Issue237

Title possible translator bug in airport
Priority critical Status resolved
Superseder Nosy List erez, gabi, malte, mkatz
Assigned To malte Keywords
Optional summary

Created on 2011-04-27.10:33:46 by erez, last changed by malte.

Files
File name Uploaded Type Edit Remove
domain.pddl malte, 2011-04-27.13:18:54 application/octet-stream
output.sas.new erez, 2011-04-27.10:34:03 application/octet-stream
output.sas.old erez, 2011-04-27.10:33:46 application/x-trash
problem.pddl malte, 2011-04-27.13:23:07 application/octet-stream
sas_plan.new erez, 2011-04-27.10:34:35 application/octet-stream
sas_plan.old erez, 2011-04-27.10:34:19 application/x-trash
Messages
msg1343 (view) Author: malte Date: 2011-04-27.22:55:32
I've integrated this into the fixes branch and informed Carlos.

List of potentially affected domains (those that use :constants):

airport
openstacks-opt08-strips
openstacks-sat08-strips
parcprinter-08-strips
pathways
pathways-noneg
pipesworld-notankage
pipesworld-tankage
psr-large
psr-middle
schedule
woodworking-opt08-strips
woodworking-sat08-strips
msg1342 (view) Author: malte Date: 2011-04-27.19:56:55
OK, fix merged and pushed.
Michael and Erez, thanks for reporting this, and Gabi, thanks for fixing! :-)

Now we need to check which domains are affected, integrate this into the IPC
fixes branch, notify Carlos, and notify everyone else who's working with our
translator.
msg1341 (view) Author: gabi Date: 2011-04-27.19:54:33
Yes, that's guaranteed because the equivalence classes are represented as sets.
The elements are variables and constants represented as string.
msg1340 (view) Author: malte Date: 2011-04-27.19:46:43
Thanks, Gabi!

In the case that there are two element in "constants" in that method, is it
actually guaranteed that they are two *different* constnats (i.e., not two
copies of the same one)?
msg1339 (view) Author: gabi Date: 2011-04-27.14:55:32
This was a bug in the invariant synthesis that could only occur when constants
were used in the action specification.
msg1338 (view) Author: malte Date: 2011-04-27.13:23:07
OK, attached a much simplified test case (still big, but much much easier to
understand). With this test case, the old translator gives a solution of length
39, while the new one completely explores the state space up to depth 69,
failing to find a solution.

The new encoding (as per test.groups) appears to claim that at most one airplane
can be facing north at a time. If I interpret the plan from the old encoding
correctly, this is wrong, so this might be the source of the problem.

Passing over to Gabi.

PS: Gabi: Using invariant_finder.py as a top-level script seems to be broken.
msg1337 (view) Author: malte Date: 2011-04-27.11:56:36
OK, all confirmed. Will look into it.
msg1336 (view) Author: erez Date: 2011-04-27.11:26:34
I used the version in the SVN trunk.
msg1335 (view) Author: malte Date: 2011-04-27.10:37:48
Adding Gabi. I can confirm the length-75 solution.
Erez, which version of the translator did you use for the "old" result?
msg1334 (view) Author: erez Date: 2011-04-27.10:35:12
I should not this was found by Michael.
msg1333 (view) Author: erez Date: 2011-04-27.10:33:46
In airport-p09, using the new translator and then astar with blind search finds 
a solution of length 75.
Using the old translator (and the new preprocessor and search) finds a solution 
of length 71.
Both solution are validated by VAL.
History
Date User Action Args
2011-04-27 22:55:32maltesetstatus: reviewing -> resolved
messages: + msg1343
title: Possible Translator Bug in Airport -> possible translator bug in airport
2011-04-27 19:56:55maltesetmessages: + msg1342
2011-04-27 19:54:34gabisetmessages: + msg1341
2011-04-27 19:46:43maltesetmessages: + msg1340
2011-04-27 14:55:32gabisetstatus: chatting -> reviewing
assignedto: gabi -> malte
messages: + msg1339
2011-04-27 13:23:08maltesetfiles: + problem.pddl
assignedto: malte -> gabi
messages: + msg1338
2011-04-27 13:18:54maltesetfiles: + domain.pddl
2011-04-27 11:56:36maltesetassignedto: malte
messages: + msg1337
2011-04-27 11:51:35maltesetpriority: bug -> critical
2011-04-27 11:26:35erezsetmessages: + msg1336
2011-04-27 10:37:48maltesetnosy: + gabi
messages: + msg1335
2011-04-27 10:35:12erezsetnosy: + mkatz
messages: + msg1334
2011-04-27 10:34:35erezsetfiles: + sas_plan.new
2011-04-27 10:34:19erezsetfiles: + sas_plan.old
2011-04-27 10:34:03erezsetfiles: + output.sas.new
2011-04-27 10:33:46erezcreate