Hard Timetables

From: Nicholas Robinson <nprREMOVE@THISbottlehall.co.uk>
Date: Tue Jun 07 2005 - 23:38:49 CEST


Despite the best attempts of the new algorithm, a solution to my timetable as
a whole remains elusive - but read on!

Paring down the problem, it seems that it is the high incidence of double
lessons combined with zero free lessons for classes and very few free lessons
for teachers. Of these, it always seems to be the doubles (especially those
that are 'same-time-as' another class or three) that contribute the largest
number of contributions to the fitness score.

To test, I ran a timetable with only the doubles and one single lesson that is
attended by four forms/classes. This leaves about 25-30% of our timetable.
Tablix didn't solve even this reduced problem. As far as I can tell, changing
the weights doesn't change the outcome although it certainly affects the type
of fitness problems and the speed at which it reaches the minimum. I then
decided to try one more simplification. Since my timetable was, with one
exception, double lessons, why not 'halve' my timetable? Reduce the doubles
to singles and reduce the number of lessons per day to 4 from 9. It solves
this problem in less than a minute with zero fitness every time!

So, if I can schedule the doubles as just described and then somehow alter the
configuration file for the remaining lessons to include the doubles as fixed
then I think I will be able to use tablix for real.

Would the following work?

1. Use the timetabled events in the result XML file as part of the input file
for the second run but replace the "time" resources that specify the place in
the timetable with a restriction such as 'fixed-period' - obviously keeping
the day/period values unchanged.

2. The 'fixed-period' handler would be implemented so that it removed all
other time, class, room and teacher resources in the resource domains for the
appropriate event tuple.

3. Since no other alternatives could ever be generated, these lessons would
remain fixed - and at no real processing load, as far as I can see.

Best Wishes

Received on Tue Jun 07 23:38:30 2005

This archive was generated by hypermail 2.1.8 : Tue Aug 16 2005 - 20:43:54 CEST