Sebastian Junges
d144a1776c
prism next state generator, rename actions to reflect labeled choices vs unlabeled choices are sync vs async
5 years ago
Sebastian Junges
3f165785d0
integrated feedback on code quality by TQ
5 years ago
Sebastian Junges
922d6f6572
export of compressed states to json with negative values fixed
5 years ago
Tim Quatmann
bdd89d87b2
Prism next state generator now deals with unbounded integer variables.
5 years ago
Tim Quatmann
d5c6a509a2
JaniNextStateGenerator: Fixed evaluation of terminal states using expressions over transient variables
5 years ago
Tim Quatmann
66e6938d20
added a few clarifying comments in JaniNextStateGenerator
5 years ago
Sebastian Junges
08f928456c
fix guard for code that considers transient assignments to also consider only transient assignments
5 years ago
Sebastian Junges
6b6f44100e
allow building parametric models of the form s --p-->, s--q-->
5 years ago
Sebastian Junges
71f60e812c
more precise analysis of whether commands will synchronize
5 years ago
Sebastian Junges
c1ec3032fa
reset to state
5 years ago
Stefan Pranger
44378ac9a1
fix for playerIndex in Choice
5 years ago
Stefan Pranger
972df05683
store tuples of player name and index
Store this instead of only the index. Needed for easier parsing of the
rpatl formulas (prism allows player indices and names!)
5 years ago
Tim Quatmann
8b68fbf948
JaniBuilder: Fixed checks for transient variable assignments
5 years ago
Tim Quatmann
875410a59e
Polished ExplicitModelBuilder:
* ChoiceInformationBuilder renamed to StateAndChoiceInformationBuilder, now also keeping track of state-based information (StateValuations, MarkovianStates, statePlayerIndications)
* ModelComponents now consider statePlayerIndications and PlayerNamesToIndices separately
5 years ago
Tim Quatmann
277f802850
* PlayerIndex is now declared in a separate file (as this can potentially be independent of PRISM input).
* Polished PrismNextStateGenerator, in particular more proper error handling
5 years ago
Tim Quatmann
97b2d751e0
* prism::Player's no longer keep track of module and action indices to reduce redundancies.
* PrismProgram::CheckValidity and PrismProgram::simplify now treat SMGs properly
* PrismProgram is now responsible for moduleIndex->playerIndex and actionIndex->playerIndex assignment
* More defined behavior for actions that don't have a player (work in progress)
5 years ago
Tim Quatmann
6fe76a009d
Polished parsing of Prism-SMGs, in particular
* Fixed issues related to module renaming that resulted from setting the module indices already in the first run
* Fixed a few uint_fast32_t vs uint_fast64_t issues, created alias PlayerIndex
5 years ago
Sebastian Junges
4a7ea35959
first version for action mask callbacks in explicit generator
5 years ago
Sebastian Junges
7a38f54d01
extend the next state generator to support prism program simulation
5 years ago
Sebastian Junges
aa6a3d2142
sampling from a distribution and from a choice
5 years ago
Stefan Pranger
c11c49d22b
store tuples of player name and index
Store this instead of only the index. Needed for easier parsing of the
rpatl formulas (prism allows player indices and names!)
5 years ago
Tim Quatmann
3184ba1611
Jani: Correctly parse the input-enable field. Throw an error in the sparse model builder, as these are not supported right now.
5 years ago
Tim Quatmann
6d6e142236
Fixed an issue with JANI models concerning properties using transient variable expressions.
5 years ago
Stefan Pranger
cea09f932b
generator now assigns player indices to states
5 years ago
Stefan Pranger
6b715792aa
added player related helpers
5 years ago
Stefan Pranger
55f4efd40a
added SMG ModelType
5 years ago
Stefan Pranger
d35e9a6a40
removed plenty of empty line whitespaces
5 years ago
Sebastian Junges
9423d01631
observation valuations added
5 years ago
Sebastian Junges
923f779a09
explicit-state-lookup, for finding states in a model based on the variable assignment
5 years ago
Tim Quatmann
5b103579f2
Fixed building state valuations for transient variables of a jani model.
6 years ago
Tim Quatmann
adfdf8c572
Refactored state valuations. They now store values for transient jani variables and do not store values for constants (solving Github issue #73 )
6 years ago
Tim Quatmann
3a86cc4391
CompressedState: Added a method to create a human readable string out of the state. Added a method to "uncompress" by extracting all values into corresponding value vectors
6 years ago
Tim Quatmann
5733072898
TransientVariableInformation: Flagging a few getters const
6 years ago
Matthias Volk
d3c8093e0f
Removed unnecessary semicolons
6 years ago
Tim Quatmann
7e9029e5bd
Optimization for PRISM model building: Avoid evaluating unnecessarily many guards.
6 years ago
Tim Quatmann
05471d94fd
Optimizations in JaniNextStateGenerator that avoid unnecessary (re-) allocations of memory. Moreover, before collecting the complete set of enabled edges for each automaton, we first check whether each automaton (that synchronizes with that action) has at least one enabled edge. This avoids checking unnecessarily many edge guards.
6 years ago
Tim Quatmann
a5d3d0e696
slight optimizations in the JaniNextStateGenerator
6 years ago
Tim Quatmann
bb3f7c52fd
DdJaniModelBuilder: Fixed canHandle
6 years ago
Tim Quatmann
23fb3bedff
all model builders: Added a canHandle method and a getSupportedJaniFeatures method.
6 years ago
Sebastian Junges
b554dabaab
state generator now takes into account observable expressions when building POMDPs
6 years ago
Tim Quatmann
cb00c21db2
Fixed type uintX -> uintX_t
6 years ago
Matthias Volk
6a4c18e4a2
Use custom FlatSet to account for allocator changes in flat_set in Boost 1.70.
Boost 1.70 changed the default allocator parameter from new_allocator<T> to void to reduce symbol lenghts.
This reverts the default to the old allocator.
7 years ago
Tim Quatmann
c8ea0f60da
JaniBuilder: Fixed several issues that occurred with branch reward expressions over non-transient variables, including GitHub issue #47
7 years ago
Jip Spel
a35cb2643a
Extend error message
7 years ago
TimQu
33127c9b6e
JaniNextStateGenerator: Fixed references to the unpreprocessed model.
7 years ago
TimQu
02977da3d7
Apply maximum progress assumption while building a Markov Automaton explicitly.
7 years ago
TimQu
c622f463ad
JaniNextStateGenerator: Fixed an issue related to CTMCs without state-action rewards
7 years ago
TimQu
ca828729ff
Fixed a few warnings
7 years ago
TimQu
0a6122258c
Used the new reward information traverser wherever one needs to find out the reward kinds of a given rewardmodel
7 years ago
TimQu
86d99d8ca5
silenced a few warnings
7 years ago