pyrieef.learning package

Subpackages

• pyrieef.learning.timeseries package
  • Submodules
  • pyrieef.learning.timeseries.functions module
  • pyrieef.learning.timeseries.gmm module
  • pyrieef.learning.timeseries.gmr module
  • pyrieef.learning.timeseries.hmm module
  • pyrieef.learning.timeseries.model module
  • pyrieef.learning.timeseries.plot module
  • Module contents

Submodules

pyrieef.learning.common_imports module

pyrieef.learning.dataset module

class pyrieef.learning.dataset.CostmapDataset(filename)

Bases: object

property epochs_completed

next_batch(batch_size, shuffle=True)

Return the next batch_size examples from this data set.

normalize_maps()

normalize all maps

property num_examples

reshape_data_to_tensors()

split_data(data)

Load datasets afresh, train_per should be between 0 and 1

class pyrieef.learning.dataset.WorkspaceData

Bases: object

pyrieef.learning.dataset.create_circles_workspace(box, ws)

Creates circle dataset from array

pyrieef.learning.dataset.get_yaml_options()

pyrieef.learning.dataset.import_tf_data(filename='costdata2d_10k.hdf5')

Works with version 1.9.0rc1

pyrieef.learning.dataset.learning_data_dir()

pyrieef.learning.dataset.load_data_from_file(filename='costdata2d_10k.hdf5')

pyrieef.learning.dataset.load_data_from_hdf5(filename, train_per)

Setup training / test data

pyrieef.learning.dataset.load_dictionary_from_file(filename='costdata2d_10k.hdf5')

pyrieef.learning.dataset.load_paths_from_file(filename='paths_1k_demos.hdf5')

pyrieef.learning.dataset.load_trajectories_from_file(filename='trajectories_1k_small.hdf5')

Load data from an hdf5 file

pyrieef.learning.dataset.load_workspace_dataset(basename='1k_small.hdf5')

pyrieef.learning.dataset.load_workspaces_from_file(filename='workspaces_1k_small.hdf5')

Load data from an hdf5 file

pyrieef.learning.dataset.save_paths_to_file(paths, filename='paths_1k_demos.hdf5')

pyrieef.learning.dataset.save_trajectories_to_file(trajectories, filename='trajectories_1k_demos.hdf5')

pyrieef.learning.dataset.write_data_to_file(data_out, filename='costdata2d_10k.hdf5')

pyrieef.learning.dataset.write_dictionary_to_file(data_out, filename='costdata2d_10k.hdf5')

pyrieef.learning.demonstrations module

pyrieef.learning.demonstrations.compute_demonstration(workspace, graph, nb_points, show_result, average_cost, verbose, no_linear_interpolation)

pyrieef.learning.demonstrations.generate_demonstrations(nb_points)

pyrieef.learning.demonstrations.generate_one_demonstration(nb_points, demo_id)

pyrieef.learning.demonstrations.getrandbits(k) → x. Generates an int with k random bits.

pyrieef.learning.demonstrations.obsatcle_potential(workspace)

pyrieef.learning.demonstrations.optimize(path, workspace, costmap, verbose=False)

pyrieef.learning.demonstrations.random() → x in the interval [0, 1).

pyrieef.learning.inverse_optimal_control module

class pyrieef.learning.inverse_optimal_control.InverseOptimalControl(nb_demonstrations)

Bases: object

Abstract class for IOC problems

abstract on_step()

abstract solution(env_id)

class pyrieef.learning.inverse_optimal_control.Learch(nb_demonstrations)

Bases: pyrieef.learning.inverse_optimal_control.InverseOptimalControl

one_step(iteration)

abstract planning(env_id)

abstract supervised_learning()

class pyrieef.learning.inverse_optimal_control.Learch2D(dataset)

Bases: pyrieef.learning.inverse_optimal_control.Learch

initialize_data()

planning()

supervised_learning()

pyrieef.learning.inverse_optimal_control.goodness_map(trajectory, nb_points, box, goodness_scalar, goodness_stddev)

pyrieef.learning.networks module

pyrieef.learning.one_demo module

pyrieef.learning.random_environment module

class pyrieef.learning.random_environment.RandomEnvironmentOptions(dataset_id=None)

Bases: object

environment_parser()

get_options()

Load dataset options from file or option parser

pyrieef.learning.random_environment.chomp_obstacle_cost(min_dist, epsilon)

Compute the cost function now (From CHOMP paper) If min_dist < 0, cost = -min_dist + epsilon/2 If min_dist >= 0 && min_dist < epsilon, have a different cost If min_dist >= epsilon, cost = 0

pyrieef.learning.random_environment.get_dataset_id(data_id)

pyrieef.learning.random_environment.getrandbits(k) → x. Generates an int with k random bits.

pyrieef.learning.random_environment.grids(workspace, grid_to_world, epsilon)

Creates a boolean matrix of occupancies To convert it to int or floats, use the following matrix.astype(int) matrix.astype(float)

pyrieef.learning.random_environment.random() → x in the interval [0, 1).

pyrieef.learning.random_environment.random_environments(opt)

pyrieef.learning.random_environment.remove_file_if_exists(file)

pyrieef.learning.random_environment.sample_circle_workspace(box, nobjs_max=3, random_max=False, maxnumtries=100)

Samples a workspace made of a maximum of nobjs_max circles that do not intersect todo replace the random environment script to use this function

pyrieef.learning.random_environment.samplerandpt(lims)

Sample a random point within limits

pyrieef.learning.random_paths module

pyrieef.learning.random_paths.cost_grid(workspace, nb_points)

pyrieef.learning.random_paths.generate_paths(nb_points)

pyrieef.learning.random_paths.getrandbits(k) → x. Generates an int with k random bits.

pyrieef.learning.random_paths.graph_search_path(graph, workspace, nb_points)

Find feasible path using Dijkstra’s algorithm

1. samples a path that has collision with the enviroment

   and perform graph search on a grid (nb_points x nb_points)

2. convert path to world coordinates

3. interpolate path continuously

pyrieef.learning.random_paths.grid_to_world_path(workspace, path, nb_points)

pyrieef.learning.random_paths.obsatcle_potential(workspace)

pyrieef.learning.random_paths.random() → x in the interval [0, 1).

pyrieef.learning.random_paths.sample_path(workspace, graph, nb_points, no_linear_interpolation)

finds a path that does not collide with enviroment but that is significantly difficult to perform

pyrieef.learning.tf_autoencoder module

pyrieef.learning.tf_autoencoder_2 module

pyrieef.learning.tf_costpredict module

pyrieef.learning.tf_costpredict_2 module

pyrieef.learning.tf_costpredict_3 module

pyrieef.learning.tf_costpredict_4 module

pyrieef.learning.tf_networks module

class pyrieef.learning.tf_networks.ConvDeconv64

Bases: pyrieef.learning.tf_networks.Network

Defines a Convolution Deconvolution network sized for the Mnist 28x28 matrix format

define(x_input)

lrelu(x, alpha=0.3)

placeholder()

resize_batch(imgs)

resize_output(imgs, i)

class pyrieef.learning.tf_networks.ConvDeconvResize

Bases: pyrieef.learning.tf_networks.Network

https://towardsdatascience.com/ autoencoders-introduction-and-implementation-3f40483b0a85

define(x_inputs)

placeholder()

resize_batch(imgs)

resize_output(imgs, i)

class pyrieef.learning.tf_networks.ConvDeconvSmall

Bases: pyrieef.learning.tf_networks.Network

Defines a Convolution Deconvolution network sized for the Mnist 28x28 matrix format

define(x_input)

input()

placeholder()

resize_batch(imgs)

resize_output(imgs, i)

class pyrieef.learning.tf_networks.Network

Bases: object

Convenience class to easily swap networks

define()

number_of_parameters()

placeholder()

resize_batch(b)

resize_output(imgs, i)

pyrieef.learning.train_supervised_cost_net module

pyrieef.learning.visualize_data module

Module contents