Observations¶

MTDVRP observations.

Observations settings are defined in file observations.py.

Observations¶

class maenvs4vrp.environments.mtdvrp.observations.Observations(feature_list: Dict = None)[source]¶

MTDVRP observations class.

__init__(feature_list: Dict = None)[source]¶

Constructor

Parameters:: feature_list (Dict) – Dictionary containing observation features list to be available to the agent. Defaults to None.

set_env(env: AECEnv)[source]¶

Set environment.

Parameters:: env (AECEnv) – Environment.
Returns:: None.

Nodes static features¶

Observations.get_feat_x_coordinate()[source]¶

Instance nodes X coordinates.

Parameters:: n/a.
Returns:: Instance nodes X coordinates.
Return type:: torch.Tensor

Observations.get_feat_y_coordinate()[source]¶

Instance nodes Y coordinates.

Parameters:: n/a.
Returns:: Instance nodes Y coordinates.
Return type:: torch.Tensor

Observations.get_feat_x_coordinate_min_max()[source]¶

Min-max normalized X coordinates of instance nodes.

Parameters:: n/a.
Returns:: Min. and max. x coordinates of instance nodes.
Return type:: torch.Tensor

Observations.get_feat_y_coordinate_min_max()[source]¶

Min-max normalized Y coordinates of instance nodes.

Parameters:: n/a.
Returns:: Min-max normalized Y coordinates of instance nodes.
Return type:: torch.Tensor

Observations.get_feat_tw_low()[source]¶

Nodes time windows starting times.

Parameters:: n/a.
Returns:: Nodes time windows starting times.
Return type:: torch.Tensor

Observations.get_feat_tw_high()[source]¶

Nodes time windows ending times.

Parameters:: n/a.
Returns:: Nodes time windows ending times.
Return type:: torch.Tensor

Observations.get_feat_linehaul_demand()[source]¶

Nodes linehaul demands.

Parameters:: n/a.
Returns:: Nodes linehaul demands.
Return type:: torch.Tensor

Observations.get_feat_backhaul_demand()[source]¶

Nodes backhaul demands.

Parameters:: n/a.
Returns:: Nodes backhaul demands.
Return type:: torch.Tensor

Observations.get_feat_service_time()[source]¶

Nodes service times.

Parameters:: n/a.
Returns:: Nodes service times.
Return type:: torch.Tensor

Observations.get_feat_tw_high_minus_tw_low_div_max_dur()[source]¶

Nodes time window amplitude divided by max tour duration.

Parameters:: n/a.
Returns:: Nodes time window amplitude divided by max tour duration.
Return type:: torch.Tensor

Observations.get_feat_is_depot()[source]¶

Checks if node is depot.

Parameters:: n/a.
Returns:: If the node is depot or not.
Return type:: torch.Tensor

Nodes dynamic features¶

Observations.get_feat_time2open_div_end_time()[source]¶

Nodes time to open divided by end time.

Parameters:: n/a.
Returns:: Nodes time to open divided by end time.
Return type:: torch.Tensor

Observations.get_feat_time2close_div_end_time()[source]¶

Nodes time to close divided by end time.

Parameters:: n/a.
Returns:: Nodes time to close divided by end time.
Return type:: torch.Tensor

Observations.get_feat_arrive2node_div_end_time()[source]¶

Agent arriving time to nodes divided by end time.

Parameters:: n/a.
Returns:: Agent arriving time to nodes divided by end time.
Return type:: torch.Tensor

Observations.get_feat_time2open_after_step_div_end_time()[source]¶

Nodes time to open, after agent step, divided by end time.

Parameters:: n/a.
Returns:: Nodes time to open, after agent step, divided by end time.
Return type:: torch.Tensor

Observations.get_feat_time2close_after_step_div_end_time()[source]¶

Nodes time to close, after agent step, divided by end time.

Parameters:: n/a.
Returns:: Nodes time to close, after agent step, divided by end time.
Return type:: torch.Tensor

Observations.get_feat_time2end_after_step_div_end_time()[source]¶

Time end, after agent step to node, divided by end time.

Parameters:: n/a.
Returns:: Time end, after agent step to node, divided by end time.
Return type:: torch.Tensor

Observations.get_feat_fract_time_after_step_div_end_time()[source]¶

Fraction of time left, after agent step to node.

Parameters:: n/a.
Returns:: Fraction of time left, after agent step to node.
Return type:: torch.Tensor

Observations.get_feat_reachable_frac_agents()[source]¶

Feasible nodes per agent.

Parameters:: n/a.
Returns:: Feasible nodes per agent.
Return type:: torch.Tensor

Current agent features¶

Observations.get_feat_agent_x_coordinate()[source]¶

Current agent X coordinate.

Parameters:: n/a.
Returns:: Current agent X coordinate.
Return type:: torch.Tensor

Observations.get_feat_agent_y_coordinate()[source]¶

Current agent Y coordinate.

Parameters:: n/a.
Returns:: Current agent Y coordinate.
Return type:: torch.Tensor

Observations.get_feat_agent_x_coordinate_min_max()[source]¶

Current agent min-max normalized X location.

Parameters:: n/a.
Returns:: Current agent min-max normalized X location.
Return type:: torch.Tensor

Observations.get_feat_agent_y_coordinate_min_max()[source]¶

Current agent min-max normalized Y location.

Parameters:: n/a.
Returns:: Current agent min-max normalized Y location.
Return type:: torch.Tensor

Observations.get_feat_agent_frac_current_time()[source]¶

Agent fraction of time elapsed.

Parameters:: n/a.
Returns:: Agent fraction of time elapsed.
Return type:: torch.Tensor

Observations.get_feat_agent_used_capacity_linehaul()[source]¶

Used linehaul capacity by agent.

Parameters:: n/a.
Returns:: Used linehaul capacity.
Return type:: torch.Tensor

Observations.get_feat_agent_used_capacity_backhaul()[source]¶

Used backhaul capacity by agent.

Parameters:: n/a.
Returns:: Used backhaul capacity.
Return type:: torch.Tensor

Observations.get_feat_agent_avaliable_load()[source]¶

Agent’s avaliable load when problem is unmixed.

Parameters:: n/a.
Returns:: Avaliable load.
Return type:: torch.Tensor

Observations.get_feat_agent_avaliable_load_vrpmpd()[source]¶

Agent’s avaliable load when problem is mixed.

Parameters:: n/a.
Returns:: Avaliable load.
Return type:: torch.Tensor

Observations.get_feat_agent_arrivedepot_div_end_time()[source]¶

Agent time to depot divided by end time.

Parameters:: n/a.
Returns:: Agent time to depot divided by end time.
Return type:: torch.Tensor

Observations.get_feat_agent_frac_feasible_nodes()[source]¶

Fraction of current agent feasible nodes, in order to the total number of instance nodes.

Parameters:: n/a.
Returns:: Fraction of current agent feasible nodes, in order to the total number of instance nodes.
Return type:: torch.Tensor

Observations.get_feat_agent_remaining_dist()[source]¶

Agent’s remaining distance.

Parameters:: n/a.
Returns:: Remaining distance.
Return type:: torch.Tensor

Observations.get_feat_agent_current_time()[source]¶

Agent’s current time.

Parameters:: n/a.
Returns:: Current time.
Return type:: torch.Tensor

Other agents features¶

Observations.get_feat_agents_x_coordinate()[source]¶

Agents X coordinates.

Parameters:: n/a.
Returns:: Agents X coordinates.
Return type:: torch.Tensor

Observations.get_feat_agents_y_coordinate()[source]¶

Agents Y coordinates.

Parameters:: n/a.
Returns:: Agents Y coordinates.
Return type:: torch.Tensor

Observations.get_feat_agents_x_coordinate_min_max()[source]¶

Agents min-max normalized X location.

Parameters:: n/a.
Returns:: Agents min-max normalized X location.
Return type:: torch.Tensor

Observations.get_feat_agents_y_coordinate_min_max()[source]¶

Agents min-max normalized Y location.

Parameters:: n/a.
Returns:: Agents min-max normalized Y location.
Return type:: torch.Tensor

Observations.get_feat_agents_frac_current_time()[source]¶

Agents fraction of elapsed time.

Parameters:: n/a.
Returns:: Agents fraction of elapsed time.
Return type:: torch.Tensor

Observations.get_feat_agents_dist2depot_div_end_time()[source]¶

Fraction of current agent distance to depot compared to its end time.

Parameters:: n/a.
Returns:: Fraction of current agent distance to depot compared to its end time.
Return type:: torch.Tensor

Observations.get_feat_agents_dist2agent_div_end_time()[source]¶

Agents distance to active agent divided by end time.

Parameters:: n/a.
Returns:: Agents distance to active agent divided by end time.
Return type:: torch.Tensor

Observations.get_feat_agents_frac_feasible_nodes()[source]¶

Fraction of agents feasible nodes, in order to the total number of instance nodes.

Parameters:: n/a.
Returns:: Fraction of agents feasible nodes, in order to the total number of instance nodes.
Return type:: torch.Tensor

Observations.get_feat_agents_time_delta2agent_div_max_dur()[source]¶

Difference between agents time and current agent time, divided by max. tour duration.

Parameters:: n/a.
Returns:: Difference between agents time and current agent time, divided by max. tour duration.
Return type:: torch.Tensor

Observations.get_feat_agents_was_last()[source]¶

Last agent performing an action.

Parameters:: n/a.
Returns:: Last agent performing an action.
Return type:: torch.Tensor

Global features¶

Observations.get_feat_global_frac_done_agents()[source]¶

Fraction of done agents.

Parameters:: n/a.
Returns:: Fraction of done agents.
Return type:: torch.Tensor

Observations.get_feat_global_frac_linehaul_demands()[source]¶

Fraction of served demands.

Parameters:: n/a.
Returns:: Fraction of served demands.
Return type:: torch.Tensor

Observations.get_feat_global_frac_backhaul_demands()[source]¶

Fraction of served demands.

Parameters:: n/a.
Returns:: Fraction of served demands.
Return type:: torch.Tensor

Observations.get_feat_global_open_routes()[source]¶

Checks if problems have open routes.

Parameters:: n/a.
Returns:: Open routes.
Return type:: torch.Tensor

Observations.get_feat_global_distance_limits()[source]¶

Chack problems distance limits.

Parameters:: n/a.
Returns:: Distance limits.
Return type:: torch.Tensor

Observations.get_feat_global_max_tw_depot()[source]¶

High tw from depot. Max tour duration.

Parameters:: N/a.
Returns:: Max tour duration.
Return type:: torch.Tensor

Observations.get_feat_global_is_problem_backhaul_mixed()[source]¶

Checks if problem is mixed. 0 means it’s unmixed, 1 means it’s mixed.

Parameters:: n/a.
Returns:: Is problem mixed?
Return type:: torch.Tensor

Computing features¶

Observations.compute_static_features()[source]¶

Compute nodes static features.

Parameters:: n/a.
Returns:: Nodes static features.
Return type:: torch.Tensor

Observations.compute_dynamic_features()[source]¶

Compute nodes dynamic features.

Parameters:: n/a.
Returns:: Nodes dynamic features.
Return type:: torch.Tensor

Observations.compute_agent_features()[source]¶

Compute current agent features.

Parameters:: n/a.
Returns:: Current agent features.
Return type:: torch.Tensor

Observations.compute_agents_features()[source]¶

Compute other agent features.

Parameters:: n/a.
Returns:: Other agent features.
Return type:: torch.Tensor

Observations.compute_global_features()[source]¶

Compute global features.

Parameters:: n/a.
Returns:: Global features.
Return type:: torch.Tensor

Observations.get_observations(is_reset=False) → TensorDict[source]¶

Compute the environment.

Parameters:: is_reset (bool) – If the environment is on reset. Defauts to False.

Returns: observations(TensorDict): Current environment observations and masks dictionary.

Internal methods¶

static Observations._concat_features(features)[source]¶

Concatenate features.

Parameters:: features (list) – Features to concatenate.
Returns:: Concatenated tensor.
Return type:: torch.Tensor

Observations._normalize_feature(x, norm)[source]¶

Normalize features.

Parameters:

x (torch.Tensor) – Tensor to be normalized.
norm (str) – Type of normalization. It can be ‘min_max’ or ‘standardize’. If None, tensor is returned.

Returns:

Tensor normalized or default tensor if norm is invalid.

Return type:

torch.Tensor

static Observations._min_max_normalization(x)[source]¶

Min. max. normalization.

Parameters:: x (torch.Tensor) – Tensor to be normalized.
Returns:: Normalized tensor.
Return type:: torch.Tensor

static Observations._min_max_normalization2d(x)[source]¶

Min. max. normalization 2 dimensions.

Parameters:: x (torch.Tensor) – Tensor to be normalized.
Returns:: Normalized tensor.
Return type:: torch.Tensor

static Observations._standardize(x)[source]¶

Tensor standardization.

Parameters:: x (torch.Tensor) – Tensor to be normalized.
Returns:: Normalized tensor.
Return type:: torch.Tensor