generate_simulation_data.GenerateSimulationData¶check_dataset_conformity(runs_dir, runs_img, title, dataset, net_input, communication)¶Generate a scatter plot to check the conformity of the dataset. The plot will show the distribution of the input sensing, in particular, as the difference between the front sensor and the mean of the rear sensors, with respect to the output control of the datasets.
runs_dir – directory containing the simulation
runs_img – directory containing the simulation images
title – title of the plot
dataset – name of the dataset
net_input – input of the net between prox_values, prox_comm or all_sensors
communication – states if the communication is used by the network
generate_dict(myt, n_sim, s, comm=None)¶Save all the information of the agent in a dictionary.
myt – agent
n_sim – index of the simulation
s – index of the timestep
comm – boolean, True if the dataset is generated using the learned model with communication
‘run’: n_sim,
‘timestep’: s,
‘name’: myt.name,
‘index’: myt.index,
‘myt_quantity’: myt.controller.N,
‘net_input’: myt.controller.net_input,
‘prox_values’: prox_values,
‘prox_comm’: utils.parse_prox_comm(prox_comm),
‘all_sensors’: all_sensors,
‘initial_position’: myt.initial_position,
‘position’: myt.position,
‘angle’: myt.angle,
‘goal_position’: myt.goal_position,
‘goal_angle’: myt.goal_angle,
‘motor_left_target’: myt.motor_left_target,
‘motor_right_target’: myt.motor_right_target,
‘goal_position_distance’: myt.goal_position[0] - myt.position[0],
‘goal_position_distance_absolute’: abs(myt.goal_position[0] - myt.position[0])
‘transmitted_comm’: communication transmitted two the neighbours
‘colour’: colour of the top led
generate_simulation(run_dir, n_simulations, controller, myt_quantity, args, model_dir=None, model=None, communication=False)¶run_dir – directory containing the simulation runs
n_simulations – number of simulations to generate
controller – controller to be used in the simulations
model_dir – directory containing the model
myt_quantity – number of agents
args – arguments from command line
model – model to be used
communication – states if the communication is used by the network
get_controller(controller, controllers, goal, myt_quantity, net_input, communication=None, model=None, model_dir=None)¶Important
This method has not yet been implemented for task 2.
controller – name of the controller (between omniscient, manual and learned)
controllers – reference to the controller class
goal – task to perform (between colour ot distribute)
myt_quantity – number of agents
net_input – input of the net between prox_values, prox_comm or all_sensors
communication – states if the communication is used by the network
model – network to be used by the controller
model_dir – directory containing the model
lambda function containing the controller to be used
init_positions(myts, net_input, avg_gap, variate_pose=False, min_distance=10.9, extension=False, x=None, epsilon=None)¶Given the agents, position them such as all x-axes are aligned. Arrange the robots in a “single file” (all x-axes aligned) and within the proximity sensor range (when running the extension of task 1 it is not mandatory that the robot are within the proximity sensor range, since a random average gap is used).
The distance between the first and the last robot is usually computed as follow:
In case of the extended task this quantity it is not initialised.
Usually, the distances among the robots are computed by drawing (myt_quantity - 1) real random gaps in using a normal
distribution with mean equal to the average gap and stdev fixed to 8. This values are clipped between 1 and the
maximum_gap. Then, the minimum_distance is added to all the distances, in order to move from the ICR
to the front of the thymio. The distances obtained are rescaled in such a way their sum corresponds to the
total gap that is known.
In case of the extended task the distances are obtained using the random average gap fixed and generating
random positions in the maximum range allowed.
myts – list containing all the agents and associated attributes
net_input – input of the net between prox_values, prox_comm or all_sensors
avg_gap – for the prox_values the default value is 8cm; for the prox_comm the default value is 25cm. Can be also a random value, between 5 and 35, in case of the extended task.
variate_pose – False by default, True only during the evaluation of the models when only one agent is used.
min_distance – the minimum distance between two Thymio [wheel - wheel] is 10.9 cm, that corresponds to the length of the agents.
extension – True if is running task1_extension
x – quantity of which variate the position along the x axis of the robot (only during the model evaluation with variate_pose True)
run(n_sim, myts, runs, complete_runs, world: pyenki.pyenki.World, comm=None, gui: bool = False, T: float = 4, dt: float = 0.1, tol: float = 0.1) → None¶Run the simulation as fast as possible or using the real time GUI. Generate two different type of simulation data, one with all the thymios and the other without including the 2 thymios at the ends, but only the ones that have to move. If all the robots have reached their target, stop the simulation.
n_sim – index of the simulation
myts – list containing the agents
runs – list containing all the data of the robots (except about the two robots at the beginning and at the end of the row) in the simulations
complete_runs – list containing all the data of the robots in the simulations
world – enki world
comm – variable that states if the communication is used
gui – variable that states if use the gui
T – duration of the simulation in seconds
dt – control step duration – update timestep in seconds, should be below 1 (typically .02-.1)
tol – tolerance used to verify if the robot reaches the target
save_simulation(complete_data, data, runs_dir)¶complete_data – list containing all the data of the robots in the simulations
data – list containing all the data of the robots (except about the two robots at the beginning and at the end of the row) in the simulations
runs_dir – directory where to save the simulation runs
setup(controller_factory, myt_quantity, aseba: bool = False)¶Set up the world as an unbounded world.
controller_factory – if the controller is passed, load the learned network
myt_quantity – number of robot in the simulation
aseba – True
world and the agents in the world
update_dict(myt, s, comm=None)¶Updated data in the dictionary instead of rewrite every field to optimise performances.
myt – agent
s – timestep
comm – boolean, True if the dataset is generated using the learned model with communication
updated dictionary
verify_target(myt, counter, dictionary, tol)¶Verify if the robot has reached the target goal.
It is always ensured that enough timesteps are performed, especially for the omniscient controller.
myt – agent
counter – counter incremented when the robot has reached the goal
dictionary – dictionary containing the agent information
tol – tolerance