ooh, "congestion potential"... I love it when people talk
scientifically... :-) I was just thinking that such network analysis is
probably similar to metabolic flux analysis in my field, and a standard
set of equations are probably applicable. Mind you, I'm sure this has
all been studied before.

Maybe I should make this my Masters project next year... I know a
couple of friendly computing/information systems students who might
like to collaborate!

Please do.

No, wait! I never uses buses - work on something to do with tubes or
bikes. Cheers.

The Tube's no fun; it either breaks down or it doesn't, there's no
delicious in-between you can model!

For bikes... perhaps an improved Journey Planner which knows traffic
levels on roads across the city, and allows you to specify a maximum
traffic level you'll cycle through, allowing you to plan routes to avoid
unpleasantly busy roads like the Euston Road. You could chuck in
pollution indices too if you like.

--
Dave Arquati
Imperial College, SW7
www.alwaystouchout.com - Transport projects in London