Distrbuted parameter sweep for UPPAAL models

Studenteropgave: Speciale (inkl. HD afgangsprojekt)

  • Brian Villumsen
  • Peter Schmidt Freiberg
  • Jimmy Merrild Krag
2. semester, Datalogi, Kandidat (Kandidatuddannelse)
The use of a tool when conducting a para-
meter sweep is essential, since the amount
of possible combination is often very large.
This thesis presents U PPAAL PARMOS,
a Parameter Sweep Application that dis-
tributes a parameter sweep of verifications
of a U PPAAL model to multiple resources,
while employing an optimization scheme
to prioritize individual model configura-
tions, thus allowing for faster access to de-
sired results. The thesis includes both a
fully described design of U PPAAL PAR -
MOS as well as an implementation and
test. Additional, a study of heuristic op-
timization algorithms has been conducted
in order to utilize three of these, specifi-
cally Hill-Climbing, Simulated Annealing,
and Pareto Archived Evolution Strategy,
in U PPAAL PARMOS. Based on experi-
ments on our implementation we conclude
that there are grounds for utilizing U PPAAL
PARMOS to acquire faster access to desired
Udgivelsesdato1 jun. 2011
Antal sider106
ID: 52712058