Verimag@Ametist

courageouscellistΤεχνίτη Νοημοσύνη και Ρομποτική

29 Οκτ 2013 (πριν από 3 χρόνια και 11 μήνες)

81 εμφανίσεις

Verimag@Ametist

Oded Maler


December 2002

Major Work Directions


SAT Solving for Difference Logic


Axxom case
-
study


Scheduling with Timed Automata

SAT for Difference Logic


The first prototype of the solver has been
implemented (Moez Mahfoudh,
collaboration with E. Asarin, M Bozga and
P. Niebert (Marseille)) with some
translations from TA and other problems.


Expected breakthrough: April 2003 or
never..

Axxom Case
-
Study


Two directions:


Treat the case
-
study itself


Marius Bozga
using IF


Reach it gradually as extending the basic
job
-
shop problem (see later)


Expected results: Mars 2003

Scheduling Framework:
Principles


Start with the cleanest problem (job
-
shop)


Model it with timed automata


Exploit the specific structure of the problem
to have better algorithms


Add general
-

and special
-
purpose heuristics


Test on benchmarks


Move on to more complex problems


(Thesis of Yasmina Abdeddaim, Nov 2002)

So far it Works!


Job
-
shop problem (in the VHS project):
discovery of the fact that for this problem
you do not need the zone technology!


In fact you do not need it in a large class of
problems


Investigation of “intelligent” search
techniques


Respectable performance results

Extensions


Job
-
shop with preemption (Ametist kick
-
off
meeting): similar resuls, no zones,
undecidability of stopwatch becomes
irrelevant, good performance results


Task graph scheduling: partially
-
ordered
tasks on parallel identical machines (with
release times and deadlines)


talk by A.
Kerbaa this meeting

Further Extensions


Different machines with different speeds for
each product


a preparation for the value
chain problem (talk by A Kerbaa)


Parallel threads with resources and
synchronization (preliminary work)


Non monotonicity (relative deadlines)


Scheduling under uncertainty (my talk)