interventions.InterventionApplier

class interventions.InterventionApplier(state_manager, shift_reward)

Bases: object

Lightweight intervention applier for symbolic state updates.

This class parses intervention actions and applies them to a StateManager. For Scenario 1, it focuses on directional shifts (left, right, forward, back).

Attributes:

state_manager: StateManager instance for tracking state changes. shift_reward: Reward bonus for successful shifts (currently unused).

__init__(state_manager, shift_reward)

Initialize the intervention applier.

Args:

state_manager: StateManager instance to apply interventions to. shift_reward: Reward bonus for successful shifts (default: 0.0).

Methods

__init__(state_manager, shift_reward)	Initialize the intervention applier.
apply(obj, action)	Apply an intervention to an object.
parse_action(action)	Parse an action string into direction and magnitude.

apply(obj, action)

Apply an intervention to an object.

Parses the action and applies it to the object via the StateManager.

Args:

obj: Name of the object to intervene on (e.g., ‘1’, ‘2’, ‘3’). action: Action string (e.g., ‘left,0.005’).

Returns:

Tuple of (success, reward_delta):

• success: True if intervention was successfully applied

• reward_delta: Immediate reward from this intervention

parse_action(action)

Parse an action string into direction and magnitude.

Args:

action: Action string (e.g., ‘left,0.005’, ‘right,0.01’).

Returns:

Tuple of (direction, magnitude), or (None, None) if parsing fails.

Examples:

>>> parse_action('left,0.005')
('left', 0.005)
>>> parse_action('forward,0.02')
('forward', 0.02)