Intervention Actors

BaseInterventionActorPolicy

class causal_world.intervention_actors.BaseInterventionActorPolicy(**kwargs)

__init__(**kwargs)

This class indicates the interface of an intervention actor

Parameters

kwargs – (params) parameters for the construction of the actor.

act(variables_dict)

This functions enables the intervention actor to decide on specific interventions.

Parameters

variables_dict – (dict) The current dict of variables that it can intervene on with their current values. (this can be a two level dict)

Returns

(dict) interventions decided by the intervention actor to intervene on.

get_params()

returns parameters that could be used in recreating this intervention actor.

Returns

(dict) specifying paramters to create this intervention actor again.

initialize(env)

This functions allows the intervention actor to query things from the env, such as intervention spaces or to have access to sampling funcs for goals..etc

Parameters

env – (causal_world.env.CausalWorld) the environment used for the intervention actor to query different methods from it.

Returns

RandomInterventionActorPolicy

class causal_world.intervention_actors.RandomInterventionActorPolicy(**kwargs)

__init__(**kwargs)

This is a random intervention actor which intervenes randomly on all available state variables except joint positions since its a trickier space.

Parameters

kwargs –

get_params()

returns parameters that could be used in recreating this intervention actor.

Returns

(dict) specifying paramters to create this intervention actor again.

initialize(env)

This functions allows the intervention actor to query things from the env, such as intervention spaces or to have access to sampling funcs for goals..etc

Parameters

env – (causal_world.env.CausalWorld) the environment used for the intervention actor to query different methods from it.

Returns

GoalInterventionActorPolicy

class causal_world.intervention_actors.GoalInterventionActorPolicy(**kwargs)

__init__(**kwargs)

This class indicates the goal intervention actor, which an intervention actor that intervenes by sampling a new goal.

Parameters

kwargs – (params) parameters for the construction of the actor.

get_params()

returns parameters that could be used in recreating this intervention actor.

Returns

(dict) specifying paramters to create this intervention actor again.

initialize(env)

This functions allows the intervention actor to query things from the env, such as intervention spaces or to have access to sampling funcs for goals..etc

Parameters

env – (causal_world.env.CausalWorld) the environment used for the intervention actor to query different methods from it.

Returns

JointsInterventionActorPolicy

class causal_world.intervention_actors.JointsInterventionActorPolicy(**kwargs)

__init__(**kwargs)

This class indicates the joint intervention actor which intervenes on the joints of the robot in a random fashion.

Parameters

kwargs –

get_params()

returns parameters that could be used in recreating this intervention actor.

Returns

(dict) specifying paramters to create this intervention actor again.

initialize(env)

This functions allows the intervention actor to query things from the env, such as intervention spaces or to have access to sampling funcs for goals..etc

Parameters

env – (causal_world.env.CausalWorld) the environment used for the intervention actor to query different methods from it.

Returns

RigidPoseInterventionActorPolicy

class causal_world.intervention_actors.RigidPoseInterventionActorPolicy(positions=True, orientations=True, **kwargs)

__init__(positions=True, orientations=True, **kwargs)

This intervention actor intervenes on the pose of the blocks available in the arena.

Parameters

• positions – (bool) True if interventions on positions should be allowed.

• orientations – (bool) True if interventions on orientations should be allowed.

• kwargs –

get_params()

returns parameters that could be used in recreating this intervention actor.

Returns

(dict) specifying paramters to create this intervention actor again.

initialize(env)

This functions allows the intervention actor to query things from the env, such as intervention spaces or to have access to sampling funcs for goals..etc

Parameters

env – (causal_world.env.CausalWorld) the environment used for the intervention actor to query different methods from it.

Returns

VisualInterventionActorPolicy

class causal_world.intervention_actors.VisualInterventionActorPolicy(**kwargs)

__init__(**kwargs)

This intervention actor intervenes on all visual components of the robot, (i.e: colors).

Parameters

kwargs –

get_params()

returns parameters that could be used in recreating this intervention actor.

Returns

(dict) specifying paramters to create this intervention actor again.

initialize(env)

This functions allows the intervention actor to query things from the env, such as intervention spaces or to have access to sampling funcs for goals..etc

Parameters

env – (causal_world.env.CausalWorld) the environment used for the intervention actor to query different methods from it.

Returns

PhysicalPropertiesInterventionActorPolicy

class causal_world.intervention_actors.PhysicalPropertiesInterventionActorPolicy(group, **kwargs)

__init__(group, **kwargs)

This intervention actor intervenes on physcial proporties such as friction, mass…etc

Parameters

• group – (str) the object that the actor will intervene on. floor, stage, robot..etc

• kwargs –

get_params()

returns parameters that could be used in recreating this intervention actor.

Returns

(dict) specifying paramters to create this intervention actor again.

initialize(env)

This functions allows the intervention actor to query things from the env, such as intervention spaces or to have access to sampling funcs for goals..etc

Parameters

env – (causal_world.env.CausalWorld) the environment used for the intervention actor to query different methods from it.

Returns