def reset(self, isTest, foodGet, temperature, display=False):
"""
Reset the environment for the training
"""
# if it is a test phase, we put the food get by the agent at the end into the list result
if (isTest):
foodGet.append(self.totalFood)
self.Temp = 0
else:
self.Temp = temperature
self.agent = agt.Agent(12, 18, 40)
self.ennemies = [
Ennemy(6, 6, self),
Ennemy(12, 2, self),
Ennemy(12, 6, self),
Ennemy(18, 6, self)
]
self.environment = [[0 if j == 2 else j for j in i]
for i in self.environment]
# we reset the location of the food
for i in range(15):
x = random.randint(0, 24)
y = random.randint(0, 24)
while not self.is_empty(x, y):
x = random.randint(0, 24)
y = random.randint(0, 24)
self.environment[x][y] = 2
self.end = False
self.killed = False
self.dead = False
self.totalFood = 0 #food eaten during the simulation
self.memory = [] # we reset the list of experiences
def testOPatch(self):
testgrid = env.State([], nn)
for i in range(env.width):
for j in range(env.heigth):
testgrid.environment[i][j] = 0
testgrid.environment[10][10] = 2
testgrid.ennemies = ([
Ennemy(5, 5, testgrid.environment),
Ennemy(-1, 0, testgrid.environment)
])
for i in range(-1, env.width + 1, 1):
for j in range(-1, env.heigth + 1, 1):
if (i == 5 and j == 5) or (i == 4 and j == 5) or (
i == 5 and j == 4) or (i == 6 and j == 5) or (
i == 5 and j == 6) or (i == 4 and j == 4) or (
i == 4 and j == 6) or (i == 6 and j == 6) or (
i == 6 and j == 4):
self.assertTrue(testgrid.Opatch(1, i, j))
self.assertFalse(testgrid.Opatch(2, i, j))
elif (i == 10 and j == 10) or (i == 9 and j == 10) or (
i == 10 and j == 9) or (i == 11 and j == 10) or (
i == 10 and j == 11) or (i == 9 and j == 9) or (
i == 9
and j == 11) or (i == 11
and j == 9) or (i == 11
and j == 11):
self.assertTrue(testgrid.Opatch(2, i, j))
self.assertFalse(testgrid.Opatch(1, i, j))
else:
self.assertFalse(testgrid.Opatch(1, i, j))
self.assertFalse(testgrid.Opatch(2, i, j))
def testEnnemyLookUp(self):
testgrid = env.State([], nn)
testgrid.ennemies = ([
Ennemy(5, 5, testgrid.environment),
Ennemy(-1, -1, testgrid.environment)
])
for i in range(-1, env.width + 1, 1):
for j in range(-1, env.heigth + 1, 1):
if i == 5 and j == 5:
self.assertTrue(testgrid.lookupEnnemies(i, j))
else:
self.assertFalse(testgrid.lookupEnnemies(i, j))
def main():
# initialize game engine
pygame.init()
# Setup main window
window_width = 400
window_height = 300
screen = pygame.display.set_mode((window_width, window_height))
pygame.display.set_caption("YARL!")
# Initialize Hero
floor = Floor("floors/dungeon.json")
hero = Hero()
print(hero.name + " just entered " + floor.name)
hero.loc_x = floor.spawn[0]
hero.loc_y = floor.spawn[1]
set_position(hero, floor)
# Initialize ennemies
nb_ennemies = 1
ennemies = []
for ennemy in range(nb_ennemies):
ennemies.append(Ennemy())
loc = get_ennemy_spawn(floor, ennemies, hero)
ennemies[-1].loc_x = loc[0]
ennemies[-1].loc_y = loc[1]
set_position(ennemies[-1], floor)
# Example of event handling
while True:
for event in pygame.event.get():
if event.type == QUIT:
pygame.quit()
sys.exit()
elif event.type == pygame.KEYDOWN:
next_loc = [hero.loc_x, hero.loc_y]
if event.key == pygame.K_LEFT:
next_loc[0] -= 1
if event.key == pygame.K_RIGHT:
next_loc[0] += 1
if event.key == pygame.K_UP:
next_loc[1] -= 1
if event.key == pygame.K_DOWN:
next_loc[1] += 1
set_loc(floor, hero, ennemies, next_loc)
set_position(hero, floor)
# Get all items to blit
blits = floor.get_blits()
blits += ui.get_blits()
blits += hero.get_blits()
for ennemy in ennemies:
blits += ennemy.get_blits()
# Blit and update display at every frame
for item in blits:
screen.blit(item[0], item[1])
pygame.display.update()
def generateEnnemies(self):
En = []
for dic in self._waves:
for key in dic.keys():
nbToCreate = dic[key]
while nbToCreate > 0:
En.append(
Ennemy(
100, self.sponePoint[0] * int(self.caseW) +
int(self.caseW / 2),
self.sponePoint[1] * int(self.caseH) +
int(self.caseH / 2), int(key),
(self.caseW, self.caseH)))
nbToCreate -= 1
self.ennemies = En
def __init__(self,
obstacles,
nn,
temperature=1 / 60,
display=False,
isTest=False,
interval=[1 / 20, 1 / 60]):
self.environment = [[0 for j in range(25)] for i in range(25)]
#0=empty, 1=obstacle, 2= food (ennemy and agent stored separetely)
for (x, y) in obstacles:
self.environment[x][y] = 1
self.ennemies = [
Ennemy(6, 6, self),
Ennemy(12, 2, self),
Ennemy(12, 6, self),
Ennemy(18, 6, self)
]
self.agent = agt.Agent(12, 18, 40)
for i in range(15):
x = random.randint(0, 24)
y = random.randint(0, 24)
while not self.is_empty(x, y):
x = random.randint(0, 24)
y = random.randint(0, 24)
self.environment[x][y] = 2
self.energy = 40
self.end = False
self.nn = nn # the neural network used for the learning
self.gamma = 0.9 # the parameters for the learning
self.Temp = temperature # The temperature for the stochastic action selector
if isTest:
self.Temp = 0
self.var_T = interval # interval that define the max and min value the temperature could take during learning
self.Ulist = [
] # list of Utility for all the actions at each state use for the learning
self.totalFood = 0 #food eaten during the simulation
self.count_without_food = 0 # a count to compute when the robot is stuck in a area without getting food
self.grille = None
self.can = None
self.agentText = None
self.PAS = None
self.can_life = None
self.life = None
self.ennemyText = None
self.killed = False
self.dead = False
self.action_proba = [] # the list of probabilities for each action
self.memory = [
] # the list that will record previous experiments for the replay buffer()
self.lessons = [] # lessons for the replay buffer that will obtain
self.display = display
if self.display:
self.initiate_simulation()