class Environment:
def __init__(self,
wrap=False,
grid_size=10,
local_size=9,
rate=100,
max_time=math.inf,
action_space=5,
food_count=1,
stick_count=1,
obstacle_count=0,
lava_count=0,
zombie_count=0,
history=30,
map_path=None):
"""
Initialise the Game Environment with default values
"""
#self.FPS = 120 # NOT USED YET
self.UPDATE_RATE = rate
self.SCALE = 20 # Scale of each block 20x20 pixels
self.GRID_SIZE = grid_size
self.LOCAL_GRID_SIZE = local_size # Has to be an odd number
self.ENABLE_WRAP = wrap
if not self.ENABLE_WRAP:
self.GRID_SIZE += 2
self.ACTION_SPACE = action_space
self.MAP_PATH = map_path
self.DISPLAY_WIDTH = self.GRID_SIZE * self.SCALE
self.DISPLAY_HEIGHT = self.GRID_SIZE * self.SCALE
# Maximum timesteps before an episode is stopped
self.MAX_TIME_PER_EPISODE = max_time
# Create and Initialise steve
self.steve = Steve(history)
# Create Food
self.NUM_OF_FOOD = food_count
self.food = Food(self.NUM_OF_FOOD)
self.NUM_OF_STICKS = stick_count
self.stick = Stick(self.NUM_OF_STICKS)
# Create Obstacles
self.NUM_OF_OBSTACLES = obstacle_count
self.obstacle = Obstacle(self.NUM_OF_OBSTACLES)
# Create Lava
self.NUM_OF_LAVA = lava_count
self.lava = Lava(self.NUM_OF_LAVA)
# Create Zombies
self.NUM_OF_ZOMBIES = zombie_count
self.zombie = Zombie(self.NUM_OF_ZOMBIES)
self.score = 0
self.time = 0
self.state = None
self.display = None
self.bg = None
self.clock = None
self.font = None
self.steps = 0
self.spawn_new_food = False
self.last_eaten_food = -1
# Used putting food and obstacles on the grid
self.grid = []
for j in range(self.GRID_SIZE):
for i in range(self.GRID_SIZE):
self.grid.append((i * self.SCALE, j * self.SCALE))
self.maze = [] # created in reset()
def prerender(self):
"""
If you want to render the game to the screen, you will have to prerender
Load textures / images
"""
pygame.init()
pygame.key.set_repeat(1, 1)
self.display = pygame.display.set_mode(
(self.DISPLAY_WIDTH, self.DISPLAY_HEIGHT))
pygame.display.set_caption('Malmo Simulation')
self.clock = pygame.time.Clock()
pygame.font.init()
self.font = pygame.font.SysFont('Default', 32, bold=False)
# Creates a visual steve
self.steve.create(pygame, self.SCALE)
# Creates visual Food
self.food.create(pygame, self.SCALE)
# Creates visual Food
self.stick.create(pygame, self.SCALE)
# Creates visual Obstacles
self.obstacle.create(pygame, self.SCALE)
# Creates visual Lava
self.lava.create(pygame, self.SCALE)
# Creates visual Multipliers
self.zombie.create(pygame, self.SCALE)
# Creates the grid background
self.bg = pygame.image.load("./Images/Grid100.png").convert()
self.bg = pygame.transform.scale(self.bg,
(self.SCALE * 100, self.SCALE * 100))
def reset(self):
"""Reset the environment"""
self.steps = 0
# Reset the score to 0
self.score = 0
self.steve.health = 3
# Positions on the grid that are not allowed to spawn things
disallowed = []
# Create obstacles and lava in the environment
self.obstacle.array.clear()
self.obstacle.array_length = 0
self.lava.array.clear()
self.lava.array_length = 0
if self.MAP_PATH != None:
self.obstacle.reset_map(self.GRID_SIZE, self.MAP_PATH,
self.ENABLE_WRAP)
self.lava.reset_map(self.GRID_SIZE, self.MAP_PATH,
self.ENABLE_WRAP)
if not self.ENABLE_WRAP:
self.obstacle.create_border(self.GRID_SIZE, self.SCALE)
# Add all obstacle positions to the disallowed list
[disallowed.append(grid_pos) for grid_pos in self.obstacle.array]
[disallowed.append(grid_pos) for grid_pos in self.lava.array]
# Add random obstacles not already on the map
self.obstacle.reset(self.grid, disallowed)
# Add all obstacle positions to the disallowed list
disallowed = []
[disallowed.append(grid_pos) for grid_pos in self.obstacle.array]
[disallowed.append(grid_pos) for grid_pos in self.lava.array]
# Create lava in the environment
# self.lava.array.clear()
# self.lava.array_length = 0
# if self.MAP_PATH != None:
# self.lava.reset_map(self.GRID_SIZE, self.MAP_PATH, self.ENABLE_WRAP)
# [disallowed.append(grid_pos) for grid_pos in self.lava.array]
# Add random lava not already on the map
self.lava.reset(self.grid, disallowed)
disallowed = []
[disallowed.append(grid_pos) for grid_pos in self.obstacle.array]
[disallowed.append(grid_pos) for grid_pos in self.lava.array]
self.maze = createGrid(self.GRID_SIZE, disallowed, self.SCALE)
# print(self.obstacle.array)
# print("\n")
# print(self.maze)
# Reseting Steve at a random (or specific) position
self.steve.reset(self.grid, disallowed)
# Initialise the movement to not moving
if self.ACTION_SPACE == 5:
self.steve.dx = 0
self.steve.dy = 0
disallowed.append(self.steve.pos)
# Create a piece of food
self.food.reset(self.NUM_OF_FOOD, self.grid, disallowed)
# self.food.make_within_range(self.GRID_SIZE, self.SCALE, self.steve)
self.spawn_new_food = False
[disallowed.append(grid_pos) for grid_pos in self.food.array]
# Spawn in a zombie
self.zombie.reset(self.grid, disallowed)
# Fill the state array with the appropriate state representation
# self.state = self.state_array()
# self.state = self.state_vector_3D()
self.state = self.local_state_vector_3D()
# Reset the time
self.time = 0
# A dictionary of information that can be useful
info = {"time": self.time, "score": self.score}
return self.state, info
def quick_reset(self):
"""Reset the environment and places all entities according to the map layout"""
self.steps = 0
# Reset the score to 0
self.score = 0
self.steve.health = 5
# Positions on the grid that are not allowed to spawn things
disallowed = []
# Create obstacles and lava in the environment
self.obstacle.array.clear()
self.obstacle.array_length = 0
self.lava.array.clear()
self.lava.array_length = 0
self.zombie.array.clear()
self.zombie.amount = 0
self.steve.pos, self.food.array, self.stick.array, self.zombie.array, barrier = reset_impossible_map(
self.GRID_SIZE, self.MAP_PATH)
self.zombie.prev_array = self.zombie.array
# self.steve.pos = (0,0)
self.steve.x = self.steve.pos[0][0]
self.steve.y = self.steve.pos[0][1]
# quick fix for something silly
self.steve.pos[0] = self.steve.x
self.steve.pos.append(self.steve.y)
self.steve.history.clear()
self.steve.history.append((self.steve.x, self.steve.y))
self.steve.hasSword = False
self.obstacle.reset_map(self.GRID_SIZE, self.MAP_PATH,
self.ENABLE_WRAP)
self.lava.reset_map(self.GRID_SIZE, self.MAP_PATH, self.ENABLE_WRAP)
self.obstacle.create_border(self.GRID_SIZE, self.SCALE)
[disallowed.append(grid_pos) for grid_pos in self.obstacle.array]
[disallowed.append(grid_pos) for grid_pos in self.lava.array]
[disallowed.append(grid_pos) for grid_pos in barrier]
self.maze = createGrid(self.GRID_SIZE, disallowed, self.SCALE)
# Initialise the movement to not moving
if self.ACTION_SPACE == 5:
self.steve.dx = 0
self.steve.dy = 0
self.spawn_new_food = False
self.zombie.amount = len(self.zombie.array)
self.food.amount = len(self.food.array)
self.stick.amount = len(self.stick.array)
# Fill the state array with the appropriate state representation
# self.state = self.state_array()
# self.state = self.state_vector_3D()
self.state = self.local_state_vector_3D()
# Reset the time
self.time = 0
# A dictionary of information that can be useful
info = {"time": self.time, "score": self.score}
return self.state, info
def render(self):
"""Renders ONLY the CURRENT state"""
# Mainly to close the window and stop program when it's running
action = self.controls()
# Set the window caption to the score
# pygame.display.set_caption("Score: " + str(self.score))
# Draw the background, steve and the food
self.display.blit(self.bg, (0, 0))
self.obstacle.draw(self.display)
self.lava.draw(self.display)
self.food.draw(self.display)
self.stick.draw(self.display)
self.zombie.draw(self.display)
self.steve.draw(self.display)
self.steve.draw_health(self.display)
# Text on screen
text = self.font.render("Score: " + str(int(self.score)), True,
(240, 240, 240, 0))
self.display.blit(text, (10, 0))
# text = self.font.render("Multiplier: "+str(self.steve.score_multiplier)+"x", True, (240, 240, 240, 0))
# self.display.blit(text,(150,0))
# Update the pygame display
pygame.display.update()
# print(pygame.display.get_surface().get_size())
# pygame.display.get_surface().lock()
# p = pygame.PixelArray(pygame.display.get_surface())
# p = pygame.surfarray.array3d(pygame.display.get_surface())
# print("PIXELS:", p.shape)
# pygame.display.get_surface().unlock()
# print(clock.get_rawtime())
# clock.tick(self.FPS) # FPS setting
# Adds a delay to the the game when rendering so that humans can watch
pygame.time.delay(self.UPDATE_RATE)
return action
def end(self):
"""
Ending the Game - This has to be at the end of the code
Clean way to end the pygame env
with a few backups...
"""
pygame.quit()
quit()
sys.exit(0) # safe backup
def wrap(self):
""" If the steve goes out the screen bounds, wrap it around"""
if self.steve.x > self.DISPLAY_WIDTH - self.SCALE:
self.steve.x = 0
if self.steve.x < 0:
self.steve.x = self.DISPLAY_WIDTH - self.SCALE
if self.steve.y > self.DISPLAY_HEIGHT - self.SCALE:
self.steve.y = 0
if self.steve.y < 0:
self.steve.y = self.DISPLAY_HEIGHT - self.SCALE
self.steve.pos = (self.steve.x, self.steve.y)
def step(self, action):
"""
Step through the game, one state at a time.
Return the reward, the new_state, whether its reached_diamond or not, and the time
"""
self.steps += 1
# Rewards:
# reward_each_time_step = 1.0
reward_each_time_step = -0.1
reward_collecting_diamond = 10.0
reward_out_of_bounds = -1.0 # not used
reward_zombie_hit = -10.0
reward_in_lava = -10.0
reward_barrier = -2.0
# Increment time step
self.time += 1
# If the steve has reached the food
reached_diamond = False
hit_obstacle = False
in_lava = False
# If the episode is finished - after a certain amount of timesteps or it crashed
done = False
# Initialze to -1 for every time step - to find the fastest route (can be a more negative reward)
reward = reward_each_time_step
# Negetive exponential distance rewards
# if len(self.zombie.array) > 0:
# distance_list = []
# for i, z in enumerate(self.zombie.array):
# distance_list.append((math.sqrt((self.steve.x - z[0])**2 + (self.steve.y - z[1])**2)/self.GRID_SIZE)/2)
# # print(distance_list)
# if min(distance_list) < 1.6:
# reward = reward_barrier
# normal_distance = ((distance/self.GRID_SIZE)*(pi/2))-pi/4
# normal_distance = ((distance/self.GRID_SIZE)*(1.0))-0.4
# reward = np.tan(normal_distance)
# reward = -np.exp(-distance)*10
# Linear distance reward
# if len(self.zombie.array) > 0:
# reward = (math.sqrt((self.steve.x - self.zombie.array[0][0])**2 + (self.steve.y - self.zombie.array[0][1])**2)/20)/self.GRID_SIZE
# Exponential distance reward
# if len(self.zombie.array) > 0:
# reward = (((self.steve.x - self.zombie.array[0][0])**2 + (self.steve.y - self.zombie.array[0][1])**2)/20**2)/self.GRID_SIZE
# Test: if moving, give a reward
# if self.steve.dx != 0 or self.steve.dy != 0:
# reward = -0.01
# Test for having a barrier around the zombie
# if reward < 0.143:
# reward = -5
# Update the position of steve
self.steve.update(self.SCALE, action, self.ACTION_SPACE)
if self.ENABLE_WRAP:
self.wrap()
else:
if self.steve.x > self.DISPLAY_WIDTH - self.SCALE:
reward = reward_out_of_bounds
done = True
if self.steve.x < 0:
reward = reward_out_of_bounds
done = True
if self.steve.y > self.DISPLAY_HEIGHT - self.SCALE:
reward = reward_out_of_bounds
done = True
if self.steve.y < 0:
reward = reward_out_of_bounds
done = True
# Check for obstacle collision
for i in range(self.obstacle.array_length):
hit_obstacle = (self.steve.pos == self.obstacle.array[i])
if hit_obstacle:
self.steve.x = self.steve.prev_pos[0]
self.steve.y = self.steve.prev_pos[1]
self.steve.pos = self.steve.prev_pos
self.steve.history.pop(len(self.steve.history) - 1)
self.steve.history.append(self.steve.pos)
# done = True
# reward = -0.1
# Check for lava collision
for i in range(self.lava.array_length):
in_lava = (self.steve.pos == self.lava.array[i])
if in_lava:
done = True
reward = reward_in_lava
# Update the position of the zombie
# self.zombie.move(self.maze, self.steve, self.steps, self.SCALE)
self.zombie.range_move(self.maze, self.steve, self.steps, self.SCALE)
# Check if zombie gets steve
for i in range(self.zombie.amount):
# print(self.steve.pos, self.zombie.array[i])
if self.steve.pos == self.zombie.array[i]:
zombie_hit = True
else:
zombie_hit = False
self.steve.isHit = False
if zombie_hit:
self.zombie.move_back(i)
self.steve.isHit = True
self.steve.health = self.steve.health - 1
# done = True
reward = reward_zombie_hit
break
# Make the most recent history have the most negative rewards
if self.steve.history_size != 0:
decay = (reward_each_time_step) / (self.steve.history_size)
for i in range(len(self.steve.history) - 1):
# print(i,-1*(1-decay*i))
if ((self.steve.pos) == self.steve.history[-i - 2]):
# reward = -1*(1-decay*i)
break
# Checking if Steve has reached the diamond
reached_diamond, diamond_index = self.food.eat(self.steve)
# ADDED FOR ALLOWING THE MODEL TO HAVE STEVE OVER THE FOOD IN THE STATE
# if self.spawn_new_food:
# disallowed = []
# [disallowed.append(grid_pos) for grid_pos in self.obstacle.array]
# # [disallowed.append(grid_pos) for grid_pos in self.zombie.array]
# # if self.steve.pos not in disallowed:
# # disallowed.append(self.steve.pos)
# self.food.make(self.grid, disallowed, index = self.last_eaten_food)
# self.spawn_new_food = False
# reached_diamond = False
# If Steve reaches the food, increment score
if reached_diamond:
self.score += 1
self.last_eaten_food = diamond_index
self.spawn_new_food = False
# Create a piece of food that is not within Steve
disallowed = []
[disallowed.append(grid_pos) for grid_pos in self.obstacle.array]
# [disallowed.append(grid_pos) for grid_pos in self.zombie.array]
self.food.make(self.grid, disallowed, index=diamond_index)
# Only collect 1 food item at a time
# done = True
# Reward functions
reward = reward_collecting_diamond
else:
self.spawn_new_food = False
# If Steve collects a stick, increment score
collected_stick, stick_index = self.stick.collect(self.steve)
if collected_stick:
self.score += 1
# Create a piece of food that is not within Steve
disallowed = []
[disallowed.append(grid_pos) for grid_pos in self.obstacle.array]
self.stick.make(self.grid, disallowed, index=stick_index)
# Only collect 1 food item at a time
# done = True
# Reward functions
reward = reward_collecting_diamond
# For fun
if self.score == 3:
self.steve.hasSword = True
done = True
# To make it compatible with malmo
if self.score == self.NUM_OF_FOOD:
# reward = 1
pass
if self.NUM_OF_FOOD != 0:
done = True
pass
if self.steve.health <= 0:
done = True
# If the episode takes longer than the max time, it ends
if self.time == self.MAX_TIME_PER_EPISODE:
# print("Steve survived :)")
# reward = -1.0
done = True
# Get the new_state
# new_state = self.state_array()
# new_state = self.state_vector_3D()
new_state = self.local_state_vector_3D()
# print(reward)
# print(self.steve.history)
# A dictionary of information that may be useful
info = {"time": self.time, "score": self.score}
return new_state, reward, done, info
def state_index(self, state_array):
"""
Given the state array, return the index of that state as an integer
Used for the Qlearning lookup table
"""
return int((self.GRID_SIZE**3) * state_array[0] +
(self.GRID_SIZE**2) * state_array[1] +
(self.GRID_SIZE**1) * state_array[2] +
(self.GRID_SIZE**0) * state_array[3])
def sample_action(self):
"""
Return a random action
Can't use action space 4 with a tail, else it will have a double chance of doing nothing
or crash into itself
"""
return np.random.randint(0, self.ACTION_SPACE)
def set_state(self, state):
"""Set the state of the game environment"""
self.state = state
def set_map(self, map_path):
self.MAP_PATH = map_path
def number_of_states(self):
"""
Return the number of states with just the steve head and 1 food
Used for Q Learning look up table
"""
return (self.GRID_SIZE**2) * ((self.GRID_SIZE**2))
def number_of_actions(self):
"""
Return the number of possible actions
Action space:
> nothing, up, down, left, right
"""
return self.ACTION_SPACE
def state_array(self):
"""
The state represented as an array or steve positions and food positions
Used for Q learning
"""
new_state = np.zeros(4)
new_state[0] = int(self.steve.x / self.SCALE)
new_state[1] = int(self.steve.y / self.SCALE)
new_state[2] = int(self.food.x / self.SCALE)
new_state[3] = int(self.food.y / self.SCALE)
return new_state
def state_vector(self):
"""
The state represented as a onehot 1D vector
Used for the feed forward NN
"""
# (rows, columns)
state = np.zeros((self.GRID_SIZE**2, 3))
# This is for STEVE, the FOOD and EMPTY
for i in range(self.GRID_SIZE): # rows
for j in range(self.GRID_SIZE): # columns
if ((self.steve.x / self.SCALE) == j
and (self.steve.y / self.SCALE) == i):
state[i * self.GRID_SIZE + j] = [1, 0, 0]
# print("steve:", i*self.GRID_SIZE+j)
elif ((self.food.x / self.SCALE) == j
and (self.food.y / self.SCALE) == i):
state[i * self.GRID_SIZE + j] = [0, 1, 0]
# print("Food:", i*self.GRID_SIZE+j)
else:
state[i * self.GRID_SIZE + j] = [0, 0, 1]
# Flatten the vector to a 1 dimensional vector for the input layer to the NN
state = state.flatten()
state = state.reshape(1, (self.GRID_SIZE**2) * 3)
# state = np.transpose(state)
return state
def state_vector_3D(self):
"""
State as a 3D vector of the whole map for the CNN
Shape = (Layers, GRID_SIZE, GRID_SIZE)
"""
state = np.zeros((3, self.GRID_SIZE, self.GRID_SIZE))
state[0,
int(self.steve.y / self.SCALE),
int(self.steve.x / self.SCALE)] = 1
state[1,
int(self.food.y / self.SCALE),
int(self.food.x / self.SCALE)] = 1
# Obstacles
for i in range(self.obstacle.array_length):
state[2,
int(self.obstacle.array[i][1] / self.SCALE),
int(self.obstacle.array[i][0] / self.SCALE)] = 1
# Zombies
# for i in range(len(self.zombie.array)):
# state[1, int(self.zombie.array[i][1]/self.SCALE), int(self.zombie.array[i][0]/self.SCALE)] = 1
return state
def local_state_vector_3D(self):
"""
State as a 3D vector of the local area around the steve
Shape = (Layers, LOCAL_GRID_SIZE, LOCAL_GRID_SIZE)
"""
s_pos = 0
d_pos = 1
z_pos = 2
h_pos = 3
l_pos = 4
o_ops = 5
#s = steve
sx = int(self.steve.x / self.SCALE)
sy = int(self.steve.y / self.SCALE)
state = np.zeros((6, self.LOCAL_GRID_SIZE, self.LOCAL_GRID_SIZE))
# Agent
local_pos = int((self.LOCAL_GRID_SIZE - 1) / 2)
state[s_pos, local_pos, local_pos] = 1
# Diamonds
for i in range(self.food.amount):
x_prime_food = local_pos + int(
self.food.array[i][0] / self.SCALE) - sx
y_prime_food = local_pos + int(
self.food.array[i][1] / self.SCALE) - sy
if x_prime_food < self.LOCAL_GRID_SIZE and x_prime_food >= 0 and y_prime_food < self.LOCAL_GRID_SIZE and y_prime_food >= 0:
state[d_pos, y_prime_food, x_prime_food] = 1
pass
# Zombies
for i in range(len(self.zombie.array)):
x_prime_zom = local_pos + int(
self.zombie.array[i][0] / self.SCALE) - sx
y_prime_zom = local_pos + int(
self.zombie.array[i][1] / self.SCALE) - sy
if x_prime_zom < self.LOCAL_GRID_SIZE and x_prime_zom >= 0 and y_prime_zom < self.LOCAL_GRID_SIZE and y_prime_zom >= 0:
state[z_pos, y_prime_zom, x_prime_zom] = 1
pass
# Obstacles
for i in range(self.obstacle.array_length):
x_prime_obs = local_pos + int(
self.obstacle.array[i][0] / self.SCALE) - sx
y_prime_obs = local_pos + int(
self.obstacle.array[i][1] / self.SCALE) - sy
if x_prime_obs < self.LOCAL_GRID_SIZE and x_prime_obs >= 0 and y_prime_obs < self.LOCAL_GRID_SIZE and y_prime_obs >= 0:
state[o_ops, y_prime_obs, x_prime_obs] = 1
pass
# Out of bounds
for j in range(0, self.LOCAL_GRID_SIZE):
for i in range(0, self.LOCAL_GRID_SIZE):
x_prime_wall = local_pos - sx
y_prime_wall = local_pos - sy
if i < x_prime_wall or j < y_prime_wall:
state[o_ops, j, i] = 1
pass
x_prime_wall = local_pos + (self.GRID_SIZE - sx) - 1
y_prime_wall = local_pos + (self.GRID_SIZE - sy) - 1
if i > x_prime_wall or j > y_prime_wall:
state[o_ops, j, i] = 1
pass
# Lava
for i in range(self.lava.array_length):
x_prime_lava = local_pos + int(
self.lava.array[i][0] / self.SCALE) - sx
y_prime_lava = local_pos + int(
self.lava.array[i][1] / self.SCALE) - sy
if x_prime_lava < self.LOCAL_GRID_SIZE and x_prime_lava >= 0 and y_prime_lava < self.LOCAL_GRID_SIZE and y_prime_lava >= 0:
state[l_pos, y_prime_lava, x_prime_lava] = 1
pass
# History
if self.steve.history_size != 0:
decay = 1 / self.steve.history_size
for i in range(len(self.steve.history) - 1):
x_prime = local_pos + int(
self.steve.history[-i - 2][0] / self.SCALE) - sx
y_prime = local_pos + int(
self.steve.history[-i - 2][1] / self.SCALE) - sy
if x_prime < self.LOCAL_GRID_SIZE and x_prime >= 0 and y_prime < self.LOCAL_GRID_SIZE and y_prime >= 0:
if 1 - decay * i >= 0 and state[h_pos, y_prime, x_prime] == 0:
state[h_pos, y_prime, x_prime] = 1 - decay * i
pass
# Delete these layers:
# state = np.delete(state, s_pos, 0)
# state = np.delete(state, d_pos, 0)
# state = np.delete(state, z_pos, 0)
# state = np.delete(state, h_pos, 0)
# state = np.delete(state, l_pos, 0)
# state = np.delete(state, o_pos, 0)
# DIAMOND DOJO
# state = np.delete(state, 2, 0)
# state = np.delete(state, 2, 0)
# ZOMBIE DOJO
# state = np.delete(state, 1, 0)
# state = np.delete(state, 2, 0)
# EXPLORE DOJO
# state = np.delete(state, 1, 0)
# state = np.delete(state, 1, 0)
# GRID 9 SETUP
# state = np.delete(state, 1, 0)
# state = np.delete(state, 2, 0)
# state = np.delete(state, 2, 0)
# state = np.delete(state, 3, 0)
# state = np.delete(state, 4, 0)
return state
def get_pixels(self):
"""
Returns the pixels in a (GRID*20, GRID*20, 3) size array/
Unfortunatly it has to render in order to gather the pixels
"""
return pygame.surfarray.array3d(pygame.display.get_surface())
def controls(self):
"""
Defines all the players controls during the game
"""
action = 0 # Do nothing as default
for event in pygame.event.get():
# print(event) # DEBUGGING
if event.type == pygame.QUIT:
self.end()
if event.type == pygame.KEYDOWN:
# print(event.key) #DEBUGGING
# In order to quit easily
if (event.key == pygame.K_q):
self.end()
# Moving up
if (event.key == pygame.K_UP or event.key == pygame.K_w):
if self.ACTION_SPACE == 5:
action = 1 # up
# Moving down
elif (event.key == pygame.K_DOWN or event.key == pygame.K_s):
if self.ACTION_SPACE == 5:
action = 2 # down
# Moving left
elif (event.key == pygame.K_LEFT or event.key == pygame.K_a):
if self.ACTION_SPACE == 5:
action = 3 # left
# Moving right
elif (event.key == pygame.K_RIGHT or event.key == pygame.K_d):
if self.ACTION_SPACE == 5:
action = 4 # right
return action
def play(self):
"""
Lets you simply play the game
Useful for debugging and testing out the environment
"""
GAME_OVER = False
self.prerender()
# self.reset()
for i in range(10):
# MAP_NUMBER = np.random.randint(5)
# MAP_NUMBER = 1
# MAP_PATH = "./Maps/Grid{}/map{}.txt".format(self.GRID_SIZE-2, MAP_NUMBER)
# MAP_PATH = "./Maps/Grid{}/impossible_map_empty{}.txt".format(self.GRID_SIZE-2, MAP_NUMBER)
# self.set_map(MAP_PATH)
# self.reset()
self.quick_reset()
GAME_OVER = False
while not GAME_OVER:
# print(self.steve.history)
action = self.controls()
# action = self.render()
# print(self.get_pixels().shape)
s, r, GAME_OVER, i = self.step(action)
# print("\n\n") # DEBUGGING
# print(self.state_vector_3D()) # DEBUGGING
# print(self.local_state_vector_3D()) # DEBUGGING
# print(r)
self.render()
if GAME_OVER:
print("Game Over: time:", i["time"])
self.end()
