refine_plan.models.semi_mdp
Class for semi-MDP based on options.
Author: Charlie Street Owner: Charlie Street
Module Contents
Classes
Class for semi-MDPs where actions=options. |
- class refine_plan.models.semi_mdp.SemiMDP(sf_list, option_list, labels, initial_state=None)
Bases:
objectClass for semi-MDPs where actions=options.
- _state_factors
A dictionary from state factor name to state factor
- _options
A dictionary from option name to options
- _labels
A list of Labels
- _initial_state
A deterministic initial state (if there is one)
- get_state_factors()
Return the state factor dictionary.
- Returns:
A dictionary from state factor name to state factor
- get_det_initial_state()
Return the deterministic initial state, if there is one.
- Returns:
The initial state if there is one, None otherwise
- get_labels()
Return the list of labels.
- Returns:
The list of labels
- get_transition_prob(state, option, next_state)
Get the transition probability for executing an option in a state.
- Parameters:
state – The current state
option – The name of the option (action) being executed
next_state – The successor state
- Returns:
The semi-MDP transition probability
- Raises:
invalid_opt_exception – Raised if an invalid option is passed in
- get_reward(state, option)
Get the reward for executing an option in a state.
- Parameters:
state – The current state
option – The name of the option (action) being executed
- Returns:
The reward for that option in the semi-MDP
- Raises:
invalid_opt_exception – Raised if an invalid option is passed in
- to_prism_string(output_file=None)
Convert the semi-MDP into a PRISM model.
Because of the objectives we are handling this is written as an MDP.
- Parameters:
output_file – Optional. A file path to write the PRISM string to.
- Returns:
The PRISM string