class App:
def __init__(self, settings):
### NAME OF SAVE ###
self.savename = "NN"
self.simulation = False
self.settings = settings
### INIT WINDOW ###
self.window = pyglet.window.Window(fullscreen=True, resizable=True)
self.window.set_caption("NEURAL NETWORK RACING by Tomas Brezina")
if not self.window.fullscreen: self.window.set_size(settings["WIDTH"], settings["HEIGHT"])
### LOAD ICON ###
try:
icon = pyglet.image.load("graphics/icon.ico")
self.window.set_icon(icon)
except:
print("Error >>> Loading icon")
### MODULES ###
self.simulation = None
self.graphics = Graphics(self.window)
### LABELS ###
self.graphics.labels["name"].text = self.savename
### VARIABLES ###
self.show = False # show track, cps, etc.
self.pause = False # pause the simulation
self.timer = 0 # number of ticks
self.timer_limit = self.settings["timeout_seconds"] // self.settings["render_timestep"] # max ticks
### BIND EVENTS ###
self.window.event(self.on_key_release)
self.window.event(self.on_close)
self.window.event(self.on_resize)
self.window.event(self.on_draw)
# when key is released
def on_key_release(self,symbol, modifiers):
# save the nn
if symbol == key.S:
if self.savename and self.simulation.best_nn:
_save_stg = self.simulation.save_stg
_save_stg["generations"] = self.simulation.gen_count
_save_stg["best_result"] = self.simulation.max_score
save_neural_network(
name=self.savename,
weigts=self.simulation.best_nn.weights,
settings=self.simulation.save_stg,
folder="saves"
)
# fullscreen on/off
if symbol == key.F:
self.window.set_fullscreen(not self.window.fullscreen)
if not self.window.fullscreen: self.window.set_size(self.settings["WIDTH"], self.settings["HEIGHT"])
# pause on/off
if symbol == key.P:
self.pause = not self.pause
# show on/off
if symbol == key.O:
self.show = not self.show
# when closed (unnecessary)
def on_close(self):
pyglet.clock.unschedule(self.update)
# when resized
def on_resize(self, width, height):
_scale = width / self.settings["WIDTH"]
self.graphics.change_scale(_scale)
self.simulation.track.change_scale(_scale)
# every frame
def on_draw(self):
self.graphics.clear()
if self.simulation.track.bg: self.graphics.draw_bg(self.simulation.track.bg)
self.graphics.car_batch.draw()
self.graphics.draw_labels()
# draw hidden details
if self.show:
for car in self.simulation.cars:
self.graphics.draw_car_sensors(car)
# draw checkpoints
self.graphics.draw_cps(self.simulation.track.cps_dict)
# draw edge of the track
self.graphics.draw_lines(self.simulation.track.vertex_list)
else:
if not self.simulation.track.bg:
self.graphics.draw_lines(self.simulation.track.vertex_list)
# create new generation from best nns
def new_generation(self):
self.graphics.labels["gen"].text = "Generation: " + str(int(self.simulation.gen_count))
self.graphics.clear_batch()
self.simulation.new_generation(
best_nns=self.simulation.get_best_nns(),
population=self.settings["population"],
mutation_rate=self.settings["mutation_rate"],
images=self.graphics.car_images,
batch=self.graphics.car_batch
)
self.graphics.update_sprites(self.simulation.cars)
self.graphics.labels["max"].text = "Best score: " + str(self.simulation.max_score)
# every frame
def update(self,dt):
if not self.pause:
# car behaviour
active = self.simulation.behave(dt)
if not active:
self.timer = 0
self.new_generation()
self.simulation.update(dt)
# update sprites position and rotation
self.graphics.update_sprites(self.simulation.cars)
self.timelimit()
# each tick
def timelimit(self):
self.timer += 1
if self.timer >= self.timer_limit:
self.timer = 0
self.new_generation()
seconds = int(self.timer * self.settings["render_timestep"])
self.graphics.labels["time"].text = "Time: " + str(seconds) + " / " + str(self.settings["timeout_seconds"])
# start of simulation
def start_simulation(self, track, nn_stg, nn_weights=False, name="New NN"):
# init simulation
self.savename = name
self.simulation = Simulation(track, nn_stg)
self.simulation.friction = self.settings["friction"]
# set labels
self.graphics.labels["name"].text = self.savename[:10]
self.graphics.labels["gen"].text = "Generation: " + str(int(self.simulation.gen_count))
self.graphics.labels["max"].text = "Best score: " + str(self.simulation.max_score)
# new save or loaded one
if nn_weights == False:
self.simulation.first_generation(
shape=self.simulation.save_stg["SHAPE"],
population=self.settings["population"],
mutation_rate=self.settings["mutation_rate"],
images=self.graphics.car_images,
batch=self.graphics.car_batch
)
else:
nn = NeuralNetwork(self.simulation.save_stg["SHAPE"])
nn.set_weights(nn_weights)
self.simulation.load_generation(
nn=nn,
population=self.settings["population"],
mutation_rate=self.settings["mutation_rate"],
images=self.graphics.car_images,
batch=self.graphics.car_batch
)
self.graphics.update_sprites(self.simulation.cars)
self.on_resize(self.window.width, self.window.height)
pyglet.clock.schedule_interval(self.update, self.settings["render_timestep"])
pyglet.app.run()
def end_simulation(self):
pyglet.clock.unschedule(self.update)
statistics = self.simulation.statistics
self.simulation = False
return statistics
# end
def exit(self):
pyglet.app.exit()
