Others

Plots

utils.my_plots.animate_simulation(out_dirs, myt_quantity)
Parameters

  • out_dirs – directory containing the simulation runs

  • myt_quantity – number of agents

utils.my_plots.get_position_distances(runs_sub, with_run=False)
Parameters

  • runs_sub – directory containing the simulation runs

  • with_run – boolean, default False

Return position_distances

array containing distances from goal

utils.my_plots.make_space_above(axes, topmargin=1)

Increase figure size to make topmargin (in inches) space for titles, without changing the axes sizes.

Parameters

  • axes – axes of the image

  • topmargin – topmargin

utils.my_plots.my_histogram(prediction, x_label, img_dir, title, filename, label=None)
Parameters

  • prediction – predictions

  • x_label – label for the x-axis

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

  • label – label for the plot, default None

utils.my_plots.my_scatterplot(x, y, x_label, y_label, img_dir, title, filename)

Plot a scatter plot. Usually with the groundtruth on x-axis and prediction on y-axis.

Parameters

  • x – values for the x-axis

  • y – values for the y-axis

  • x_label – label for the x-axis

  • y_label – label for the y-axis

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

utils.my_plots.plot_compared_distance_compressed(dataset_folders, img_dir, datasets, title, filename, absolute=True)

Warning

Limits on x and y axes have not yet been fixed. Their value depends if it is used goal_position_distance or goal_position_distance_absolute.

Parameters

  • dataset_folders – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • datasets – names of the datasets to be uses

  • title – title of the image

  • filename – name of the image

  • absolute – boolean value that states is use absolute distances from goal (default: True)

utils.my_plots.plot_compared_distance_from_goal(dataset_folders, img_dir, title, filename, absolute=True)
Parameters

  • dataset_folders – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

  • absolute – boolean value that states is use absolute distances from goal (default: True)

utils.my_plots.plot_distance_from_goal(runs_dir, img_dir, title, filename)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

utils.my_plots.plot_losses(train_loss, valid_loss, img_dir, title, filename, goal)
Parameters

  • train_loss – the training losses

  • valid_loss – the testing losses

  • img_dir – directory for the output image

  • title – title of the image

  • filename – name of the image

utils.my_plots.plot_regressor(x, y, x_label, y_label, img_dir, title, filename)
Parameters

  • x – values for the x-axis

  • y – values for the y-axis

  • x_label – label for the x-axis

  • y_label – label for the y-axis

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

utils.my_plots.plot_response(x, y, x_label, img_dir, title, filename, index=None, y_label='control')
Parameters

  • x – values for the x-axis

  • y – values for the y-axis

  • x_label – label for the x-axis

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

  • index – this parameter is different from None only when x is the input sensing, otherwise, x is a 1D vector

  • y_label – label for the y-axis (optional, default: ‘control’)

utils.my_plots.plot_sensing_timestep(runs_dir, img_dir, net_input, model)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • net_input – input of the net between prox_values, prox_comm or all_sensors

  • model – model to be used

utils.my_plots.plot_simulations(out_dirs, myt_quantity)
Parameters

  • out_dirs – directory containing the simulation runs

  • myt_quantity – number of agents

utils.my_plots.plot_target_distribution(y_g, y_p, img_dir, title, filename)
Parameters

  • y_g – validation groundtruth

  • y_p – validation prediction

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

utils.my_plots.save_visualisation(filename, img_dir, make_space=False, axes=None)
Parameters

  • filename – name of the image

  • img_dir – path where to save the image

  • make_space – if make space above the image

  • axes – axes of the image

utils.my_plots.test_controller_given_init_positions(model_img, model, net_input)
Parameters

  • model_img – directory for the output image of the model

  • model – name of the model

  • net_input – input of the network (between: prox_values, prox_comm and all_sensors)

utils.my_plots.thymio_quantity_distribution(runs_dir, img_dir, title, filename)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • title – title of the image

  • filename – name of the image

utils.my_plots.visualise_communication_simulation(runs_dir, img_dir, simulation, title)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • simulation – simulation to use

  • title – title of the image

utils.my_plots.visualise_communication_vs_control(runs_dir, img_dir, title)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • title – title of the image

utils.my_plots.visualise_communication_vs_distance(runs_dir, img_dir, title)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • title – title of the image

utils.my_plots.visualise_simulation(runs_dir, img_dir, simulation, title, net_input)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • simulation – simulation to use

  • title – title of the image

  • net_input – input of the net between prox_values, prox_comm or all_sensors

utils.my_plots.visualise_simulation_all_sensors(runs_dir, img_dir, simulation, title, net_input)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • simulation – simultion to use

  • title – title of the image

  • net_input – input of the net between prox_values, prox_comm or all_sensors

utils.my_plots.visualise_simulation_over_time_all_sensors(runs_dir, img_dir, simulation, title)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • simulation – simulation to use

  • title – title of the image

utils.my_plots.visualise_simulations_comparison(runs_dir, img_dir, title, net_input)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • title – title of the image

  • net_input – input of the net between prox_values, prox_comm or all_sensors

utils.my_plots.visualise_simulations_comparison_all_sensors(runs_dir, img_dir, title, net_input)
Parameters

  • runs_dir – directory containing the simulation runs

  • img_dir – directory containing the simulation images

  • title – title of the image

  • net_input – input of the net between prox_values, prox_comm or all_sensors

Utils

class utils.utils.ThymioState(state_dict)

Object containing all the agent information :param state_dict

utils.utils.cartesian_product(*arrays)
Parameters

arrays – arrays used to compute the cartesian product

Return arr.reshape(-1, la)

cartesian product

utils.utils.check_dir(directory)

Check if the path is a directory, if not create it.

Parameters

directory – path to the directory

utils.utils.dataset_split(file_name, num_run=1000)
Parameters

  • file_name – path to the file where to save the splits of the dataset

  • num_run – number of simulations, by default 1000

utils.utils.directory_for_dataset(dataset, controller)
Parameters

  • dataset – name of the dataset

  • controller – name of the controller

Return run_dir, run_img_dir, run_video_dir

output directories for the simulations

utils.utils.directory_for_model(args)
Parameters

args – command line arguments

Return model_dir, model_img_dir, model_video_dir, metrics_path

output directories for the models

utils.utils.export_network(model_dir, model, in_put, input_shape)
Parameters

  • model_dir – directory containing the model

  • model – name of the model

  • in_put – array of the same shape of the first network input

  • input_shape – array of the same shape of the second network input

utils.utils.extract_colour_output(runs, communication=False, input_combination=True)

The output is the colour of the top led that depends by the position of the robot in the row.

Parameters

  • runs – dataframe containing all the simulation runs

  • communication – states if the communication is used by the network

  • input_combination – states if using the input combination of the sensors, that means using only the central frontal sensor and the mean of the rear sensors

Return out_put, runs

output array for the network and dataframe with the runs

utils.utils.extract_input_output(runs, in_label, N, communication=False, input_combination=True, myt_quantities=None, task='distribute')

Whether the input is prox_values, prox_comm or all sensors, it corresponds to the response values of ​​the sensors [array of 7 floats]. The input is normalised so that the average is around 1 or a constant (e.g. for all (dividing by 1000)). The output is the speed of the wheels (which we assume equals left and right) [array of 1 float]. There is no need to normalize the outputs.

Parameters

  • runs – dataframe containing all the simulation runs

  • in_label – input of the net between prox_values, prox_comm or all_sensors

  • N – number of agents

  • communication – states if the communication is used by the network

  • input_combination – states if using the input combination of the sensors, that means using only the central frontal sensor and the mean of the rear sensors

  • myt_quantities – array containing the number agents for each simulation run

  • task – task to perform (can be distribute or colour)

Return in_put, out_put, out_myt_quantities, runs, columns

input and output arrays for the network, array with the number of agents, dataframe with the runs and columns of the dataframe referred to the input label

utils.utils.from_dataset_to_tensors(train_sample, train_target, valid_sample, valid_target, test_sample, test_target, q_train, q_valid, q_test)
Parameters

  • train_sample – training set samples

  • train_target – training set targets

  • valid_sample – validation set samples

  • valid_target – validation set targets

  • test_sample – testing set samples

  • test_target – testing set targets

  • q_train – mask containing the number of agents for each sample of the training set

  • q_valid – mask containing the number of agents for each sample of the validation set

  • q_test – mask containing the number of agents for each sample of the testing set

Return test, train, valid

test, train and valid TensorDataset

utils.utils.from_indices_to_dataset(runs_dir, train_indices, validation_indices, test_indices, net_input, communication=False, task='distribute')
Parameters

  • runs_dir – directory containing the simulations

  • train_indices – indices of the sample belonging to the training set

  • validation_indices – indices of the sample belonging to the validation set

  • test_indices – indices of the sample belonging to the testing set

  • net_input – input of the net between prox_values, prox_comm or all_sensors

  • communication – states if the communication is used by the network

  • task – task to perform (can be distribute or colour)

Returns

(train_sample, valid_sample, test_sample), train_target, valid_target, test_target, train_quantities, valid_quantities, test_quantities: all the train, validation and test samples, targets and masks

utils.utils.generate_fake_simulations(run_dir, model, initial_positions, myt_quantity)
Parameters

  • run_dir – directory containing the simulation runs

  • model – name of the model

  • initial_positions – initial position for the agents

  • myt_quantity – number of agents

Return out_dirs

directory containing the simulation run

utils.utils.get_all_sensors(prox_values, prox_comm)
Parameters

  • prox_values – prox_values reading

  • prox_comm – prox_comm reading

Return all_sensors

combination of the two sensor readings

utils.utils.get_input_columns(in_label)
Parameters

in_label – input of the net between prox_values, prox_comm or all_sensors

Return columns

columns of the dataframe referred to the input label

utils.utils.get_input_sensing(in_label, myt, normalise=True)
Parameters

  • in_label – input of the net between prox_values, prox_comm or all_sensors

  • myt – agent

  • normalise – states if normalise the input sensing (default: True)

Return sensing

sensing perceived by the agent

utils.utils.get_key_value_of_nested_dict(nested_dict)

Access a nested dictionary and return a list of tuples (rv) and values. Used to return the list of intensities given a prox_comm dictionary containing multiple senders.

Parameters

nested_dict – nested dictionary, usually containing prox_comm_events

Return rv, values

rv is a list of tuples where, in each of these, the first element is a list of keys and the second is the final value. Values is the list of inner values.

utils.utils.get_prox_comm(myt)

Create a dictionary containing all the senders as key and the corresponding intensities as value.

Parameters

myt – agent

Return prox_comm

prox_comm sensing

utils.utils.get_received_communication(myt, goal='distribute')

Create a list containing the messages received from the back and front.

Parameters

  • myt – agent

  • goal – goal of the task, by default distribute

Return communication

the communication received from left to right

utils.utils.get_transmitted_communication(myt)

Return the values transmitted during the communication.

Parameters

myt – agent

Return communication

the communication to be transmitted

utils.utils.load_dataset(runs_dir, dataset)
Parameters

  • runs_dir – directory containing the simulation runs

  • dataset – name of the dataset

Return dataframe

resulting dataframe

utils.utils.parse_prox_comm(prox_comm)
Parameters

prox_comm – prox_comm dictionary

Return prox_comm

parsed prox_comm list

utils.utils.prepare_dataset(run_dir, split, num_run)
Parameters

  • run_dir – directory containing the simulation runs

  • split – states if generate or load the split file

  • num_run – number of runs used in the simulation

Return file, indices

file containing the splits and the splits indices

utils.utils.signed_distance(state)
Parameters

state – object containing all the agent information

Return b - a

signed distance between current and the goal position, along the current theta of the robot

← Examples