probfd Namespace Reference

Description

The top-level namespace of probabilistic Fast Downward.

Namespaces
namespace	algorithms
	This namespace contains implementations of SSP search algorithms.
namespace	bisimulation
	This namespace contains the implementation of deterministic bisimulation quotients for SSPs, based on merge-and-shrink.
namespace	graphviz
	This namespace contains code used for dumping search spaces as dot graphs.
namespace	heuristics
	This namespace contains heuristic implementations.
namespace	iterators
	Namespace dedicated to custom iterators.
namespace	occupation_measures
	Namespace dedicated to occupation measure heuristic base classes.
namespace	open_lists
	This namespace contains implementations of open lists.
namespace	pdbs
	Namespace dedicated to probabilistic pattern databases.
namespace	preprocessing
	This namespace contains preprocessing algorithms for SSPs.
namespace	solvers
	This namespace contains the solver interface base class for various search algorithms.
namespace	storage
	This namespace contains commonly used storage and container types.
namespace	successor_samplers
	This namespace contains implementations of transition successor samplers.
namespace	tasks
	This namespace contains implementations of probabilistic planning tasks.

Classes
struct	approx_eq_to
	Unary function object for approximate equality comparison. More...
struct	approx_less
	Unary function object for approximate inequality comparison. More...
struct	approx_neq_to
	Unary function object for approximate inequality comparison. More...
struct	CompositeMDP
	Composes a state space and a cost function to an MDP. More...
class	CostFunction
	The interface specifying action and state termination costs, aswell as the goal states of a state space. More...
class	Distribution
	A convenience class that represents a finite probability distribution. More...
class	Evaluator
	The interface representing heuristic functions. More...
struct	Interval
	Represents a closed interval over the extended reals as a pair of lower and upper bound. More...
class	ItemProbabilityPair
	An item-probability pair. More...
class	MDP
	Basic interface for MDPs. More...
class	MDPAlgorithm
	Interface for MDP algorithm implementations. More...
class	MultiPolicy
	Represents a policy that can specify a set of actions for a state. More...
struct	no_normalize_t
	Disambiguator tag type. More...
struct	overloaded
class	Policy
	Represents a deterministic, stationary, partial policy. More...
struct	PolicyDecision
	Stores an action and a bounding interval for the Q* value of this action. More...
class	ProbabilisticEffectConditionsProxy
	Proxy class used to inspect the list of effect conditions of a conditional effect of a probabilistic operator. Can be used as a range of FactProxies, one for each effect condition. More...
class	ProbabilisticEffectProxy
	Proxy class used to inspect a probabilistic effect of a probabilistic operator. More...
class	ProbabilisticEffectsProxy
	Proxy class used to inspect the list of probabilistic effects of a probabilistic operator. Can be used as a range of ProbabilisticEffectProxies, one for each probabilistic effect. More...
class	ProbabilisticOperatorProxy
	Proxy class used to inspect a single probabilistic operator. More...
class	ProbabilisticOperatorsProxy
	Proxy class used to inspect a list of probabilistic operators of a probabilistic task. Can be used as a range of ProbabilisticOperatorProxies, one for each probabilistic operator. More...
class	ProbabilisticOutcomeProxy
	Proxy class used to inspect a single probabilistic outcome of a probabilistic operator. More...
class	ProbabilisticOutcomesProxy
	Proxy class used to inspect the list of probabilistic outcomes of a probabilistic operator. Can be used as a range of ProbabilisticOutcomeProxies, one for each probabilistic outcome. More...
class	ProbabilisticTask
	Represents a probabilistic planning task with axioms and conditional effects. More...
class	ProbabilisticTaskProxy
	Proxy class used to inspect a probabilistic planning task. More...
class	ProgressReport
	A registry for print functions related to search progress. More...
class	SimpleMDP
	Basic interface for MDPs. More...
class	SolverInterface
	An interface that describes an MDP solver. More...
struct	StateID
	A StateID represents a state within a StateIDMap. Just like Fast Downward's StateID type, it is trivial to copy and intended for long term storage. More...
class	StateSpace
	An interface representing a Markov Decision Process (MDP) without objective function. More...
class	TerminationInfo
	Specifies the termination cost and goal status of a state. More...

Typedefs
using	FDROpenList = algorithms::OpenList<OperatorID>
	Type alias for OpenLists for MDPs in FDR.
using	FDRPolicyPicker = algorithms::PolicyPicker<State, OperatorID>
	Type alias for policy pickers for MDPs in FDR.
using	FDRTransitionSorter = algorithms::TransitionSorter<State, OperatorID>
	Type alias for TransitionSorters for MDPs in FDR.
using	value_t = double
	Typedef for the state value type.
using	FDRStateSpace = StateSpace<State, OperatorID>
	Type alias for state spaces in FDR.
using	FDRCostFunction = SimpleCostFunction<State, OperatorID>
	Type alias for cost functions for MDPs in FDR.
using	FDRSimpleCostFunction = SimpleCostFunction<State, OperatorID>
	Type alias for simple cost functions for MDPs in FDR.
using	FDRMDP = MDP<State, OperatorID>
	Type alias for MDPs with states in FDR.
using	FDRSimpleMDP = SimpleMDP<State, OperatorID>
	Type alias for simple MDPs with states in FDR.
using	FDREvaluator = Evaluator<State>
	Type alias for evaluators for states in FDR.
using	Printer = std::function<void(std::ostream&)>
	A function that prints something to an output stream.
using	BoundProperty = std::function<Interval()>
	A function that produces a state value bound.
template<typename T >
using	param_type
	Alias template defining the best way to pass a parameter of a given type.

Functions
std::string	usage (const std::string &progname)
	Returns a string documenting usage of the planner. Receives the program name as input.
template<typename State , typename Action >
void	compute_value_table (MDP< State, Action > &mdp, param_type< State > initial_state, const Evaluator< State > &heuristic, std::span< value_t > value_table, double max_time=std::numeric_limits< double >::infinity())
	Computes the optimal value function of the abstraction, complete up to forward reachability from the initial state.
template<typename State , typename Action >
std::unique_ptr< MultiPolicy< State, Action > >	compute_optimal_projection_policy (MDP< State, Action > &mdp, std::span< const value_t > value_table, param_type< State > initial_state, utils::RandomNumberGenerator &rng, bool wildcard)
	Extract an abstract optimal policy from the value table.
template<typename State , typename Action >
std::unique_ptr< MultiPolicy< State, Action > >	compute_greedy_projection_policy (MDP< State, Action > &mdp, std::span< const value_t > value_table, param_type< State > initial_state, utils::RandomNumberGenerator &rng, bool wildcard)
	Extracts an abstract greedy policy from the value table, which may not be optimal if traps are existent.
template<typename State , typename Action >
void	verify (MDP< State, Action > &mdp, std::span< const value_t > value_table, lp::LPSolverType type)
	Computes the optimal value function of the abstraction, complete up to forward reachability from the initial state.
Interval	operator+ (Interval lhs, Interval rhs)
	Computes the component-wise addition of two intervals.
Interval	operator* (value_t scale_factor, Interval val)
	Scales an interval.
Interval	operator* (Interval val, value_t scale_factor)
	Scales an interval.
std::ostream &	operator<< (std::ostream &os, Interval val)
	Stream output operator. Prints '[<val.lower>,<val.upper>]'.
bool	does_fire (const ProbabilisticEffectProxy &effect, const State &state)
	Checks if the conditions of a probabilistic effect are fulfilled in a state.
constexpr value_t	operator""_vt (long double value)
	User-defined floating-point literals for state values.
constexpr value_t	operator""_vt (unsigned long long value)
	User-defined integer literals for state values.
bool	is_approx_equal (value_t v1, value_t v2, value_t epsilon=g_epsilon)
	Equivalent to \(|v_1 - v_2| \leq \epsilon\).
bool	is_approx_less (value_t v1, value_t v2, value_t epsilon=g_epsilon)
	Equivalent to \(v_2 - v_1 > \epsilon\).
bool	is_approx_greater (value_t v1, value_t v2, value_t epsilon=g_epsilon)
	Equivalent to \(v_1 - v_2 > \epsilon\).

Variables
constexpr no_normalize_t	no_normalize = no_normalize_t{}
	Disambiguator tag for Distribution constructor to indicate that the probabilities are already normalized to one.
value_t	g_epsilon
	The default tolerance value for approximate comparisons.

Typedef Documentation

FDROpenList

using probfd::FDROpenList = algorithms::OpenList<OperatorID>

Type alias for OpenLists for MDPs in FDR.

FDRPolicyPicker

using probfd::FDRPolicyPicker = algorithms::PolicyPicker<State, OperatorID>

Type alias for policy pickers for MDPs in FDR.

FDRTransitionSorter

using probfd::FDRTransitionSorter = algorithms::TransitionSorter<State, OperatorID>

Type alias for TransitionSorters for MDPs in FDR.

value_t

using probfd::value_t = double

Typedef for the state value type.

FDRStateSpace

using probfd::FDRStateSpace = StateSpace<State, OperatorID>

Type alias for state spaces in FDR.

FDRCostFunction

using probfd::FDRCostFunction = SimpleCostFunction<State, OperatorID>

Type alias for cost functions for MDPs in FDR.

FDRSimpleCostFunction

using probfd::FDRSimpleCostFunction = SimpleCostFunction<State, OperatorID>

Type alias for simple cost functions for MDPs in FDR.

FDRMDP

using probfd::FDRMDP = MDP<State, OperatorID>

Type alias for MDPs with states in FDR.

FDRSimpleMDP

using probfd::FDRSimpleMDP = SimpleMDP<State, OperatorID>

Type alias for simple MDPs with states in FDR.

FDREvaluator

using probfd::FDREvaluator = Evaluator<State>

Type alias for evaluators for states in FDR.

Printer

using probfd::Printer = std::function<void(std::ostream&)>

A function that prints something to an output stream.

BoundProperty

using probfd::BoundProperty = std::function<Interval()>

A function that produces a state value bound.

param_type

template<typename T >

using probfd::param_type

Initial value:

 
    typename std::conditional_t<is_cheap_to_copy_v<T>, T, const T&>

Alias template defining the best way to pass a parameter of a given type.

Function Documentation

usage()

std::string probfd::usage ( const std::string & progname )

extern

Returns a string documenting usage of the planner. Receives the program name as input.

compute_value_table()

template<typename State , typename Action >

void probfd::compute_value_table	(	MDP< State, Action > &	mdp,
		param_type< State >	initial_state,
		const Evaluator< State > &	heuristic,
		std::span< value_t >	value_table,
		double	max_time = std::numeric_limits<double>::infinity() )

Computes the optimal value function of the abstraction, complete up to forward reachability from the initial state.

compute_optimal_projection_policy()

template<typename State , typename Action >

std::unique_ptr< MultiPolicy< State, Action > > probfd::compute_optimal_projection_policy	(	MDP< State, Action > &	mdp,
		std::span< const value_t >	value_table,
		param_type< State >	initial_state,
		utils::RandomNumberGenerator &	rng,
		bool	wildcard )

Extract an abstract optimal policy from the value table.

Tie-breaking is performed randomly using the input RNG. If the wildcard option is specified, a wildcard policy will be returned, i.e., a policy that assigns multiple equivalent operators to an abstract state.

compute_greedy_projection_policy()

template<typename State , typename Action >

std::unique_ptr< MultiPolicy< State, Action > > probfd::compute_greedy_projection_policy	(	MDP< State, Action > &	mdp,
		std::span< const value_t >	value_table,
		param_type< State >	initial_state,
		utils::RandomNumberGenerator &	rng,
		bool	wildcard )

Extracts an abstract greedy policy from the value table, which may not be optimal if traps are existent.

Tie-breaking is performed randomly using the input RNG. If the wildcard option is specified, a wildcard policy will be returned, i.e., a policy that assigns multiple equivalent operators to an abstract state.

verify()

template<typename State , typename Action >

void probfd::verify	(	MDP< State, Action > &	mdp,
		std::span< const value_t >	value_table,
		lp::LPSolverType	type )

Computes the optimal value function of the abstraction, complete up to forward reachability from the initial state.

operator+()

Interval probfd::operator+	(	Interval	lhs,
		Interval	rhs )

Computes the component-wise addition of two intervals.

Parameters

lhs	- An interval \([a_1, b_1]\)
rhs	- An interval \([a_2, b_2]\)

Returns

The interval \([a_1 + a_2, b_1 + b_2]\)

operator*() [1/2]

Interval probfd::operator*	(	value_t	scale_factor,
		Interval	val )

Scales an interval.

Parameters

scale_factor	- The scaling factor \(n\)
val	- An interval \([a, b]\)

Returns

The interval \([n \cdot a, n \cdot b]\)

operator*() [2/2]

Interval probfd::operator*	(	Interval	val,
		value_t	scale_factor )

Scales an interval.

Parameters

scale_factor	- The scaling factor \(n\)
val	- An interval \([a, b]\)

Returns

The interval \([n \cdot a, n \cdot b]\)

operator<<()

std::ostream & probfd::operator<<	(	std::ostream &	os,
		Interval	val )

Stream output operator. Prints '[<val.lower>,<val.upper>]'.

does_fire()

bool probfd::does_fire	(	const ProbabilisticEffectProxy &	effect,
		const State &	state )

Checks if the conditions of a probabilistic effect are fulfilled in a state.

operator""_vt() [1/2]

value_t probfd::operator""_vt ( long double value )

constexpr

User-defined floating-point literals for state values.

operator""_vt() [2/2]

value_t probfd::operator""_vt ( unsigned long long value )

constexpr

User-defined integer literals for state values.

is_approx_equal()

bool probfd::is_approx_equal	(	value_t	v1,
		value_t	v2,
		value_t	epsilon = g_epsilon )

Equivalent to \(|v_1 - v_2| \leq \epsilon\).

is_approx_less()

bool probfd::is_approx_less	(	value_t	v1,
		value_t	v2,
		value_t	epsilon = g_epsilon )

Equivalent to \(v_2 - v_1 > \epsilon\).

is_approx_greater()

bool probfd::is_approx_greater	(	value_t	v1,
		value_t	v2,
		value_t	epsilon = g_epsilon )

Equivalent to \(v_1 - v_2 > \epsilon\).

Variable Documentation

no_normalize

no_normalize_t probfd::no_normalize = no_normalize_t{}

inlineconstexpr

Disambiguator tag for Distribution constructor to indicate that the probabilities are already normalized to one.

g_epsilon

value_t probfd::g_epsilon

extern

The default tolerance value for approximate comparisons.