On Wed, 20 Aug 2003 08:37:30 +0100, Steve Fitzgerald
] wrote:

In message , Rupert Goodwins
writes

I can't see why the ability to change the destination of trains after
the start of service should have to be bought at the cost of four
hours' extra work. After all, the Northern Line has to reconfigure
itself frequently because of signalling, stock or other problems:

That will be service recovery though - slightly different as they are
trying to get the service back 'on book'.


Absolutely, but it must involve many of the same problems. As the
original poster noted, service recovery isn't particularly efficient
at the moment, so work here may have immediate benefits.

The main problem I can foresee with your suggestion is that the trains
will end up far away from where they are scheduled to be, thus away from
the relieving driver. You would find more instances of a train having
no driver and having to be put away in a depot or sidings. This would
cause the service to deteriorate even more.

So you don't let that happen.

Working out the correct combination of destination and driver changes
so that everyone is happy, is a very difficult -- classicly so --
mathematical problem (probably related to the travelling salesman, but
I haven't thought that through). You may well have to do a brute-force
search through the solution space, which is the sort of chess-game
approach that even five years ago would seem hopelessly time
consuming. But as we now have PCs that can do close on ten billion
calculations a second, I'd imagine that it's the sort of project one
motivated chap could sensibly attack. Even if it wasn't possible to
produce a working system in the first case, a partial simulation to
prove the concept would be tempting.

The Northern Line has a great advantage over mainline services in that
the passengers don't need to know the timetable. They need to know
first and last trains, and that at any particular time there will be a
train within x minutes going to their destination. Internally, of
course, there has to be a detailed timetable, but because that doesn't
need to be public the line managers have the huge potential advantage
of being able to make as many changes as they like during the day to
maintain that level of service, without involving the public.

They are constrained by the need to have all the trains back at the
right place at the end of play (although this also can be flexible, to
an extent), and of ensuring the drivers also end up where they need to
be when they need to be there.

Very hard. But the benefits to the efficiency of the service would be
considerable -- and it's not unreasonable to see that this level of
flexibility could be of great benefit to staff, as it would allow much
better handling of problems during the day and also allow a much more
flexible scheme of driver rostering in the first place.

At this point, I run out of steam as I don't know how driver rostering
works, or many of the other requirements of running the NL! It would
be very interesting if this information was available somewhere, to
see whether such a flight of fantasy had legs. Or wheels, or whatever.

R