def init_game(random = True):
def get_random_orientation():
return np.random.choice([0, 90, 180, 270])
game = tron.TronGame(width = args.width, height = args.height, save_history = args.save_replay)
if random:
game.set_player_pos(game.get_random_pos(), game.get_random_pos())
game.set_player_orientation([get_random_orientation(), get_random_orientation()])
return game
def demo_game(net1, net2):
print("PLAYING DEMO GAME!")
game = tron.TronGame(width=20, height=20)
finished = False
while not finished:
evolve(game, net1, net2)
gameState = game.get_game_state()
finished = gameState[0]
gameState = game.get_game_state()
print(gameState)
def game_outcome(net1, net2):
game = tron.TronGame(width=gameSize[0], height=gameSize[1])
finished = False
while not finished:
evolve(game, net1, net2)
gameState = game.get_game_state()
finished = gameState[0]
winners = gameState[4]
if (winners[0] and not winners[1]):
#print("PLAYER 2 has WON")
return 2
if (not winners[0] and winners[1]):
#print("PLAYER 1 has WON")
return 1
else:
return 0
from strategy.reinforcement_learning.rl_strategy_train import *
import tron
import pickle
import os
from time import time
env = tron.TronGame(width=30, height=30)
STRATEGY_FILE = 'tmp/rl_strategy.npy'
def calculate_reward(player_idx):
return env.tick + (-10 if env.player_lost[player_idx] else 20)
USE_CHECK_POINT = True
NUM_GAMES = 1000000
if USE_CHECK_POINT and os.path.exists(STRATEGY_FILE):
with open(STRATEGY_FILE, 'rb') as f:
RLS = pickle.load(f)
else:
RLS = [
QLearningTable(actions=list(env.get_available_actions()))
for x in range(2)
]
def check_point():
with open(STRATEGY_FILE, 'wb') as f:
pickle.dump(RLS, f)
def init_game(args):
env = tron.TronGame(width=args.width, height=args.width)
env.reset(set_random=True)
return env
def init_game(random=True):
game = tron.TronGame(width=args.width, height=args.height)
if random:
game.set_player_pos(game.get_random_pos(), game.get_random_pos())
return game
def main():
args = get_args()
used_replays = replay_helper.get_all_replays(args.replay_folder)
pygame.font.init()
font = pygame.font.SysFont("Verdana", 20)
text_color = (120, 120, 120)
text_position = (5, 5)
game_offset = (5, 100)
for replay_file, replay in used_replays:
info = replay.info
start_positions = [body[0] for body in replay.bodies]
start_orientations = [
orientation[0] for orientation in replay.orientations
]
width, height = replay.info['width'], replay.info['height']
game = tron.TronGame(width=width, height=height, save_history=False)
game.set_player_orientation(start_orientations)
game.set_player_pos(*start_positions)
pygame.init()
screen = pygame.display.set_mode(
(int(width * args.player_dim) + game_offset[0] * 2,
int(height * args.player_dim) + game_offset[1]))
pygame.display.set_caption('Tron')
pygame.mouse.set_visible(0)
texts = [replay_file.split('/')[-1]]
if 'strategies' in info:
strategies = info['strategies']
texts.append(" vs. ".join(strategies))
def render_text(t):
return font.render(t, True, text_color)
texts = [render_text(t) for t in texts]
clock = pygame.time.Clock()
for idx, game_field in enumerate(replay.game_field):
clock.tick(1000)
resume_this_game = draw_game(screen,
game_field,
clock,
args.player_dim,
args.tick_delay,
background_offset=game_offset)
for idx, text in enumerate(texts +
[render_text('Tick: {}'.format(idx))]):
screen.blit(text,
(text_position[0], text_position[1] + (30 * idx)))
pygame.display.flip()
pygame.time.wait(args.tick_delay)
if not resume_this_game:
break
pygame.time.wait(args.delay_after_gameover)
# OLD
if False:
for p1_action, p2_action in zip(*replay.actions):
game.set_action(0, p1_action)
game.set_action(1, p2_action)
'FALSE'
resume_this_game = draw_game(screen, game, clock,
args.player_dim, args.tick_delay)