pomdp_py.problems.multi_object_search.agent package


pomdp_py.problems.multi_object_search.agent.agent module

class pomdp_py.problems.multi_object_search.agent.agent.MosAgent(robot_id, init_robot_state, object_ids, dim, sensor, sigma=0.01, epsilon=1, belief_rep='histogram', prior={}, num_particles=100, grid_map=None)[source]

Bases: Agent

One agent is one robot.


Custum function; clear history

pomdp_py.problems.multi_object_search.agent.belief module

class pomdp_py.problems.multi_object_search.agent.belief.MosOOBelief(robot_id, object_beliefs)[source]

Bases: OOBelief

This is needed to make sure the belief is sampling the right type of State for this problem.

mpe(self, return_oostate=False, **kwargs)[source]

Returns most likely state.

random(self, return_oostate=False, **kwargs)[source]

Returns a random state

pomdp_py.problems.multi_object_search.agent.belief.initialize_belief(dim, robot_id, object_ids, prior={}, representation='histogram', robot_orientations={}, num_particles=100)[source]

Returns a GenerativeDistribution that is the belief representation for the multi-object search problem.

  • dim (tuple) – a tuple (width, length) of the search space gridworld.

  • robot_id (int) – robot id that this belief is initialized for.

  • object_ids (dict) – a set of object ids that we want to model the belief distribution over; They are assumed to be the target objects, not obstacles, because the robot doesn’t really care about obstacle locations and modeling them just adds computation cost.

  • prior (dict) – A mapping {(objid|robot_id) -> {(x,y) -> [0,1]}}. If used, then all locations not included in the prior will be treated to have 0 probability. If unspecified for an object, then the belief over that object is assumed to be a uniform distribution.

  • robot_orientations (dict) – Mapping from robot id to their initial orientation (radian). Assumed to be 0 if robot id not in this dictionary.

  • num_particles (int) – Maximum number of particles used to represent the belief


the initial belief representation.

Return type:


Module contents