You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
 
 
 
 
 
 

1337 lines
43 KiB

//==============================================================================
//
// Copyright (c) 2002-
// Authors:
// * Dave Parker <david.parker@comlab.ox.ac.uk> (University of Oxford)
//
//------------------------------------------------------------------------------
//
// This file is part of PRISM.
//
// PRISM is free software; you can redistribute it and/or modify
// it under the terms of the GNU General Public License as published by
// the Free Software Foundation; either version 2 of the License, or
// (at your option) any later version.
//
// PRISM is distributed in the hope that it will be useful,
// but WITHOUT ANY WARRANTY; without even the implied warranty of
// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
// GNU General Public License for more details.
//
// You should have received a copy of the GNU General Public License
// along with PRISM; if not, write to the Free Software Foundation,
// Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
//
//==============================================================================
package explicit;
import java.io.File;
import java.util.BitSet;
import java.util.List;
import explicit.rewards.ConstructRewards;
import explicit.rewards.MCRewards;
import explicit.rewards.MDPRewards;
import explicit.rewards.Rewards;
import explicit.rewards.STPGRewards;
import parser.ast.Coalition;
import parser.ast.Expression;
import parser.ast.ExpressionProb;
import parser.ast.ExpressionReward;
import parser.ast.ExpressionSS;
import parser.ast.ExpressionStrategy;
import parser.ast.ExpressionTemporal;
import parser.ast.ExpressionUnaryOp;
import parser.ast.RewardStruct;
import parser.ast.TemporalOperatorBound;
import parser.type.TypeBool;
import parser.type.TypeDouble;
import parser.type.TypePathBool;
import parser.type.TypePathDouble;
import prism.IntegerBound;
import prism.ModelType;
import prism.OpRelOpBound;
import prism.Prism;
import prism.PrismComponent;
import prism.PrismException;
import prism.PrismLog;
import prism.PrismNotSupportedException;
import prism.PrismSettings;
import prism.PrismUtils;
/**
* Super class for explicit-state probabilistic model checkers.
*/
public class ProbModelChecker extends NonProbModelChecker
{
// Flags/settings
// (NB: defaults do not necessarily coincide with PRISM)
// Method used to solve linear equation systems
protected LinEqMethod linEqMethod = LinEqMethod.GAUSS_SEIDEL;
// Method used to solve MDPs
protected MDPSolnMethod mdpSolnMethod = MDPSolnMethod.GAUSS_SEIDEL;
// Iterative numerical method termination criteria
protected TermCrit termCrit = TermCrit.RELATIVE;
// Parameter for iterative numerical method termination criteria
protected double termCritParam = 1e-8;
// Max iterations for numerical solution
protected int maxIters = 100000;
// Use precomputation algorithms in model checking?
protected boolean precomp = true;
protected boolean prob0 = true;
protected boolean prob1 = true;
// should we suppress log output during precomputations?
protected boolean silentPrecomputations = false;
// Use predecessor relation? (e.g. for precomputation)
protected boolean preRel = true;
// Direction of convergence for value iteration (lfp/gfp)
protected ValIterDir valIterDir = ValIterDir.BELOW;
// Method used for numerical solution
protected SolnMethod solnMethod = SolnMethod.VALUE_ITERATION;
// Is non-convergence of an iterative method an error?
protected boolean errorOnNonConverge = true;
// Adversary export
protected boolean exportAdv = false;
protected String exportAdvFilename;
// Delay between occasional updates for slow processes, e.g. numerical solution (milliseconds)
public static final int UPDATE_DELAY = 5000;
// Enums for flags/settings
// Method used for numerical solution
public enum LinEqMethod {
POWER, JACOBI, GAUSS_SEIDEL, BACKWARDS_GAUSS_SEIDEL, JOR, SOR, BACKWARDS_SOR;
public String fullName()
{
switch (this) {
case POWER:
return "Power method";
case JACOBI:
return "Jacobi";
case GAUSS_SEIDEL:
return "Gauss-Seidel";
case BACKWARDS_GAUSS_SEIDEL:
return "Backwards Gauss-Seidel";
case JOR:
return "JOR";
case SOR:
return "SOR";
case BACKWARDS_SOR:
return "Backwards SOR";
default:
return this.toString();
}
}
};
// Method used for solving MDPs
public enum MDPSolnMethod {
VALUE_ITERATION, GAUSS_SEIDEL, POLICY_ITERATION, MODIFIED_POLICY_ITERATION, LINEAR_PROGRAMMING;
public String fullName()
{
switch (this) {
case VALUE_ITERATION:
return "Value iteration";
case GAUSS_SEIDEL:
return "Gauss-Seidel";
case POLICY_ITERATION:
return "Policy iteration";
case MODIFIED_POLICY_ITERATION:
return "Modified policy iteration";
case LINEAR_PROGRAMMING:
return "Linear programming";
default:
return this.toString();
}
}
};
// Iterative numerical method termination criteria
public enum TermCrit {
ABSOLUTE, RELATIVE
};
// Direction of convergence for value iteration (lfp/gfp)
public enum ValIterDir {
BELOW, ABOVE
};
// Method used for numerical solution
public enum SolnMethod {
VALUE_ITERATION, GAUSS_SEIDEL, POLICY_ITERATION, MODIFIED_POLICY_ITERATION, LINEAR_PROGRAMMING
};
/**
* Create a new ProbModelChecker, inherit basic state from parent (unless null).
*/
public ProbModelChecker(PrismComponent parent) throws PrismException
{
super(parent);
// If present, initialise settings from PrismSettings
if (settings != null) {
String s;
// PRISM_LIN_EQ_METHOD
s = settings.getString(PrismSettings.PRISM_LIN_EQ_METHOD);
if (s.equals("Power")) {
setLinEqMethod(LinEqMethod.POWER);
} else if (s.equals("Jacobi")) {
setLinEqMethod(LinEqMethod.JACOBI);
} else if (s.equals("Gauss-Seidel")) {
setLinEqMethod(LinEqMethod.GAUSS_SEIDEL);
} else if (s.equals("Backwards Gauss-Seidel")) {
setLinEqMethod(LinEqMethod.BACKWARDS_GAUSS_SEIDEL);
} else if (s.equals("JOR")) {
setLinEqMethod(LinEqMethod.JOR);
} else if (s.equals("SOR")) {
setLinEqMethod(LinEqMethod.SOR);
} else if (s.equals("Backwards SOR")) {
setLinEqMethod(LinEqMethod.BACKWARDS_SOR);
} else {
throw new PrismNotSupportedException("Explicit engine does not support linear equation solution method \"" + s + "\"");
}
// PRISM_MDP_SOLN_METHOD
s = settings.getString(PrismSettings.PRISM_MDP_SOLN_METHOD);
if (s.equals("Value iteration")) {
setMDPSolnMethod(MDPSolnMethod.VALUE_ITERATION);
} else if (s.equals("Gauss-Seidel")) {
setMDPSolnMethod(MDPSolnMethod.GAUSS_SEIDEL);
} else if (s.equals("Policy iteration")) {
setMDPSolnMethod(MDPSolnMethod.POLICY_ITERATION);
} else if (s.equals("Modified policy iteration")) {
setMDPSolnMethod(MDPSolnMethod.MODIFIED_POLICY_ITERATION);
} else if (s.equals("Linear programming")) {
setMDPSolnMethod(MDPSolnMethod.LINEAR_PROGRAMMING);
} else {
throw new PrismNotSupportedException("Explicit engine does not support MDP solution method \"" + s + "\"");
}
// PRISM_TERM_CRIT
s = settings.getString(PrismSettings.PRISM_TERM_CRIT);
if (s.equals("Absolute")) {
setTermCrit(TermCrit.ABSOLUTE);
} else if (s.equals("Relative")) {
setTermCrit(TermCrit.RELATIVE);
} else {
throw new PrismNotSupportedException("Unknown termination criterion \"" + s + "\"");
}
// PRISM_TERM_CRIT_PARAM
setTermCritParam(settings.getDouble(PrismSettings.PRISM_TERM_CRIT_PARAM));
// PRISM_MAX_ITERS
setMaxIters(settings.getInteger(PrismSettings.PRISM_MAX_ITERS));
// PRISM_PRECOMPUTATION
setPrecomp(settings.getBoolean(PrismSettings.PRISM_PRECOMPUTATION));
// PRISM_PROB0
setProb0(settings.getBoolean(PrismSettings.PRISM_PROB0));
// PRISM_PROB1
setProb1(settings.getBoolean(PrismSettings.PRISM_PROB1));
// PRISM_USE_PRE
setPreRel(settings.getBoolean(PrismSettings.PRISM_PRE_REL));
// PRISM_EXPORT_ADV
s = settings.getString(PrismSettings.PRISM_EXPORT_ADV);
if (!(s.equals("None")))
setExportAdv(true);
// PRISM_EXPORT_ADV_FILENAME
setExportAdvFilename(settings.getString(PrismSettings.PRISM_EXPORT_ADV_FILENAME));
}
}
// Settings methods
/**
* Inherit settings (and the log) from another ProbModelChecker object.
* For model checker objects that inherit a PrismSettings object, this is superfluous
* since this has been done already.
*/
public void inheritSettings(ProbModelChecker other)
{
super.inheritSettings(other);
setLinEqMethod(other.getLinEqMethod());
setMDPSolnMethod(other.getMDPSolnMethod());
setTermCrit(other.getTermCrit());
setTermCritParam(other.getTermCritParam());
setMaxIters(other.getMaxIters());
setPrecomp(other.getPrecomp());
setProb0(other.getProb0());
setProb1(other.getProb1());
setValIterDir(other.getValIterDir());
setSolnMethod(other.getSolnMethod());
setErrorOnNonConverge(other.geterrorOnNonConverge());
}
/**
* Print summary of current settings.
*/
public void printSettings()
{
super.printSettings();
mainLog.print("linEqMethod = " + linEqMethod + " ");
mainLog.print("mdpSolnMethod = " + mdpSolnMethod + " ");
mainLog.print("termCrit = " + termCrit + " ");
mainLog.print("termCritParam = " + termCritParam + " ");
mainLog.print("maxIters = " + maxIters + " ");
mainLog.print("precomp = " + precomp + " ");
mainLog.print("prob0 = " + prob0 + " ");
mainLog.print("prob1 = " + prob1 + " ");
mainLog.print("valIterDir = " + valIterDir + " ");
mainLog.print("solnMethod = " + solnMethod + " ");
mainLog.print("errorOnNonConverge = " + errorOnNonConverge + " ");
}
// Set methods for flags/settings
/**
* Set verbosity level, i.e. amount of output produced.
*/
public void setVerbosity(int verbosity)
{
this.verbosity = verbosity;
}
/**
* Set flag for suppressing log output during precomputations (prob0, prob1, ...)
* @param value silent?
* @return the previous value of this flag
*/
public boolean setSilentPrecomputations(boolean value)
{
boolean old = silentPrecomputations;
silentPrecomputations = value;
return old;
}
/**
* Set method used to solve linear equation systems.
*/
public void setLinEqMethod(LinEqMethod linEqMethod)
{
this.linEqMethod = linEqMethod;
}
/**
* Set method used to solve MDPs.
*/
public void setMDPSolnMethod(MDPSolnMethod mdpSolnMethod)
{
this.mdpSolnMethod = mdpSolnMethod;
}
/**
* Set termination criteria type for numerical iterative methods.
*/
public void setTermCrit(TermCrit termCrit)
{
this.termCrit = termCrit;
}
/**
* Set termination criteria parameter (epsilon) for numerical iterative methods.
*/
public void setTermCritParam(double termCritParam)
{
this.termCritParam = termCritParam;
}
/**
* Set maximum number of iterations for numerical iterative methods.
*/
public void setMaxIters(int maxIters)
{
this.maxIters = maxIters;
}
/**
* Set whether or not to use precomputation (Prob0, Prob1, etc.).
*/
public void setPrecomp(boolean precomp)
{
this.precomp = precomp;
}
/**
* Set whether or not to use Prob0 precomputation
*/
public void setProb0(boolean prob0)
{
this.prob0 = prob0;
}
/**
* Set whether or not to use Prob1 precomputation
*/
public void setProb1(boolean prob1)
{
this.prob1 = prob1;
}
/**
* Set whether or not to use pre-computed predecessor relation
*/
public void setPreRel(boolean preRel)
{
this.preRel = preRel;
}
/**
* Set direction of convergence for value iteration (lfp/gfp).
*/
public void setValIterDir(ValIterDir valIterDir)
{
this.valIterDir = valIterDir;
}
/**
* Set method used for numerical solution.
*/
public void setSolnMethod(SolnMethod solnMethod)
{
this.solnMethod = solnMethod;
}
/**
* Set whether non-convergence of an iterative method an error
*/
public void setErrorOnNonConverge(boolean errorOnNonConverge)
{
this.errorOnNonConverge = errorOnNonConverge;
}
public void setExportAdv(boolean exportAdv)
{
this.exportAdv = exportAdv;
}
public void setExportAdvFilename(String exportAdvFilename)
{
this.exportAdvFilename = exportAdvFilename;
}
// Get methods for flags/settings
public int getVerbosity()
{
return verbosity;
}
public LinEqMethod getLinEqMethod()
{
return linEqMethod;
}
public MDPSolnMethod getMDPSolnMethod()
{
return mdpSolnMethod;
}
public TermCrit getTermCrit()
{
return termCrit;
}
public double getTermCritParam()
{
return termCritParam;
}
public int getMaxIters()
{
return maxIters;
}
public boolean getPrecomp()
{
return precomp;
}
public boolean getProb0()
{
return prob0;
}
public boolean getProb1()
{
return prob1;
}
public boolean getPreRel()
{
return preRel;
}
public ValIterDir getValIterDir()
{
return valIterDir;
}
public SolnMethod getSolnMethod()
{
return solnMethod;
}
/**
* Is non-convergence of an iterative method an error?
*/
public boolean geterrorOnNonConverge()
{
return errorOnNonConverge;
}
// Model checking functions
@Override
public StateValues checkExpression(Model model, Expression expr, BitSet statesOfInterest) throws PrismException
{
StateValues res;
// <<>> or [[]] operator
if (expr instanceof ExpressionStrategy) {
res = checkExpressionStrategy(model, (ExpressionStrategy) expr, statesOfInterest);
}
// P operator
else if (expr instanceof ExpressionProb) {
res = checkExpressionProb(model, (ExpressionProb) expr, statesOfInterest);
}
// R operator
else if (expr instanceof ExpressionReward) {
res = checkExpressionReward(model, (ExpressionReward) expr, statesOfInterest);
}
// S operator
else if (expr instanceof ExpressionSS) {
res = checkExpressionSteadyState(model, (ExpressionSS) expr);
}
// Otherwise, use the superclass
else {
res = super.checkExpression(model, expr, statesOfInterest);
}
return res;
}
/**
* Model check a <<>> or [[]] operator expression and return the values for the statesOfInterest.
* * @param statesOfInterest the states of interest, see checkExpression()
*/
protected StateValues checkExpressionStrategy(Model model, ExpressionStrategy expr, BitSet statesOfInterest) throws PrismException
{
// Only support <<>>/[[]] for MDPs right now
if (!(this instanceof MDPModelChecker))
throw new PrismNotSupportedException("The " + expr.getOperatorString() + " operator is only supported for MDPs currently");
// Will we be quantifying universally or existentially over strategies/adversaries?
boolean forAll = !expr.isThereExists();
// Extract coalition info
Coalition coalition = expr.getCoalition();
// For non-games (i.e., models with a single player), deal with the coalition operator here and then remove it
if (coalition != null && !model.getModelType().multiplePlayers()) {
if (coalition.isEmpty()) {
// An empty coalition negates the quantification ("*" has no effect)
forAll = !forAll;
}
coalition = null;
}
// For now, just support a single expression (which may encode a Boolean combination of objectives)
List<Expression> exprs = expr.getOperands();
if (exprs.size() > 1) {
throw new PrismException("Cannot currently check strategy operators wth lists of expressions");
}
Expression exprSub = exprs.get(0);
// Pass onto relevant method:
// P operator
if (exprSub instanceof ExpressionProb) {
return checkExpressionProb(model, (ExpressionProb) exprSub, forAll, coalition, statesOfInterest);
}
// R operator
else if (exprSub instanceof ExpressionReward) {
return checkExpressionReward(model, (ExpressionReward) exprSub, forAll, coalition, statesOfInterest);
}
// Anything else is an error
else {
throw new PrismException("Unexpected operators in " + expr.getOperatorString() + " operator");
}
}
/**
* Model check a P operator expression and return the values for the statesOfInterest.
* @param statesOfInterest the states of interest, see checkExpression()
*/
protected StateValues checkExpressionProb(Model model, ExpressionProb expr, BitSet statesOfInterest) throws PrismException
{
// Use the default semantics for a standalone P operator
// (i.e. quantification over all strategies, and no game-coalition info)
return checkExpressionProb(model, expr, true, null, statesOfInterest);
}
/**
* Model check a P operator expression and return the values for the states of interest.
* @param model The model
* @param expr The P operator expression
* @param forAll Are we checking "for all strategies" (true) or "there exists a strategy" (false)? [irrelevant for numerical (=?) queries]
* @param coalition If relevant, info about which set of players this P operator refers to (null if irrelevant)
* @param statesOfInterest the states of interest, see checkExpression()
*/
protected StateValues checkExpressionProb(Model model, ExpressionProb expr, boolean forAll, Coalition coalition, BitSet statesOfInterest) throws PrismException
{
// Get info from P operator
OpRelOpBound opInfo = expr.getRelopBoundInfo(constantValues);
MinMax minMax = opInfo.getMinMax(model.getModelType(), forAll);
// Compute probabilities
StateValues probs = checkProbPathFormula(model, expr.getExpression(), minMax, statesOfInterest);
// Print out probabilities
if (getVerbosity() > 5) {
mainLog.print("\nProbabilities (non-zero only) for all states:\n");
probs.print(mainLog);
}
// For =? properties, just return values
if (opInfo.isNumeric()) {
return probs;
}
// Otherwise, compare against bound to get set of satisfying states
else {
BitSet sol = probs.getBitSetFromInterval(opInfo.getRelOp(), opInfo.getBound());
probs.clear();
return StateValues.createFromBitSet(sol, model);
}
}
/**
* Compute probabilities for the contents of a P operator.
* @param statesOfInterest the states of interest, see checkExpression()
*/
protected StateValues checkProbPathFormula(Model model, Expression expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
// Test whether this is a simple path formula (i.e. PCTL)
// and whether we want to use the corresponding algorithms
boolean useSimplePathAlgo = expr.isSimplePathFormula();
if (useSimplePathAlgo &&
settings.getBoolean(PrismSettings.PRISM_PATH_VIA_AUTOMATA) &&
LTLModelChecker.isSupportedLTLFormula(model.getModelType(), expr)) {
// If PRISM_PATH_VIA_AUTOMATA is true, we want to use the LTL engine
// whenever possible
useSimplePathAlgo = false;
}
if (useSimplePathAlgo) {
return checkProbPathFormulaSimple(model, expr, minMax, statesOfInterest);
} else {
return checkProbPathFormulaLTL(model, expr, false, minMax, statesOfInterest);
}
}
/**
* Compute probabilities for a simple, non-LTL path operator.
*/
protected StateValues checkProbPathFormulaSimple(Model model, Expression expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
boolean negated = false;
StateValues probs = null;
if (Expression.containsTemporalRewardBounds(expr)) {
throw new PrismNotSupportedException("Can not handle reward bounds.");
}
expr = Expression.convertSimplePathFormulaToCanonicalForm(expr);
// Negation
if (expr instanceof ExpressionUnaryOp &&
((ExpressionUnaryOp)expr).getOperator() == ExpressionUnaryOp.NOT) {
negated = true;
minMax = minMax.negate();
expr = ((ExpressionUnaryOp)expr).getOperand();
}
if (expr instanceof ExpressionTemporal) {
ExpressionTemporal exprTemp = (ExpressionTemporal) expr;
// Next
if (exprTemp.getOperator() == ExpressionTemporal.P_X) {
probs = checkProbNext(model, exprTemp, minMax, statesOfInterest);
}
// Until
else if (exprTemp.getOperator() == ExpressionTemporal.P_U) {
if (exprTemp.hasBounds()) {
probs = checkProbBoundedUntil(model, exprTemp, minMax, statesOfInterest);
} else {
probs = checkProbUntil(model, exprTemp, minMax, statesOfInterest);
}
}
}
if (probs == null)
throw new PrismException("Unrecognised path operator in P operator");
if (negated) {
// Subtract from 1 for negation
probs.timesConstant(-1.0);
probs.plusConstant(1.0);
}
return probs;
}
/**
* Compute probabilities for a next operator.
*/
protected StateValues checkProbNext(Model model, ExpressionTemporal expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
// Model check the operand for all states
BitSet target = checkExpression(model, expr.getOperand2(), null).getBitSet();
// Compute/return the probabilities
ModelCheckerResult res = null;
switch (model.getModelType()) {
case CTMC:
res = ((CTMCModelChecker) this).computeNextProbs((CTMC) model, target);
break;
case DTMC:
res = ((DTMCModelChecker) this).computeNextProbs((DTMC) model, target);
break;
case MDP:
res = ((MDPModelChecker) this).computeNextProbs((MDP) model, target, minMax.isMin());
break;
case STPG:
res = ((STPGModelChecker) this).computeNextProbs((STPG) model, target, minMax.isMin1(), minMax.isMin2());
break;
default:
throw new PrismNotSupportedException("Cannot model check " + expr + " for " + model.getModelType() + "s");
}
result.setStrategy(res.strat);
return StateValues.createFromDoubleArray(res.soln, model);
}
/**
* Compute probabilities for a bounded until operator.
*/
protected StateValues checkProbBoundedUntil(Model model, ExpressionTemporal expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
// This method just handles discrete time
// Continuous-time model checkers will override this method
// Get info from bounded until
Integer lowerBound;
IntegerBound bounds;
int i;
// get and check bounds information
bounds = IntegerBound.fromExpressionTemporal(expr, constantValues, true);
// Model check operands for all states
BitSet remain = checkExpression(model, expr.getOperand1(), null).getBitSet();
BitSet target = checkExpression(model, expr.getOperand2(), null).getBitSet();
if (bounds.hasLowerBound()) {
lowerBound = bounds.getLowestInteger();
} else {
lowerBound = 0;
}
Integer windowSize = null; // unbounded
if (bounds.hasUpperBound()) {
windowSize = bounds.getHighestInteger() - lowerBound;
}
// compute probabilities for Until<=windowSize
StateValues sv = null;
if (windowSize == null) {
// unbounded
ModelCheckerResult res = null;
switch (model.getModelType()) {
case DTMC:
res = ((DTMCModelChecker) this).computeUntilProbs((DTMC) model, remain, target);
break;
case MDP:
res = ((MDPModelChecker) this).computeUntilProbs((MDP) model, remain, target, minMax.isMin());
break;
case STPG:
res = ((STPGModelChecker) this).computeUntilProbs((STPG) model, remain, target, minMax.isMin1(), minMax.isMin2());
break;
default:
throw new PrismException("Cannot model check " + expr + " for " + model.getModelType() + "s");
}
result.setStrategy(res.strat);
sv = StateValues.createFromDoubleArray(res.soln, model);
} else if (windowSize == 0) {
// A trivial case: windowSize=0 (prob is 1 in target states, 0 otherwise)
sv = StateValues.createFromBitSetAsDoubles(target, model);
} else {
// Otherwise: numerical solution
ModelCheckerResult res = null;
switch (model.getModelType()) {
case DTMC:
res = ((DTMCModelChecker) this).computeBoundedUntilProbs((DTMC) model, remain, target, windowSize);
break;
case MDP:
res = ((MDPModelChecker) this).computeBoundedUntilProbs((MDP) model, remain, target, windowSize, minMax.isMin());
break;
case STPG:
res = ((STPGModelChecker) this).computeBoundedUntilProbs((STPG) model, remain, target, windowSize, minMax.isMin1(), minMax.isMin2());
break;
default:
throw new PrismNotSupportedException("Cannot model check " + expr + " for " + model.getModelType() + "s");
}
result.setStrategy(res.strat);
sv = StateValues.createFromDoubleArray(res.soln, model);
}
// perform lowerBound restricted next-step computations to
// deal with lower bound.
if (lowerBound > 0) {
double[] probs = sv.getDoubleArray();
for (i = 0; i < lowerBound; i++) {
switch (model.getModelType()) {
case DTMC:
probs = ((DTMCModelChecker) this).computeRestrictedNext((DTMC) model, remain, probs);
break;
case MDP:
probs = ((MDPModelChecker) this).computeRestrictedNext((MDP) model, remain, probs, minMax.isMin());
break;
case STPG:
// TODO (JK): Figure out if we can handle lower bounds for STPG in the same way
throw new PrismNotSupportedException("Lower bounds not yet supported for STPGModelChecker");
default:
throw new PrismNotSupportedException("Cannot model check " + expr + " for " + model.getModelType() + "s");
}
}
sv = StateValues.createFromDoubleArray(probs, model);
}
return sv;
}
/**
* Compute probabilities for an (unbounded) until operator.
*/
protected StateValues checkProbUntil(Model model, ExpressionTemporal expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
// Model check operands for all states
BitSet remain = checkExpression(model, expr.getOperand1(), null).getBitSet();
BitSet target = checkExpression(model, expr.getOperand2(), null).getBitSet();
// Compute/return the probabilities
ModelCheckerResult res = null;
switch (model.getModelType()) {
case CTMC:
res = ((CTMCModelChecker) this).computeUntilProbs((CTMC) model, remain, target);
break;
case DTMC:
res = ((DTMCModelChecker) this).computeUntilProbs((DTMC) model, remain, target);
break;
case MDP:
res = ((MDPModelChecker) this).computeUntilProbs((MDP) model, remain, target, minMax.isMin());
break;
case STPG:
res = ((STPGModelChecker) this).computeUntilProbs((STPG) model, remain, target, minMax.isMin1(), minMax.isMin2());
break;
default:
throw new PrismNotSupportedException("Cannot model check " + expr + " for " + model.getModelType() + "s");
}
result.setStrategy(res.strat);
return StateValues.createFromDoubleArray(res.soln, model);
}
/**
* Compute probabilities for an LTL path formula
*/
protected StateValues checkProbPathFormulaLTL(Model model, Expression expr, boolean qual, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
// To be overridden by subclasses
throw new PrismNotSupportedException("Computation not implemented yet");
}
/**
* Model check an R operator expression and return the values for all states.
*/
protected StateValues checkExpressionReward(Model model, ExpressionReward expr, BitSet statesOfInterest) throws PrismException
{
// Use the default semantics for a standalone R operator
// (i.e. quantification over all strategies, and no game-coalition info)
return checkExpressionReward(model, expr, true, null, statesOfInterest);
}
/**
* Model check an R operator expression and return the values for all states.
*/
protected StateValues checkExpressionReward(Model model, ExpressionReward expr, boolean forAll, Coalition coalition, BitSet statesOfInterest) throws PrismException
{
// Get info from R operator
OpRelOpBound opInfo = expr.getRelopBoundInfo(constantValues);
MinMax minMax = opInfo.getMinMax(model.getModelType(), forAll);
// Build rewards
int r = expr.getRewardStructIndexByIndexObject(modelInfo, constantValues);
mainLog.println("Building reward structure...");
// we allow negative (and positive) rewards for DTMCs
boolean allowNegative = model.getModelType() == ModelType.DTMC;
Rewards rewards = constructRewards(model, r, allowNegative);
// Compute rewards
StateValues rews = checkRewardFormula(model, rewards, expr.getExpression(), minMax, statesOfInterest);
// Print out rewards
if (getVerbosity() > 5) {
mainLog.print("\nRewards (non-zero only) for all states:\n");
rews.print(mainLog);
}
// For =? properties, just return values
if (opInfo.isNumeric()) {
return rews;
}
// Otherwise, compare against bound to get set of satisfying states
else {
BitSet sol = rews.getBitSetFromInterval(opInfo.getRelOp(), opInfo.getBound());
rews.clear();
return StateValues.createFromBitSet(sol, model);
}
}
/**
* Construct rewards for the reward structure with index r of the model generator and a model.
* Ensures non-negative rewards.
* <br>
* Note: Relies on the stored ModelGenerator for constructing the reward structure.
*/
protected Rewards constructRewards(Model model, int r) throws PrismException
{
return constructRewards(model, r, false);
}
/**
* Construct rewards for the reward structure with index r of the model generator and a model.
* <br>
* If {@code allowNegativeRewards} is true, the rewards may be positive and negative, i.e., weights.
* <br>
* Note: Relies on the stored ModelGenerator for constructing the reward structure.
*/
protected Rewards constructRewards(Model model, int r, boolean allowNegativeRewards) throws PrismException
{
ConstructRewards constructRewards = new ConstructRewards(mainLog);
if (allowNegativeRewards)
constructRewards.allowNegativeRewards();
return constructRewards.buildRewardStructure(model, modelGen, r);
}
/**
* Construct rewards from a (non-negative) reward structure and a model.
* <br>
* Note: Deprecated, use the methods with reward structure index r instead
* to allow construction from model generators.
*/
@Deprecated
protected Rewards constructRewards(Model model, RewardStruct rewStruct) throws PrismException
{
return constructRewards(model, rewStruct, false);
}
/**
* Construct rewards from a reward structure and a model.
* <br>
* If {@code allowNegativeRewards} is true, the rewards may be positive and negative, i.e., weights.
* <br>
* Note: Deprecated, use the methods with reward structure index r instead
* to allow construction from model generators.
*/
@Deprecated
protected Rewards constructRewards(Model model, RewardStruct rewStruct, boolean allowNegativeRewards) throws PrismException
{
Rewards rewards;
ConstructRewards constructRewards = new ConstructRewards(mainLog);
if (allowNegativeRewards)
constructRewards.allowNegativeRewards();
switch (model.getModelType()) {
case CTMC:
case DTMC:
rewards = constructRewards.buildMCRewardStructure((DTMC) model, rewStruct, constantValues);
break;
case MDP:
rewards = constructRewards.buildMDPRewardStructure((MDP) model, rewStruct, constantValues);
break;
default:
throw new PrismNotSupportedException("Cannot build rewards for " + model.getModelType() + "s");
}
return rewards;
}
/**
* Compute rewards for the contents of an R operator.
*/
protected StateValues checkRewardFormula(Model model, Rewards modelRewards, Expression expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
StateValues rewards = null;
if (expr.getType() instanceof TypePathDouble) {
ExpressionTemporal exprTemp = (ExpressionTemporal) expr;
switch (exprTemp.getOperator()) {
case ExpressionTemporal.R_I:
rewards = checkRewardInstantaneous(model, modelRewards, exprTemp, minMax, statesOfInterest);
break;
case ExpressionTemporal.R_C:
if (exprTemp.hasBounds()) {
rewards = checkRewardCumulative(model, modelRewards, exprTemp, minMax);
} else {
rewards = checkRewardTotal(model, modelRewards, exprTemp, minMax);
}
break;
default:
throw new PrismNotSupportedException("Explicit engine does not yet handle the " + exprTemp.getOperatorSymbol() + " reward operator");
}
} else if (expr.getType() instanceof TypePathBool || expr.getType() instanceof TypeBool) {
rewards = checkRewardPathFormula(model, modelRewards, expr, minMax, statesOfInterest);
}
if (rewards == null)
throw new PrismException("Unrecognised operator in R operator");
return rewards;
}
/**
* Compute rewards for an instantaneous reward operator.
*/
protected StateValues checkRewardInstantaneous(Model model, Rewards modelRewards, ExpressionTemporal expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
if (expr.getBounds().hasRewardBounds()) {
throw new PrismNotSupportedException("Instantaneous reward operator with reward bounds is not supported.");
}
// Compute/return the rewards
ModelCheckerResult res = null;
switch (model.getModelType()) {
case DTMC: {
int k = expr.getBounds().getStandardBound(model.getModelType()).getUpperBound().evaluateInt(constantValues);
res = ((DTMCModelChecker) this).computeInstantaneousRewards((DTMC) model, (MCRewards) modelRewards, k, statesOfInterest);
break;
}
case CTMC: {
double t = expr.getBounds().getStandardBound(model.getModelType()).getUpperBound().evaluateDouble(constantValues);
res = ((CTMCModelChecker) this).computeInstantaneousRewards((CTMC) model, (MCRewards) modelRewards, t);
break;
}
case MDP: {
int k = expr.getBounds().getStandardBound(model.getModelType()).getUpperBound().evaluateInt(constantValues);
res = ((MDPModelChecker) this).computeInstantaneousRewards((MDP) model, (MDPRewards) modelRewards, k, minMax.isMin());
break;
}
default:
throw new PrismNotSupportedException("Explicit engine does not yet handle the " + expr.getOperatorSymbol() + " reward operator for " + model.getModelType()
+ "s");
}
result.setStrategy(res.strat);
return StateValues.createFromDoubleArray(res.soln, model);
}
/**
* Compute rewards for a cumulative reward operator.
*/
protected StateValues checkRewardCumulative(Model model, Rewards modelRewards, ExpressionTemporal expr, MinMax minMax) throws PrismException
{
int timeInt = -1;
double timeDouble = -1;
if (expr.getBounds().hasRewardBounds()) {
throw new PrismNotSupportedException("Cumulative reward operator with reward bounds is not supported.");
}
TemporalOperatorBound bound = expr.getBounds().getStandardBound(model.getModelType());
if (bound == null || bound.getUpperBound() == null) {
throw new PrismException("This is not a cumulative reward operator");
}
// Get time bound
if (model.getModelType().continuousTime()) {
timeDouble = bound.getUpperBound().evaluateDouble(constantValues);
if (timeDouble < 0) {
throw new PrismException("Invalid time bound " + timeDouble + " in cumulative reward formula");
}
} else {
timeInt = bound.getUpperBound().evaluateInt(constantValues);
if (timeInt < 0) {
throw new PrismException("Invalid time bound " + timeInt + " in cumulative reward formula");
}
}
// Compute/return the rewards
// A trivial case: "C<=0" (prob is 1 in target states, 0 otherwise)
if (timeInt == 0 || timeDouble == 0) {
return new StateValues(TypeDouble.getInstance(), model.getNumStates(), new Double(0));
}
// Otherwise: numerical solution
ModelCheckerResult res = null;
switch (model.getModelType()) {
case DTMC:
res = ((DTMCModelChecker) this).computeCumulativeRewards((DTMC) model, (MCRewards) modelRewards, timeInt);
break;
case CTMC:
res = ((CTMCModelChecker) this).computeCumulativeRewards((CTMC) model, (MCRewards) modelRewards, timeDouble);
break;
case MDP:
res = ((MDPModelChecker) this).computeCumulativeRewards((MDP) model, (MDPRewards) modelRewards, timeInt, minMax.isMin());
break;
default:
throw new PrismNotSupportedException("Explicit engine does not yet handle the " + expr.getOperatorSymbol() + " reward operator for " + model.getModelType()
+ "s");
}
result.setStrategy(res.strat);
return StateValues.createFromDoubleArray(res.soln, model);
}
/**
* Compute expected rewards for a total reward operator.
*/
protected StateValues checkRewardTotal(Model model, Rewards modelRewards, ExpressionTemporal expr, MinMax minMax) throws PrismException
{
if (expr.getBounds().hasRewardBounds()) {
throw new PrismNotSupportedException("Cumulative reward operator with reward bounds is not supported.");
}
if (expr.hasBounds()) {
throw new PrismException("This is not a total reward operator");
}
// Compute/return the rewards
ModelCheckerResult res = null;
switch (model.getModelType()) {
case DTMC:
res = ((DTMCModelChecker) this).computeTotalRewards((DTMC) model, (MCRewards) modelRewards);
break;
case CTMC:
res = ((CTMCModelChecker) this).computeTotalRewards((CTMC) model, (MCRewards) modelRewards);
break;
case MDP:
res = ((MDPModelChecker) this).computeTotalRewards((MDP) model, (MDPRewards) modelRewards, minMax.isMin());
break;
default:
throw new PrismNotSupportedException("Explicit engine does not yet handle the " + expr.getOperatorSymbol() + " reward operator for " + model.getModelType()
+ "s");
}
result.setStrategy(res.strat);
return StateValues.createFromDoubleArray(res.soln, model);
}
/**
* Compute rewards for a path formula in a reward operator.
*/
protected StateValues checkRewardPathFormula(Model model, Rewards modelRewards, Expression expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
if (Expression.isReachWithStateFormula(expr)) {
return checkRewardReach(model, modelRewards, (ExpressionTemporal) expr, minMax, statesOfInterest);
}
else if (Expression.isCoSafeLTLSyntactic(expr, true)) {
return checkRewardCoSafeLTL(model, modelRewards, expr, minMax, statesOfInterest);
}
throw new PrismException("R operator contains a path formula that is not syntactically co-safe: " + expr);
}
/**
* Compute rewards for a reachability reward operator.
*/
protected StateValues checkRewardReach(Model model, Rewards modelRewards, ExpressionTemporal expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
// No time bounds allowed
if (expr.hasBounds()) {
throw new PrismNotSupportedException("R operator cannot contain a bounded F operator: " + expr);
}
// Model check the operand for all states
BitSet target = checkExpression(model, expr.getOperand2(), null).getBitSet();
// Compute/return the rewards
ModelCheckerResult res = null;
switch (model.getModelType()) {
case DTMC:
res = ((DTMCModelChecker) this).computeReachRewards((DTMC) model, (MCRewards) modelRewards, target);
break;
case CTMC:
res = ((CTMCModelChecker) this).computeReachRewards((CTMC) model, (MCRewards) modelRewards, target);
break;
case MDP:
res = ((MDPModelChecker) this).computeReachRewards((MDP) model, (MDPRewards) modelRewards, target, minMax.isMin());
break;
case STPG:
res = ((STPGModelChecker) this).computeReachRewards((STPG) model, (STPGRewards) modelRewards, target, minMax.isMin1(), minMax.isMin2());
break;
default:
throw new PrismNotSupportedException("Explicit engine does not yet handle the " + expr.getOperatorSymbol() + " reward operator for " + model.getModelType()
+ "s");
}
result.setStrategy(res.strat);
return StateValues.createFromDoubleArray(res.soln, model);
}
/**
* Compute rewards for a co-safe LTL reward operator.
*/
protected StateValues checkRewardCoSafeLTL(Model model, Rewards modelRewards, Expression expr, MinMax minMax, BitSet statesOfInterest) throws PrismException
{
// To be overridden by subclasses
throw new PrismException("Computation not implemented yet");
}
/**
* Model check an S operator expression and return the values for all states.
*/
protected StateValues checkExpressionSteadyState(Model model, ExpressionSS expr) throws PrismException
{
// Get info from S operator
OpRelOpBound opInfo = expr.getRelopBoundInfo(constantValues);
MinMax minMax = opInfo.getMinMax(model.getModelType());
// Compute probabilities
StateValues probs = checkSteadyStateFormula(model, expr.getExpression(), minMax);
// Print out probabilities
if (getVerbosity() > 5) {
mainLog.print("\nProbabilities (non-zero only) for all states:\n");
probs.print(mainLog);
}
// For =? properties, just return values
if (opInfo.isNumeric()) {
return probs;
}
// Otherwise, compare against bound to get set of satisfying states
else {
BitSet sol = probs.getBitSetFromInterval(opInfo.getRelOp(), opInfo.getBound());
probs.clear();
return StateValues.createFromBitSet(sol, model);
}
}
/**
* Compute steady-state probabilities for an S operator.
*/
protected StateValues checkSteadyStateFormula(Model model, Expression expr, MinMax minMax) throws PrismException
{
// Model check operand for all states
BitSet b = checkExpression(model, expr, null).getBitSet();
// Compute/return the probabilities
ModelCheckerResult res = null;
switch (model.getModelType()) {
case DTMC:
double multProbs[] = Utils.bitsetToDoubleArray(b, model.getNumStates());
res = ((DTMCModelChecker) this).computeSteadyStateBackwardsProbs((DTMC) model, multProbs);
break;
default:
throw new PrismNotSupportedException("Explicit engine does not yet handle the S operator for " + model.getModelType() + "s");
}
return StateValues.createFromDoubleArray(res.soln, model);
}
// Utility methods for probability distributions
/**
* Generate a probability distribution, stored as a StateValues object, from a file.
* If {@code distFile} is null, so is the return value.
*/
public StateValues readDistributionFromFile(File distFile, Model model) throws PrismException
{
StateValues dist = null;
if (distFile != null) {
mainLog.println("\nImporting probability distribution from file \"" + distFile + "\"...");
// Build an empty vector
dist = new StateValues(TypeDouble.getInstance(), model);
// Populate vector from file
dist.readFromFile(distFile);
}
return dist;
}
/**
* Build a probability distribution, stored as a StateValues object,
* from the initial states info of the current model: either probability 1 for
* the (single) initial state or equiprobable over multiple initial states.
*/
public StateValues buildInitialDistribution(Model model) throws PrismException
{
StateValues dist = null;
// Build an empty vector
dist = new StateValues(TypeDouble.getInstance(), model);
// Populate vector (equiprobable over initial states)
double d = 1.0 / model.getNumInitialStates();
for (int in : model.getInitialStates()) {
dist.setDoubleValue(in, d);
}
return dist;
}
/**
* Export (non-zero) state rewards for one reward structure of a model.
* @param model The model
* @param r Index of reward structure to export (0-indexed)
* @param exportType The format in which to export
* @param out Where to export
*/
public void exportStateRewardsToFile(Model model, int r, int exportType, PrismLog out) throws PrismException
{
int numStates = model.getNumStates();
int nonZeroRews = 0;
if (exportType != Prism.EXPORT_PLAIN) {
throw new PrismNotSupportedException("Exporting state rewards in the requested format is currently not supported by the explicit engine");
}
// we allow negative (and positive) rewards for DTMCs
boolean allowNegative = model.getModelType() == ModelType.DTMC;
Rewards modelRewards = constructRewards(model, r, allowNegative);
switch (model.getModelType()) {
case DTMC:
case CTMC:
MCRewards mcRewards = (MCRewards) modelRewards;
for (int s = 0; s < numStates; s++) {
double d = mcRewards.getStateReward(s);
if (d != 0) {
nonZeroRews++;
}
}
out.println(numStates + " " + nonZeroRews);
for (int s = 0; s < numStates; s++) {
double d = mcRewards.getStateReward(s);
if (d != 0) {
out.println(s + " " + PrismUtils.formatDouble(d));
}
}
break;
case MDP:
case STPG:
MDPRewards mdpRewards = (MDPRewards) modelRewards;
for (int s = 0; s < numStates; s++) {
double d = mdpRewards.getStateReward(s);
if (d != 0) {
nonZeroRews++;
}
}
out.println(numStates + " " + nonZeroRews);
for (int s = 0; s < numStates; s++) {
double d = mdpRewards.getStateReward(s);
if (d != 0) {
out.println(s + " " + PrismUtils.formatDouble(d));
}
}
break;
default:
throw new PrismNotSupportedException("Explicit engine does not yet export state rewards for " + model.getModelType() + "s");
}
}
}