sp
/
tempest
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Browse Source

more on total reward formulas for mdps

tempestpy_adaptions
TimQu 7 years ago
parent
b3edae8707
commit
b5566fa861
2 changed files with 6 additions and 3 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 1
      src/storm/modelchecker/prctl/SparseMdpPrctlModelChecker.h
  2. 8
      src/storm/modelchecker/prctl/helper/SparseMdpPrctlHelper.cpp

1
src/storm/modelchecker/prctl/SparseMdpPrctlModelChecker.h
View File

@ -27,6 +27,7 @@ namespace storm {
virtual std::unique_ptr<CheckResult> computeConditionalProbabilities(Environment const& env, CheckTask<storm::logic::ConditionalFormula, ValueType> const& checkTask) override;
virtual std::unique_ptr<CheckResult> computeCumulativeRewards(Environment const& env, storm::logic::RewardMeasureType rewardMeasureType, CheckTask<storm::logic::CumulativeRewardFormula, ValueType> const& checkTask) override;
virtual std::unique_ptr<CheckResult> computeInstantaneousRewards(Environment const& env, storm::logic::RewardMeasureType rewardMeasureType, CheckTask<storm::logic::InstantaneousRewardFormula, ValueType> const& checkTask) override;
virtual std::unique_ptr<CheckResult> computeTotalRewards(Environment const& env, storm::logic::RewardMeasureType rewardMeasureType, CheckTask<storm::logic::TotalRewardFormula, ValueType> const& checkTask) override;
virtual std::unique_ptr<CheckResult> computeReachabilityRewards(Environment const& env, storm::logic::RewardMeasureType rewardMeasureType, CheckTask<storm::logic::EventuallyFormula, ValueType> const& checkTask) override;
virtual std::unique_ptr<CheckResult> computeLongRunAverageProbabilities(Environment const& env, CheckTask<storm::logic::StateFormula, ValueType> const& checkTask) override;
virtual std::unique_ptr<CheckResult> computeLongRunAverageRewards(Environment const& env, storm::logic::RewardMeasureType rewardMeasureType, CheckTask<storm::logic::LongRunAverageRewardFormula, ValueType> const& checkTask) override;

8
src/storm/modelchecker/prctl/helper/SparseMdpPrctlHelper.cpp
View File

@ -846,7 +846,7 @@ namespace storm {
MDPSparseModelCheckingHelperReturnType<ValueType> SparseMdpPrctlHelper<ValueType>::computeTotalRewards(Environment const& env, storm::solver::SolveGoal<ValueType>&& goal, storm::storage::SparseMatrix<ValueType> const& transitionMatrix, storm::storage::SparseMatrix<ValueType> const& backwardTransitions, RewardModelType const& rewardModel, bool qualitative, bool produceScheduler, ModelCheckerHint const& hint) {
// Reduce to reachability rewards
if (goal.minimize) {
if (goal.minimize()) {
STORM_LOG_ERROR_COND(!produceScheduler, "Can not produce scheduler for this property (functionality not implemented");
// Identify the states from which no reward can be collected under some scheduler
storm::storage::BitVector choicesWithoutReward = rewardModel.getChoicesWithZeroReward(transitionMatrix);
@ -883,7 +883,7 @@ namespace storm {
return computeReachabilityRewardsHelper(env, std::move(goal), ecElimResult.matrix, ecElimResult.matrix.transpose(true),
[&] (uint_fast64_t rowCount, storm::storage::SparseMatrix<ValueType> const& transitionMatrix, storm::storage::BitVector const& maybeStates) {
std::vector<ValueType> result;
std::vector<ValueType> oldChoiceRewards = rewardModel.getTotalRewardVector(rowCount, transitionMatrix);
std::vector<ValueType> oldChoiceRewards = rewardModel.getTotalRewardVector(transitionMatrix);
result.reserve(rowCount);
for (uint64_t newState : maybeStates) {
for (uint64_t newChoice = transitionMatrix.getRowGroupIndices()[newState]; newChoice < transitionMatrix.getRowGroupIndices()[newState + 1]; ++newChoice) {
@ -904,7 +904,7 @@ namespace storm {
[&] () {
storm::storage::BitVector newChoicesWithoutReward(ecElimResult.matrix.getRowGroupCount(), false);
for (uint64_t newChoice = 0; newChoice < ecElimResult.matrix.getRowCount(); ++newChoice) {
if (choicesWithoutReward.get(ecElimResult.newToOldChoiceMapping[newChoice])) {
if (choicesWithoutReward.get(ecElimResult.newToOldRowMapping[newChoice])) {
newChoicesWithoutReward.set(newChoice);
}
}
@ -1761,6 +1761,7 @@ namespace storm {
template std::vector<double> SparseMdpPrctlHelper<double>::computeInstantaneousRewards(Environment const& env, storm::solver::SolveGoal<double>&& goal, storm::storage::SparseMatrix<double> const& transitionMatrix, storm::models::sparse::StandardRewardModel<double> const& rewardModel, uint_fast64_t stepCount);
template std::vector<double> SparseMdpPrctlHelper<double>::computeCumulativeRewards(Environment const& env, storm::solver::SolveGoal<double>&& goal, storm::storage::SparseMatrix<double> const& transitionMatrix, storm::models::sparse::StandardRewardModel<double> const& rewardModel, uint_fast64_t stepBound);
template MDPSparseModelCheckingHelperReturnType<double> SparseMdpPrctlHelper<double>::computeReachabilityRewards(Environment const& env, storm::solver::SolveGoal<double>&& goal, storm::storage::SparseMatrix<double> const& transitionMatrix, storm::storage::SparseMatrix<double> const& backwardTransitions, storm::models::sparse::StandardRewardModel<double> const& rewardModel, storm::storage::BitVector const& targetStates, bool qualitative, bool produceScheduler, ModelCheckerHint const& hint);
template MDPSparseModelCheckingHelperReturnType<double> SparseMdpPrctlHelper<double>::computeTotalRewards(Environment const& env, storm::solver::SolveGoal<double>&& goal, storm::storage::SparseMatrix<double> const& transitionMatrix, storm::storage::SparseMatrix<double> const& backwardTransitions, storm::models::sparse::StandardRewardModel<double> const& rewardModel, bool qualitative, bool produceScheduler, ModelCheckerHint const& hint);
template std::vector<double> SparseMdpPrctlHelper<double>::computeLongRunAverageRewards(Environment const& env, storm::solver::SolveGoal<double>&& goal, storm::storage::SparseMatrix<double> const& transitionMatrix, storm::storage::SparseMatrix<double> const& backwardTransitions, storm::models::sparse::StandardRewardModel<double> const& rewardModel);
template double SparseMdpPrctlHelper<double>::computeLraForMaximalEndComponent(Environment const& env, OptimizationDirection dir, storm::storage::SparseMatrix<double> const& transitionMatrix, storm::models::sparse::StandardRewardModel<double> const& rewardModel, storm::storage::MaximalEndComponent const& mec);
template double SparseMdpPrctlHelper<double>::computeLraForMaximalEndComponentVI(Environment const& env, OptimizationDirection dir, storm::storage::SparseMatrix<double> const& transitionMatrix, storm::models::sparse::StandardRewardModel<double> const& rewardModel, storm::storage::MaximalEndComponent const& mec);
@ -1771,6 +1772,7 @@ namespace storm {
template std::vector<storm::RationalNumber> SparseMdpPrctlHelper<storm::RationalNumber>::computeInstantaneousRewards(Environment const& env, storm::solver::SolveGoal<storm::RationalNumber>&& goal, storm::storage::SparseMatrix<storm::RationalNumber> const& transitionMatrix, storm::models::sparse::StandardRewardModel<storm::RationalNumber> const& rewardModel, uint_fast64_t stepCount);
template std::vector<storm::RationalNumber> SparseMdpPrctlHelper<storm::RationalNumber>::computeCumulativeRewards(Environment const& env, storm::solver::SolveGoal<storm::RationalNumber>&& goal, storm::storage::SparseMatrix<storm::RationalNumber> const& transitionMatrix, storm::models::sparse::StandardRewardModel<storm::RationalNumber> const& rewardModel, uint_fast64_t stepBound);
template MDPSparseModelCheckingHelperReturnType<storm::RationalNumber> SparseMdpPrctlHelper<storm::RationalNumber>::computeReachabilityRewards(Environment const& env, storm::solver::SolveGoal<storm::RationalNumber>&& goal, storm::storage::SparseMatrix<storm::RationalNumber> const& transitionMatrix, storm::storage::SparseMatrix<storm::RationalNumber> const& backwardTransitions, storm::models::sparse::StandardRewardModel<storm::RationalNumber> const& rewardModel, storm::storage::BitVector const& targetStates, bool qualitative, bool produceScheduler, ModelCheckerHint const& hint);
template MDPSparseModelCheckingHelperReturnType<storm::RationalNumber> SparseMdpPrctlHelper<storm::RationalNumber>::computeTotalRewards(Environment const& env, storm::solver::SolveGoal<storm::RationalNumber>&& goal, storm::storage::SparseMatrix<storm::RationalNumber> const& transitionMatrix, storm::storage::SparseMatrix<storm::RationalNumber> const& backwardTransitions, storm::models::sparse::StandardRewardModel<storm::RationalNumber> const& rewardModel, bool qualitative, bool produceScheduler, ModelCheckerHint const& hint);
template std::vector<storm::RationalNumber> SparseMdpPrctlHelper<storm::RationalNumber>::computeLongRunAverageRewards(Environment const& env, storm::solver::SolveGoal<storm::RationalNumber>&& goal, storm::storage::SparseMatrix<storm::RationalNumber> const& transitionMatrix, storm::storage::SparseMatrix<storm::RationalNumber> const& backwardTransitions, storm::models::sparse::StandardRewardModel<storm::RationalNumber> const& rewardModel);
template storm::RationalNumber SparseMdpPrctlHelper<storm::RationalNumber>::computeLraForMaximalEndComponent(Environment const& env, OptimizationDirection dir, storm::storage::SparseMatrix<storm::RationalNumber> const& transitionMatrix, storm::models::sparse::StandardRewardModel<storm::RationalNumber> const& rewardModel, storm::storage::MaximalEndComponent const& mec);
template storm::RationalNumber SparseMdpPrctlHelper<storm::RationalNumber>::computeLraForMaximalEndComponentVI(Environment const& env, OptimizationDirection dir, storm::storage::SparseMatrix<storm::RationalNumber> const& transitionMatrix, storm::models::sparse::StandardRewardModel<storm::RationalNumber> const& rewardModel, storm::storage::MaximalEndComponent const& mec);

Write Preview
Loading…
Cancel
Save