State for sa-rectangular uncertainty (or no uncertainty) in an MDP. More...

#include <State.hpp>

Public Member Functions
	SAState (const vector< AType > &actions)
	Initializes state with actions and sets them all to valid.

size_t	action_count () const
	Number of actions.

size_t	size () const
	Number of actions.

AType &	create_action (long actionid)
	Creates an action given by actionid if it does not exists. More...

AType &	create_action ()
	Creates an action at the last position of the state.

const AType &	get_action (long actionid) const
	Returns an existing action.

const AType &	operator[] (long actionid) const
	Returns an existing action.

AType &	get_action (long actionid)
	Returns an existing action.

AType &	operator[] (long actionid)
	Returns an existing action.

bool	is_valid (long actionid) const
	Returns whether the actions is valid.

void	set_valid (long actionid, bool value=true)
	Set action validity. More...

const vector< AType > &	get_actions () const
	Returns set of all actions.

bool	is_terminal () const
	True if the state is considered terminal (no actions). More...

void	normalize ()
	Normalizes transition probabilities to sum to one. More...

bool	is_action_correct (long aid, numvec nataction) const
	Checks whether the prescribed action and outcome are correct.

bool	is_action_correct (long aid) const
	Checks whether the prescribed action correct.

prec_t	mean_reward (long actionid, numvec nataction) const
	Returns the mean reward following the action (and outcome). More...

prec_t	mean_reward (long actionid) const
	Returns the mean reward following the action. More...

Transition	mean_transition (long action, numvec nataction) const
	Returns the mean transition probabilities following the action and outcome. More...

Transition	mean_transition (long action) const
	Returns the mean transition probabilities following the action and outcome. More...

string	to_json (long stateid=-1) const
	Returns json representation of the state. More...

Protected Attributes
vector< AType >	actions
	list of actions

vector< bool >	valid
	whether actions can be used in computation. If false, that means that they should not be used in algorithms or in computation.

Detailed Description

template<class AType>
class craam::SAState< AType >

State for sa-rectangular uncertainty (or no uncertainty) in an MDP.

Actions within a state are sequentially labeled. That is, adding an action with an id = 3 will also create actions 0,1,2. These additional actions are marked as "invalid" state. This means that they will not be used in computations and algorithms will simply skip over them. Use create_action to make and action valid.

Template Parameters

Type	of action used in the state. This type determines the type of uncertainty set.

Member Function Documentation

◆ create_action()

template<class AType>

AType& craam::SAState< AType >::create_action ( long actionid )

inline

Creates an action given by actionid if it does not exists.

Otherwise returns the existing one.

All newly created actions with id < actionid are created as invalid (action.get_valid() = false). Action actionid is created as valid.

◆ is_terminal()

template<class AType>

bool craam::SAState< AType >::is_terminal ( ) const

inline

True if the state is considered terminal (no actions).

◆ mean_reward() [1/2]

template<class AType>

prec_t craam::SAState< AType >::mean_reward	(	long	actionid,
		numvec	nataction
	)		const

inline

Returns the mean reward following the action (and outcome).

◆ mean_reward() [2/2]

template<class AType>

prec_t craam::SAState< AType >::mean_reward ( long actionid ) const

inline

Returns the mean reward following the action.

◆ mean_transition() [1/2]

template<class AType>

Transition craam::SAState< AType >::mean_transition	(	long	action,
		numvec	nataction
	)		const

inline

Returns the mean transition probabilities following the action and outcome.

This class assumes a deterministic policy of the decision maker and a randomized policy of nature.

Parameters

action	Deterministic action of the decision maker
nataction	Randomized action of nature

◆ mean_transition() [2/2]

template<class AType>

Transition craam::SAState< AType >::mean_transition ( long action ) const

inline

Returns the mean transition probabilities following the action and outcome.

Parameters

action Deterministic action of decision maker

◆ normalize()

template<class AType>

void craam::SAState< AType >::normalize ( )

inline

Normalizes transition probabilities to sum to one.

◆ set_valid()

template<class AType>

void craam::SAState< AType >::set_valid	(	long	actionid,
		bool	value = `true`
	)

inline

Set action validity.

A valid action can be used in computations. An invalid action is just a placeholder.

◆ to_json()

template<class AType>

string craam::SAState< AType >::to_json ( long stateid = -1 ) const

inline

Returns json representation of the state.

Parameters

stateid Includes also state id

The documentation for this class was generated from the following file:

craam/State.hpp

Public Member Functions

Protected Attributes

Detailed Description

template<class AType> class craam::SAState< AType >

Member Function Documentation

◆ create_action()

◆ is_terminal()

◆ mean_reward() [1/2]

◆ mean_reward() [2/2]

◆ mean_transition() [1/2]

◆ mean_transition() [2/2]

◆ normalize()

◆ set_valid()

◆ to_json()

template<class AType>
class craam::SAState< AType >