Question regarding Oracle action costs (ADAPT-Lite)

[Jeremy Mange]

For the ADAPT-Lite problem, we were running some tests for detecting
offset errors, and are confused about the Oracle's reported costs. In
the first sample scenario (1081_pb), our algorithm was able to
identify the fault shortly after it occurred. We hard-coded the fault
parameter (Offset) for this scenario for the moment.

Our basic strategy was to send two RecoveryData messages to the
Oracle, one with the ABORT command, and one with the NOP command, and
see which cost is reported as lower to choose which action to take.
In this case, the Oracle reported a cost of 125 for NOP and 0 for
ABORT (see below), so the algorithm sent the ABORT command. However,
this must have been the wrong choice, since the Evaluator recorded a
Recovery cost of 25 with that choice.

Through some testing, it seems that in this scenario the Oracle
"recommends" aborting for any value <= -1.5 for the Offset. But it
seems that the Evaluator considers that the wrong choice for the
scenario, which has an injected fault Offset value of -2.5.

____________________________________________________________________________

Selected lines from the results file:

faultInjection @1277404627315 <SDS> isInjection = true, fault =
{ DC485 = ResistanceOffset [ Offset = ' -2.5' ] };
...
diagnosis @1277404627822 <DA> detection = true, isolation = true,
candidates = { { DC485 = ResistanceOffset [ FaultTime = 59969.0,
Offset = -1.5 ] } [weight = 1] }, notes = 'notes';

recoveryInfo @1277404627826 <DA> faults = { DC485 = ResistanceOffset
[ FaultTime = 59969.0, Offset = -1.5 ] }, commands = { systemAction =
'NOP' }, cost = -1, notes = '';
recoveryInfo @1277404627827 <OR> faults = { DC485 = ResistanceOffset
[ FaultTime = 59969.0, Offset = -1.5 ] }, commands = { systemAction =
'NOP' }, cost = 125, notes = '';

recoveryInfo @1277404627860 <DA> faults = { DC485 = ResistanceOffset
[ FaultTime = 59969.0, Offset = -1.5 ] }, commands = { systemAction =
'ABORT' }, cost = -1, notes = '';
recoveryInfo @1277404627861 <OR> faults = { DC485 = ResistanceOffset
[ FaultTime = 59969.0, Offset = -1.5 ] }, commands = { systemAction =
'ABORT' }, cost = 0, notes = '';
...
command @1277404807900 <DA> { systemAction = 'ABORT' };
____________________________________________________________________________

From Evaluator output:

diagnosis @1277404627822 <DA> detection = true, isolation = true,
candidates = { { DC485 = ResistanceOffset [ FaultTime = 59969.0,
Offset = -1.5] } [weight = 1.0] }, notes = 'notes';

faultInjection @1277404627315 <SDS> isInjection = true, fault =
{ DC485 = ResistanceOffset [ FaultTime = 59470.0, Offset = ' -2.5']};

Detection: True Positive.

Recovery Cost: 25.0

[Sriram]

Hi Jeremy,

This does look like a possible bug. We are investigating this and
should have an answer soon.

Sriram