MLE/04_pacman_rl/pacman.py

217 lines
6.6 KiB
Python

import pygame
import math
import reinforcement_learning as rl
# Initialize pygame
pygame.init()
# Define constants
SCREEN_WIDTH = 400
SCREEN_HEIGHT = 400
CELL_SIZE = 40
# Define colors
YELLOW = (255, 255, 0)
RED = (255, 0, 0)
WHITE = (255, 255, 255)
BLUE = (0, 0, 255)
BLACK = (0, 0, 0)
# Labyrinth as a string
labyrinth = [
"##########",
"#........#",
"#.##..##.#",
"#........#",
"##########"
]
# Get labyrinth dimensions
ROWS = len(labyrinth)
COLS = len(labyrinth[0])
# Q-Learning Constants
GAMMA = 0.90
ALPHA = 0.2
# Initialize game screen
screen = pygame.display.set_mode((COLS * CELL_SIZE, ROWS * CELL_SIZE))
pygame.display.set_caption("Micro-Pacman")
# Pacman class
class Pacman:
def __init__(self, x, y):
self.x = x
self.y = y
self.count = 0
def move(self, dx, dy):
new_x, new_y = self.x + dx, self.y + dy
if labyrinth[new_y][new_x] != "#":
self.x = new_x
self.y = new_y
def draw(self):
radius = CELL_SIZE // 2 - 4
start_angle = math.pi / 6
end_angle = -math.pi / 6
pygame.draw.circle(screen, YELLOW, (self.x * CELL_SIZE + CELL_SIZE // 2, self.y * CELL_SIZE + CELL_SIZE // 2), CELL_SIZE // 2 - 4)
# Calculate the points for the mouth
start_pos = (self.x* CELL_SIZE + CELL_SIZE // 2 + int(radius*1.3 * math.cos(start_angle)),
self.y* CELL_SIZE + CELL_SIZE // 2 - int(radius*1.3 * math.sin(start_angle)))
end_pos = (self.x* CELL_SIZE + CELL_SIZE // 2 + int(radius*1.3 * math.cos(end_angle)),
self.y* CELL_SIZE + CELL_SIZE // 2 - int(radius*1.3 * math.sin(end_angle)))
self.count += 1
if self.count%2==0:
# Draw the mouth by filling a polygon
pygame.draw.polygon(screen, BLACK, [(self.x* CELL_SIZE + CELL_SIZE // 2, self.y* CELL_SIZE + CELL_SIZE // 2), start_pos, end_pos])
# Ghost class with pixel art
class Ghost:
# Define the pixel art for the ghost using strings
ghost_pixels = [
" #### ",
"######",
"## # #",
"######",
"######",
"# # # "
]
def __init__(self, x, y):
self.x = x
self.y = y
def move_towards_pacman(self, pacman):
if self.x < pacman.x and labyrinth[self.y][self.x + 1] != "#":
self.x += 1
elif self.x > pacman.x and labyrinth[self.y][self.x - 1] != "#":
self.x -= 1
elif self.y < pacman.y and labyrinth[self.y + 1][self.x] != "#":
self.y += 1
elif self.y > pacman.y and labyrinth[self.y - 1][self.x] != "#":
self.y -= 1
def draw(self):
pixel_size = CELL_SIZE // len(self.ghost_pixels) # Size of each pixel in the ghost art
for row_idx, row in enumerate(self.ghost_pixels):
for col_idx, pixel in enumerate(row):
if pixel == "#":
pixel_x = self.x * CELL_SIZE + col_idx * pixel_size
pixel_y = self.y * CELL_SIZE + row_idx * pixel_size
pygame.draw.rect(screen, RED, (pixel_x, pixel_y, pixel_size, pixel_size))
# Draw walls and cookies
def draw_labyrinth():
for y, row in enumerate(labyrinth):
for x, cell in enumerate(row):
if cell == "#":
pygame.draw.rect(screen, BLUE, (x * CELL_SIZE, y * CELL_SIZE, CELL_SIZE, CELL_SIZE))
elif cell == ".":
pygame.draw.circle(screen, WHITE, (x * CELL_SIZE + CELL_SIZE // 2, y * CELL_SIZE + CELL_SIZE // 2), 5)
def move_pacman(pacman, a):
if a == 0: # left
pacman.move(-1, 0)
if a == 1: # right
pacman.move(1, 0)
if a == 2: # up
pacman.move(0, -1)
if a == 3: # down
pacman.move(0, 1)
# Main game function
def main():
global labyrinth
q = rl.q_init()
clock = pygame.time.Clock()
# Game loop
not_won = True
outer_iter = 0
while not_won:
labyrinth = [
"##########",
"#........#",
"#.##..##.#",
"#........#",
"##########"
]
running = True
iter = 0
# Initialize Pacman and Ghost positions
pacman = Pacman(1, 1)
ghost = Ghost(COLS - 2, ROWS - 2)
s = (pacman.x, pacman.y, ghost.x, ghost.y) # as a tuple so the state becomes hashable
# Handle events
for event in pygame.event.get():
if event.type == pygame.QUIT:
not_won = False
print(outer_iter)
while running or iter < 100:
screen.fill(BLACK)
iter = iter + 1
# Check for collisions (game over if ghost catches pacman)
if pacman.x == ghost.x and pacman.y == ghost.y:
print("Game Over! The ghost caught Pacman.")
outer_iter = outer_iter + 1
running = False
break
# Eat cookies
if labyrinth[pacman.y][pacman.x] == ".":
labyrinth[pacman.y] = labyrinth[pacman.y][:pacman.x] + " " + labyrinth[pacman.y][pacman.x+1:]
# Check if all cookies are eaten (game over)
if all("." not in row for row in labyrinth):
print("You Win! Pacman ate all the cookies.")
running = False
not_won = False
break
# Q-Learning part ############################################################################
a = rl.epsilon_greedy(q, s) # 0 = Left; 1 = Right ; 2 = Up ; 3 = Down
s_new, r, labyrinth = rl.take_action(s, a, labyrinth)
# print(s) # debugging
# print(q[s]) # debugging
q[s][a] += ALPHA * (r + GAMMA * rl.max_q(q, s_new, labyrinth) - q[s][a])
s = s_new
move_pacman(pacman, a)
if iter % 3 == 0:
# Ghost moves towards Pacman
ghost.move_towards_pacman(pacman)
# Update state
s = (pacman.x, pacman.y, ghost.x, ghost.y)
# End of Q-Learning part ######################################################################
# Draw the labyrinth, pacman, and ghost
draw_labyrinth()
pacman.draw()
ghost.draw()
# Update display
pygame.display.flip()
# Cap the frame rate
# tick_speed = 100
tick_speed = 5 if outer_iter % 20 == 0 else 50
clock.tick(tick_speed)
pygame.quit()
if __name__ == "__main__":
main()