PomdpX

class pgmpy.readwrite.PomdpX.PomdpXReader(path=None, string=None)

Initialize an instance of PomdpX reader class

Parameters:

• path (file or str) – Path of the file containing PomdpX information.

• string (str) – String containing PomdpX information.

Example

reader = PomdpXReader(‘TestPomdpX.xml’)

Reference

http://bigbird.comp.nus.edu.sg/pmwiki/farm/appl/index.php?n=Main.PomdpXDocumentation

get_description()

Return the problem description

Examples

>>> reader = PomdpXReader('Test_Pomdpx.xml')
>>> reader.get_description()
'RockSample problem for map size 1 x 3.
Rock is at 0, Rover’s initial position is at 1.
Exit is at 2.'
>>> reader = PomdpXReader('Test_PomdpX.xml')
>>> reader.get_description()
'RockSample problem for map size 1 x 3.
 Rock is at 0, Rover’s initial position is at 1.
 Exit is at 2.'

get_discount()

Returns the discount factor for the problem

Example

>>> reader = PomdpXReader('Test_PomdpX.xml')
>>> reader.get_discount()
0.95

get_initial_beliefs()

Returns the state, action and observation variables as a dictionary in the case of table type parameter and a nested structure in case of decision diagram parameter

Examples

>>> reader = PomdpXReader('Test_PomdpX.xml')
>>> reader.get_initial_beliefs()
[{'Var': 'rover_0',
  'Parent': ['null'],
  'Type': 'TBL',
  'Parameter': [{'Instance': ['-'],
  'ProbTable': ['0.0', '1.0', '0.0']}]
 },
 {'Var': '',
  '...': ...,'
  '...': '...',
  }]

get_obs_function()

Returns the observation function as nested dict in the case of table- type parameter and a nested structure in case of decision diagram parameter

Example

>>> reader = PomdpXReader('Test_PomdpX.xml')
>>> reader.get_obs_function()
[{'Var': 'obs_sensor',
      'Parent': ['action_rover', 'rover_1', 'rock_1'],
      'Type': 'TBL',
      'Parameter': [{'Instance': ['amw', '*', '*', '-'],
                     'ProbTable': ['1.0', '0.0']},
                 ...
                ]
}]

get_parameter(var)

This method supports the functional tags by providing the actual values in the function as list of dict in case of table type parameter or as nested dict in case of decision diagram

get_parameter_dd(parameter)

This method returns parameters as nested dicts in case of decision diagram parameter.

get_parameter_tbl(parameter)

This method returns parameters as list of dict in case of table type parameter

get_reward_function()

Returns the reward function as nested dict in the case of table- type parameter and a nested structure in case of decision diagram parameter

Example

>>> reader = PomdpXReader('Test_PomdpX.xml')
>>> reader.get_reward_function()
[{'Var': 'reward_rover',
      'Parent': ['action_rover', 'rover_0', 'rock_0'],
      'Type': 'TBL',
      'Parameter': [{'Instance': ['ame', 's1', '*'],
                     'ValueTable': ['10']},
                 ...
                ]
}]

get_state_transition_function()

Returns the transition of the state variables as nested dict in the case of table type parameter and a nested structure in case of decision diagram parameter

Example

>>> reader = PomdpXReader('Test_PomdpX.xml')
>>> reader.get_state_transition_function()
[{'Var': 'rover_1',
  'Parent': ['action_rover', 'rover_0'],
  'Type': 'TBL',
  'Parameter': [{'Instance': ['amw', 's0', 's2'],
                 'ProbTable': ['1.0']},
                 {'Instance': ['amw', 's1', 's0'],
                 'ProbTable': ['1.0']},
                 ...
                ]
}]

get_variables()

Returns list of variables of the network

Example

>>> reader = PomdpXReader("pomdpx.xml")
>>> reader.get_variables()
{'StateVar': [
                {'vnamePrev': 'rover_0',
                 'vnameCurr': 'rover_1',
                 'ValueEnum': ['s0', 's1', 's2'],
                 'fullyObs': True},
                {'vnamePrev': 'rock_0',
                 'vnameCurr': 'rock_1',
                 'fullyObs': False,
                 'ValueEnum': ['good', 'bad']}],
                'ObsVar': [{'vname': 'obs_sensor',
                            'ValueEnum': ['ogood', 'obad']}],
                'RewardVar': [{'vname': 'reward_rover'}],
                'ActionVar': [{'vname': 'action_rover',
                               'ValueEnum': ['amw', 'ame',
                                             'ac', 'as']}]
                }

class pgmpy.readwrite.PomdpX.PomdpXWriter(model_data, encoding='utf-8', prettyprint=True)

Initialise a PomdpXWriter Object

Parameters:

model (A Bayesian of Markov Model) – The model to write

encoding: String(optional)

Encoding for text data

prettyprint: Bool(optional)

Indentation in output XML if true

add_conditions(condition, condprob)

helper function for adding probability conditions for model :param condition: contains and element of conditions list :type condition: dictionary :param condprob: the tag to which condition is added :type condprob: etree SubElement

Return type:

None

add_initial_belief()

add initial belief tag to pomdpx model

Return type:

string containing the xml for initial belief tag

add_obs_function()

add observation function tag to pomdpx model

Return type:

string containing the xml for observation function tag

add_parameter_dd(dag_tag, node_dict)

helper function for adding parameters in condition

Parameters:

• dag_tag (etree SubElement) – the DAG tag is contained in this subelement

• node_dict (dictionary) – the decision diagram dictionary

Return type:

None

add_reward_function()

add reward function tag to pomdpx model

Return type:

string containing the xml for reward function tag

add_state_transition_function()

add state transition function tag to pomdpx model

Return type:

string containing the xml for state transition tag

get_variables()

Add variables to PomdpX

Return type:

xml containing variables tag

indent(elem, level=0)

Inplace prettyprint formatter.