pomdp_py.problems.multi_object_search.domain package

Submodules

pomdp_py.problems.multi_object_search.domain.action module

Defines the Action for the 2D Multi-Object Search domain;

Action space:

Motion \(\cup\) Look \(\cup\) Find

• Motion Actions scheme 1: South, East, West, North.

• Motion Actions scheme 2: Left 45deg, Right 45deg, Forward

• Look: Interprets sensor input as observation

• Find: Marks objects observed in the last Look action as (differs from original paper; reduces action space)

It is possible to force “Look” after every N/S/E/W action; then the Look action could be dropped. This is optional behavior.

class pomdp_py.problems.multi_object_search.domain.action.Action(name)

Bases: Action

Mos action; Simple named action.

class pomdp_py.problems.multi_object_search.domain.action.MotionAction(motion, scheme='xy', distance_cost=1, motion_name=None)

Bases: Action

SCHEME_XYTH = 'xyth'

EAST = (1, 0, 0)

WEST = (-1, 0, 3.141592653589793)

NORTH = (0, -1, 4.71238898038469)

SOUTH = (0, 1, 1.5707963267948966)

SCHEME_VW = 'vw'

FORWARD = (1, 0)

BACKWARD = (-1, 0)

LEFT = (0, -0.7853981633974483)

RIGHT = (0, 0.7853981633974483)

SCHEME_XY = 'xy'

EAST2D = (1, 0)

WEST2D = (-1, 0)

NORTH2D = (0, -1)

SOUTH2D = (0, 1)

SCHEMES = {'vw', 'xy', 'xyth'}

class pomdp_py.problems.multi_object_search.domain.action.LookAction

Bases: Action

class pomdp_py.problems.multi_object_search.domain.action.FindAction

Bases: Action

pomdp_py.problems.multi_object_search.domain.observation module

Defines the Observation for the 2D Multi-Object Search domain;

Origin: Multi-Object Search using Object-Oriented POMDPs (ICRA 2019) (extensions: action space changes, different sensor model, gridworld instead of topological graph)

Observation:

{objid : pose(x,y) or NULL}. The sensor model could vary; it could be a fan-shaped model as the original paper, or it could be something else. But the resulting observation should be a map from object id to observed pose or NULL (not observed).

class pomdp_py.problems.multi_object_search.domain.observation.ObjectObservation(objid, pose)

Bases: Observation

The xy pose of the object is observed; or NULL if not observed

NULL = None

class pomdp_py.problems.multi_object_search.domain.observation.MosOOObservation(objposes)

Bases: OOObservation

Observation for Mos that can be factored by objects; thus this is an OOObservation.

for_obj(objid)

factor(next_state, *params, **kwargs)

Factor this OO-observation by objects

classmethod merge(object_observations, next_state, *params, **kwargs)

Merge object_observations into a single OOObservation object;

object_observation (dict): Maps from objid to ObjectObservation

pomdp_py.problems.multi_object_search.domain.state module

Defines the State for the 2D Multi-Object Search domain;

Origin: Multi-Object Search using Object-Oriented POMDPs (ICRA 2019) (extensions: action space changes, different sensor model, gridworld instead of topological graph)

Description: Multi-Object Search in a 2D grid world.

State space:

\(S_1 \times S_2 \times ... S_n \times S_r\) where \(S_i (1\leq i\leq n)\) is the object state, with attribute “pose” \((x,y)\) and Sr is the state of the robot, with attribute “pose” \((x,y)\) and “objects_found” (set).

class pomdp_py.problems.multi_object_search.domain.state.ObjectState(objid, objclass, pose)

Bases: ObjectState

property pose

property objid

class pomdp_py.problems.multi_object_search.domain.state.RobotState(robot_id, pose, objects_found, camera_direction)

Bases: ObjectState

property pose

property robot_pose

property objects_found

class pomdp_py.problems.multi_object_search.domain.state.MosOOState(object_states)

Bases: OOState

object_pose(objid)

pose(objid)

property object_poses

Module contents

Defines state, action and observation