stormvogel.model.reward_model

Reward model.

Classes

RewardModel

Represent a reward model supporting state rewards and transition rewards.

Module Contents

class stormvogel.model.reward_model.RewardModel[ValueType: stormvogel.model.value.Value]

Represent a reward model supporting state rewards and transition rewards.

Parameters:

  • name – Name of the reward model.

  • model – The model this reward model belongs to.

  • rewards – Mapping from states to their state reward values.

  • transition_rewards – Sparse mapping from (source, action, target) triples to reward values collected on that specific transition.

name: str
model: stormvogel.model.model.Model
rewards: dict[stormvogel.model.state.State, ValueType]
transition_rewards: dict[tuple[stormvogel.model.state.State, stormvogel.model.action.Action, stormvogel.model.state.State], ValueType]
set_from_rewards_vector(vector: list[ValueType], state_action: bool = False) → None

Set the rewards of this model according to a (stormpy) rewards vector.

Parameters:

  • vector – The reward values.

  • state_action – If True, the vector has one entry per (state, action) pair; only the first entry for each state is used.

get_state_reward(state: stormvogel.model.state.State) → ValueType | None

Get the reward at the given state.

Parameters:

state – The state to look up.

Returns:

The reward value, or None if no reward is present.

set_state_reward(state: stormvogel.model.state.State, value: ValueType)

Set the reward at the given state.

Parameters:

  • state – The state to set the reward for.

  • value – The reward value to assign.

set_unset_rewards(value: ValueType)

Fill up rewards that were not set yet with the specified value.

Parameters:

value – The default reward value to assign to unset states.

has_transition_rewards() → bool

Return True if any nonzero transition reward is present.

set_transition_reward(s: stormvogel.model.state.State, a: stormvogel.model.action.Action, s_next: stormvogel.model.state.State, value: ValueType) → None

Set the reward collected when transitioning from s via a to s_next.

get_transition_reward(s: stormvogel.model.state.State, a: stormvogel.model.action.Action, s_next: stormvogel.model.state.State) → ValueType

Return the transition reward for (s, a, s_next), or 0 if absent.

__iter__()
get_reward_vector() → list[float]

Return a list of all rewards ordered appropriately.

Returns:

A flat list of reward values as floats.