def run_simulation(self, environments_copy):
weights = self.prepare_weights_for_ann()
ann = CTRAnn(weights=weights, hidden_layers=self.hidden_layers, activation_functions=self.activation_functions,
gains=self.get_gains(), time_constants=self.get_time_constants())
for j in range(len(environments_copy)):
for _ in range(TIMESTEPS):
environments_copy[j].drop_object_one_level()
agent_sensor_output = environments_copy[j].agent.get_sensor_array(environments_copy[j])
ann_inputs = agent_sensor_output
prediction = ann.predict(inputs=ann_inputs)
environments_copy[j].prediction_to_maneuver(prediction=prediction)
def __init__(self, phenotype, environments, fitness_log_average,
fitness_log_best, standard_deviation_log):
tk.Tk.__init__(self)
self.delay = INITIAL_DELAY
self.phenotype = phenotype
self.environments = environments
self.ann = CTRAnn(weights=phenotype.prepare_weights_for_ann(),
hidden_layers=phenotype.hidden_layers,
activation_functions=phenotype.activation_functions,
gains=phenotype.get_gains(),
time_constants=phenotype.get_time_constants())
self.current_timestep = 0
self.max_timestep = TIMESTEPS
self.standard_deviation_log = standard_deviation_log
self.fitness_log_best = fitness_log_best
self.fitness_log_average = fitness_log_average
self.title("Beer Tracker")
self.grid = {}
self.score_texts = []
self.pull_score_texts = []
self.sensor_texts = []
self.fitness_texts = []
self.step_text = None
self.beer_components = []
self.agent_components = []
self.cell_size = (SCREEN_WIDTH - (max(2, len(environments) + 1)) * GRID_OFFSET) / \
(WORLD_WIDTH * max(2, len(environments)))
self.canvas = tk.Canvas(self,
width=SCREEN_WIDTH,
height=(self.cell_size + 8) *
self.environments[0].height,
background='white',
borderwidth=0)
self.canvas.pack(side="top", fill="both", expand="true")
self.pause = True
self.bind('<space>', self.toggle_pause)
self.bind('<n>', self.decrease_simulation_speed)
self.bind('<m>', self.increase_simulation_speed)
# self.bind('<a>', self.move_agent_left)
# self.bind('<d>', self.move_agent_right)
self.reset_button = tk.Button(
self,
text="Reset board",
command=self.reset_gui_with_new_environment).pack()
self.draw_text()
self.draw_board()
self.update_text()
self.draw_agents()
self.draw_beer_objects()
self.run_simulation()
def run_simulation(self, environments_copy):
weights = self.prepare_weights_for_ann()
ann = CTRAnn(weights=weights,
hidden_layers=self.hidden_layers,
activation_functions=self.activation_functions,
gains=self.get_gains(),
time_constants=self.get_time_constants())
for j in range(len(environments_copy)):
for _ in range(TIMESTEPS):
environments_copy[j].drop_object_one_level()
agent_sensor_output = environments_copy[
j].agent.get_sensor_array(environments_copy[j])
ann_inputs = agent_sensor_output
prediction = ann.predict(inputs=ann_inputs)
environments_copy[j].prediction_to_maneuver(
prediction=prediction)
def __init__(self, phenotype, environments,
fitness_log_average, fitness_log_best, standard_deviation_log):
tk.Tk.__init__(self)
self.delay = INITIAL_DELAY
self.phenotype = phenotype
self.environments = environments
self.ann = CTRAnn(weights=phenotype.prepare_weights_for_ann(), hidden_layers=phenotype.hidden_layers,
activation_functions=phenotype.activation_functions,
gains=phenotype.get_gains(), time_constants=phenotype.get_time_constants())
self.current_timestep = 0
self.max_timestep = TIMESTEPS
self.standard_deviation_log = standard_deviation_log
self.fitness_log_best = fitness_log_best
self.fitness_log_average = fitness_log_average
self.title("Beer Tracker")
self.grid = {}
self.score_texts = []
self.pull_score_texts = []
self.sensor_texts = []
self.fitness_texts = []
self.step_text = None
self.beer_components = []
self.agent_components = []
self.cell_size = (SCREEN_WIDTH - (max(2, len(environments) + 1)) * GRID_OFFSET) / \
(WORLD_WIDTH * max(2, len(environments)))
self.canvas = tk.Canvas(self, width=SCREEN_WIDTH, height=(self.cell_size + 8)*self.environments[0].height,
background='white', borderwidth=0)
self.canvas.pack(side="top", fill="both", expand="true")
self.pause = True
self.bind('<space>', self.toggle_pause)
self.bind('<n>', self.decrease_simulation_speed)
self.bind('<m>', self.increase_simulation_speed)
# self.bind('<a>', self.move_agent_left)
# self.bind('<d>', self.move_agent_right)
self.reset_button = tk.Button(self, text="Reset board", command=self.reset_gui_with_new_environment).pack()
self.draw_text()
self.draw_board()
self.update_text()
self.draw_agents()
self.draw_beer_objects()
self.run_simulation()
class BeerTrackerGui(tk.Tk):
def __init__(self, phenotype, environments,
fitness_log_average, fitness_log_best, standard_deviation_log):
tk.Tk.__init__(self)
self.delay = INITIAL_DELAY
self.phenotype = phenotype
self.environments = environments
self.ann = CTRAnn(weights=phenotype.prepare_weights_for_ann(), hidden_layers=phenotype.hidden_layers,
activation_functions=phenotype.activation_functions,
gains=phenotype.get_gains(), time_constants=phenotype.get_time_constants())
self.current_timestep = 0
self.max_timestep = TIMESTEPS
self.standard_deviation_log = standard_deviation_log
self.fitness_log_best = fitness_log_best
self.fitness_log_average = fitness_log_average
self.title("Beer Tracker")
self.grid = {}
self.score_texts = []
self.pull_score_texts = []
self.sensor_texts = []
self.fitness_texts = []
self.step_text = None
self.beer_components = []
self.agent_components = []
self.cell_size = (SCREEN_WIDTH - (max(2, len(environments) + 1)) * GRID_OFFSET) / \
(WORLD_WIDTH * max(2, len(environments)))
self.canvas = tk.Canvas(self, width=SCREEN_WIDTH, height=(self.cell_size + 8)*self.environments[0].height,
background='white', borderwidth=0)
self.canvas.pack(side="top", fill="both", expand="true")
self.pause = True
self.bind('<space>', self.toggle_pause)
self.bind('<n>', self.decrease_simulation_speed)
self.bind('<m>', self.increase_simulation_speed)
# self.bind('<a>', self.move_agent_left)
# self.bind('<d>', self.move_agent_right)
self.reset_button = tk.Button(self, text="Reset board", command=self.reset_gui_with_new_environment).pack()
self.draw_text()
self.draw_board()
self.update_text()
self.draw_agents()
self.draw_beer_objects()
self.run_simulation()
def toggle_pause(self, event=None):
self.pause = not self.pause
def increase_simulation_speed(self, event=None):
self.delay = max(self.delay - 10, 1)
def decrease_simulation_speed(self, event=None):
self.delay += 10
def move_agent_left(self, event=None):
self.environments[0].agent.move_left()
def move_agent_right(self, event=None):
self.environments[0].agent.move_right()
def draw_board(self):
for i in range(len(self.environments)):
agent = self.environments[i].agent
beer_object = self.environments[i].beer_object
offset = (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET))
for y in range(WORLD_HEIGHT):
for x in range(WORLD_WIDTH):
self.grid[i, x, y] = self.canvas.create_rectangle(
x * self.cell_size + GRID_OFFSET + offset,
y * self.cell_size + GRID_OFFSET,
(x + 1) * self.cell_size + GRID_OFFSET + offset,
(y + 1) * self.cell_size + GRID_OFFSET)
def draw_beer_objects(self):
for k in range(len(self.beer_components)):
for l in range(len(self.beer_components[k])):
self.canvas.delete(self.beer_components[k][l])
self.beer_components = []
for i in range(len(self.environments)):
beer_object = self.environments[i].beer_object
offset = (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET))
beer_components = []
if beer_object.size > BEER_MAX_WANTED_SIZE:
color = "red"
else:
color = "blue"
for x in beer_object.range:
beer_components.append(self.canvas.create_oval(
x * self.cell_size + GRID_OFFSET + offset,
beer_object.y * self.cell_size + GRID_OFFSET,
(x + 1) * self.cell_size + GRID_OFFSET + offset,
(beer_object.y + 1) * self.cell_size + GRID_OFFSET,
fill=color))
self.beer_components.append(beer_components)
def draw_text(self):
for i in range(len(self.environments)):
self.score_texts.append(self.canvas.create_text(
GRID_OFFSET + (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.pull_score_texts.append(self.canvas.create_text(
GRID_OFFSET + (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
2 * GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.sensor_texts.append(self.canvas.create_text(
GRID_OFFSET + (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
3 * GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.fitness_texts.append(self.canvas.create_text(
GRID_OFFSET + (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
4 * GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.step_text = self.canvas.create_text(GRID_OFFSET, GRID_OFFSET/2)
def update_text(self):
for i in range(len(self.environments)):
self.canvas.itemconfig(self.score_texts[i], text="Score:" + str(self.environments[i].score))
self.canvas.itemconfig(self.pull_score_texts[i], text="Pull Score:" + str(self.environments[i].pull_score))
self.canvas.itemconfig(self.sensor_texts[i], text="Sensor:" + str(self.environments[i].agent.get_sensor_array(self.environments[i])))
self.canvas.itemconfig(self.step_text, text="Step: " + str(self.current_timestep + 1))
def draw_agents(self):
self.clean_agents()
for i in range(len(self.environments)):
agent = self.environments[i].agent
offset = (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET))
agent_components = []
color = "green"
if self.environments[i].pulling:
color = "yellow"
self.environments[i].pulling = False
for x in agent.range:
agent_components.append(self.canvas.create_oval(
x * self.cell_size + GRID_OFFSET + offset,
(WORLD_HEIGHT - 1) * self.cell_size + GRID_OFFSET,
(x + 1) * self.cell_size + GRID_OFFSET + offset,
(WORLD_HEIGHT) * self.cell_size + GRID_OFFSET,
fill=color))
self.agent_components.append(agent_components)
def clean_agents(self):
for k in range(len(self.agent_components)):
for l in range(len(self.agent_components[k])):
self.canvas.delete(self.agent_components[k][l])
self.agent_components = []
def start_simulation(self):
self.current_timestep = 0
self.run_simulation()
def run_simulation(self):
if not self.pause:
if self.current_timestep < self.max_timestep:
for i in range(len(self.environments)):
self.environments[i].drop_object_one_level()
agent_sensor_output = self.environments[i].agent.get_sensor_array(self.environments[i])
ann_inputs = agent_sensor_output
prediction = self.ann.predict(inputs=ann_inputs)
self.environments[i].prediction_to_maneuver(prediction=prediction)
self.draw_agents()
self.draw_beer_objects()
self.update_text()
self.current_timestep += 1
else:
print "Simulation over"
print self.ann.weights
PhenotypeBeerTracker.environments_for_fitness = self.environments
for l in range(len(self.environments)):
PhenotypeBeerTracker.environments_for_fitness[l].reset()
fitness = PhenotypeBeerTracker.fitness_evaluation(self.phenotype)
self.canvas.itemconfig(self.fitness_texts[l], text="Fitness:" + str(round(fitness[0], 3)) + " "+ str(fitness[1][l]))
#self.plot_data(self.fitness_log_average, self.fitness_log_best, self.standard_deviation_log)
return
self.after(self.delay, lambda: self.run_simulation())
def reset_gui_with_new_environment(self, event=None):
self.environments = [BeerTrackerWorld(width=self.environments[0].width,
height=self.environments[0].height,
agent_type=self.environments[0].agent.agent_type)
for _ in range(len(self.environments))]
self.pause = True
for key, val in self.grid.items():
self.canvas.delete(val)
self.draw_board()
self.draw_agents()
self.draw_beer_objects()
self.update_text()
self.start_simulation()
@staticmethod
def plot_data(fitness_log_average, fitness_log_best, standard_deviation_log):
plt.figure(1)
plt.subplot(211)
plt.plot(fitness_log_average[-1], label="Average fitness")
plt.plot(fitness_log_best[-1], label="Best fitness")
plt.legend(bbox_to_anchor=(0., 1.02, 1., .102), loc=3, ncol=2, mode="expand", borderaxespad=0.)
plt.subplot(212)
plt.plot(standard_deviation_log[-1], label="Standard deviation")
#plt.legend(bbox_to_anchor=(0., 1.02, 1., .102), loc=3, ncol=2, mode="expand", borderaxespad=0.)
plt.show()
class BeerTrackerGui(tk.Tk):
def __init__(self, phenotype, environments, fitness_log_average,
fitness_log_best, standard_deviation_log):
tk.Tk.__init__(self)
self.delay = INITIAL_DELAY
self.phenotype = phenotype
self.environments = environments
self.ann = CTRAnn(weights=phenotype.prepare_weights_for_ann(),
hidden_layers=phenotype.hidden_layers,
activation_functions=phenotype.activation_functions,
gains=phenotype.get_gains(),
time_constants=phenotype.get_time_constants())
self.current_timestep = 0
self.max_timestep = TIMESTEPS
self.standard_deviation_log = standard_deviation_log
self.fitness_log_best = fitness_log_best
self.fitness_log_average = fitness_log_average
self.title("Beer Tracker")
self.grid = {}
self.score_texts = []
self.pull_score_texts = []
self.sensor_texts = []
self.fitness_texts = []
self.step_text = None
self.beer_components = []
self.agent_components = []
self.cell_size = (SCREEN_WIDTH - (max(2, len(environments) + 1)) * GRID_OFFSET) / \
(WORLD_WIDTH * max(2, len(environments)))
self.canvas = tk.Canvas(self,
width=SCREEN_WIDTH,
height=(self.cell_size + 8) *
self.environments[0].height,
background='white',
borderwidth=0)
self.canvas.pack(side="top", fill="both", expand="true")
self.pause = True
self.bind('<space>', self.toggle_pause)
self.bind('<n>', self.decrease_simulation_speed)
self.bind('<m>', self.increase_simulation_speed)
# self.bind('<a>', self.move_agent_left)
# self.bind('<d>', self.move_agent_right)
self.reset_button = tk.Button(
self,
text="Reset board",
command=self.reset_gui_with_new_environment).pack()
self.draw_text()
self.draw_board()
self.update_text()
self.draw_agents()
self.draw_beer_objects()
self.run_simulation()
def toggle_pause(self, event=None):
self.pause = not self.pause
def increase_simulation_speed(self, event=None):
self.delay = max(self.delay - 10, 1)
def decrease_simulation_speed(self, event=None):
self.delay += 10
def move_agent_left(self, event=None):
self.environments[0].agent.move_left()
def move_agent_right(self, event=None):
self.environments[0].agent.move_right()
def draw_board(self):
for i in range(len(self.environments)):
agent = self.environments[i].agent
beer_object = self.environments[i].beer_object
offset = (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET))
for y in range(WORLD_HEIGHT):
for x in range(WORLD_WIDTH):
self.grid[i, x, y] = self.canvas.create_rectangle(
x * self.cell_size + GRID_OFFSET + offset,
y * self.cell_size + GRID_OFFSET,
(x + 1) * self.cell_size + GRID_OFFSET + offset,
(y + 1) * self.cell_size + GRID_OFFSET)
def draw_beer_objects(self):
for k in range(len(self.beer_components)):
for l in range(len(self.beer_components[k])):
self.canvas.delete(self.beer_components[k][l])
self.beer_components = []
for i in range(len(self.environments)):
beer_object = self.environments[i].beer_object
offset = (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET))
beer_components = []
if beer_object.size > BEER_MAX_WANTED_SIZE:
color = "red"
else:
color = "blue"
for x in beer_object.range:
beer_components.append(
self.canvas.create_oval(
x * self.cell_size + GRID_OFFSET + offset,
beer_object.y * self.cell_size + GRID_OFFSET,
(x + 1) * self.cell_size + GRID_OFFSET + offset,
(beer_object.y + 1) * self.cell_size + GRID_OFFSET,
fill=color))
self.beer_components.append(beer_components)
def draw_text(self):
for i in range(len(self.environments)):
self.score_texts.append(
self.canvas.create_text(
GRID_OFFSET +
(i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.pull_score_texts.append(
self.canvas.create_text(
GRID_OFFSET +
(i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
2 * GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.sensor_texts.append(
self.canvas.create_text(
GRID_OFFSET +
(i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
3 * GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.fitness_texts.append(
self.canvas.create_text(
GRID_OFFSET +
(i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET)),
4 * GRID_OFFSET + WORLD_HEIGHT * self.cell_size,
anchor=tk.NW))
self.step_text = self.canvas.create_text(GRID_OFFSET, GRID_OFFSET / 2)
def update_text(self):
for i in range(len(self.environments)):
self.canvas.itemconfig(self.score_texts[i],
text="Score:" +
str(self.environments[i].score))
self.canvas.itemconfig(self.pull_score_texts[i],
text="Pull Score:" +
str(self.environments[i].pull_score))
self.canvas.itemconfig(
self.sensor_texts[i],
text="Sensor:" +
str(self.environments[i].agent.get_sensor_array(
self.environments[i])))
self.canvas.itemconfig(self.step_text,
text="Step: " + str(self.current_timestep + 1))
def draw_agents(self):
self.clean_agents()
for i in range(len(self.environments)):
agent = self.environments[i].agent
offset = (i * (WORLD_WIDTH * self.cell_size + GRID_OFFSET))
agent_components = []
color = "green"
if self.environments[i].pulling:
color = "yellow"
self.environments[i].pulling = False
for x in agent.range:
agent_components.append(
self.canvas.create_oval(
x * self.cell_size + GRID_OFFSET + offset,
(WORLD_HEIGHT - 1) * self.cell_size + GRID_OFFSET,
(x + 1) * self.cell_size + GRID_OFFSET + offset,
(WORLD_HEIGHT) * self.cell_size + GRID_OFFSET,
fill=color))
self.agent_components.append(agent_components)
def clean_agents(self):
for k in range(len(self.agent_components)):
for l in range(len(self.agent_components[k])):
self.canvas.delete(self.agent_components[k][l])
self.agent_components = []
def start_simulation(self):
self.current_timestep = 0
self.run_simulation()
def run_simulation(self):
if not self.pause:
if self.current_timestep < self.max_timestep:
for i in range(len(self.environments)):
self.environments[i].drop_object_one_level()
agent_sensor_output = self.environments[
i].agent.get_sensor_array(self.environments[i])
ann_inputs = agent_sensor_output
prediction = self.ann.predict(inputs=ann_inputs)
self.environments[i].prediction_to_maneuver(
prediction=prediction)
self.draw_agents()
self.draw_beer_objects()
self.update_text()
self.current_timestep += 1
else:
print "Simulation over"
print self.ann.weights
PhenotypeBeerTracker.environments_for_fitness = self.environments
for l in range(len(self.environments)):
PhenotypeBeerTracker.environments_for_fitness[l].reset()
fitness = PhenotypeBeerTracker.fitness_evaluation(
self.phenotype)
self.canvas.itemconfig(self.fitness_texts[l],
text="Fitness:" +
str(round(fitness[0], 3)) + " " +
str(fitness[1][l]))
#self.plot_data(self.fitness_log_average, self.fitness_log_best, self.standard_deviation_log)
return
self.after(self.delay, lambda: self.run_simulation())
def reset_gui_with_new_environment(self, event=None):
self.environments = [
BeerTrackerWorld(width=self.environments[0].width,
height=self.environments[0].height,
agent_type=self.environments[0].agent.agent_type)
for _ in range(len(self.environments))
]
self.pause = True
for key, val in self.grid.items():
self.canvas.delete(val)
self.draw_board()
self.draw_agents()
self.draw_beer_objects()
self.update_text()
self.start_simulation()
@staticmethod
def plot_data(fitness_log_average, fitness_log_best,
standard_deviation_log):
plt.figure(1)
plt.subplot(211)
plt.plot(fitness_log_average[-1], label="Average fitness")
plt.plot(fitness_log_best[-1], label="Best fitness")
plt.legend(bbox_to_anchor=(0., 1.02, 1., .102),
loc=3,
ncol=2,
mode="expand",
borderaxespad=0.)
plt.subplot(212)
plt.plot(standard_deviation_log[-1], label="Standard deviation")
#plt.legend(bbox_to_anchor=(0., 1.02, 1., .102), loc=3, ncol=2, mode="expand", borderaxespad=0.)
plt.show()