from gym import logger
from pong.pong_game import PongGame
from pong.gym_agents import RandomAgent
if __name__ == '__main__':
# You can set the level to logger.DEBUG or logger.WARN if you
# want to change the amount of output.
logger.set_level(logger.INFO)
env = PongGame()
env.seed(0)
agent1 = RandomAgent()
agent2 = RandomAgent()
episode_count = 1
reward = 0
done = False
print(env.action_space.n)
for i in range(episode_count):
ob = env.reset()
while True:
action1 = agent1.act(ob, player=0)
action2 = agent2.act(ob, player=1)
ob, reward = env.step(action1, a2=action2)
if done:
break
env.render()
# Close the env and write monitor result info to disk
from pong.gym_agents import *
from pong.monitor import PongMonitor
possible_opponents = {
1: RandomAgent,
2: GreedyAgent,
3: AggressiveAgent,
4: LazyAgent
}
print("Welcome in Pong")
selected_opponent, = input(
"Select opponent for MCTS (1 - Random, 2 - Safe, 3 - Aggressive, 4 - Lazy): "
).split()
game = PongGame()
game = PongMonitor(game, ".", force=True)
game.reset()
opponent = possible_opponents[int(selected_opponent)]()
mcts_agent = GreedyAgent()
tree = Mcts(game, simulation_agent=mcts_agent)
# tree = Mcts(game)
count = 0
while not game.done:
count = count + 1
start = time()
tree.run(30, verbose=True)
from mcts.mcts import Mcts
from pong.pong_game import PongGame
from time import time
from pong.gym_agents import *
from pong.monitor import PongMonitor
from ddqn.ddqn_agent import DdqnAgent
game = PongGame()
# game = PongMonitor(game, ".", force=True)
opponent = AggressiveAgent()
# opponent = GreedyAgent()
ddqn_agent = RandomAgent()
_scores = []
_bounces = []
for i in range(15):
game.reset()
scores = [0, 0]
prev_bounce = 0
while 21 not in scores:
ob = game._get_obs()
game.render()
action1 = ddqn_agent.act(ob, player=0)
game.act(action1)
action2 = opponent.act(ob, player=1)
reward = game.act(action2)
# game.render()
# for run in [60]:
# for parameter in [False]:
# for i in range(0, 15):
# playouts.append({
# 'runs': run,
# 'agent': agent,
# 'method': method,
# 'skip_actions': parameter,
# 'exploration_parameter': 1.41
# })
for playout in playouts:
print('Playing pong with {} runs, using {} method, against {} opponent'.format(playout['runs'], playout['method'],
opponent_names[
playout['agent']]))
game = PongGame()
filename = './logs-defence/' + playout['method'] + '-' + \
('no-skip' if playout['exploration_parameter'] is False else 'with-skip') + \
'-' + str(playout['exploration_parameter']) + \
'/pong-' + playout['method'] + '-' + str(playout['runs']) + '-against-' + opponent_names[
playout['agent']] + '_' + datetime.now().strftime("%Y%m%d-%H%M%S")
print(filename)
game = PongMonitor(game, filename, force=False)
game.reset()
pong_logger = PDLogger(filename)
opponent = possible_opponents[agent]()
mcts_agent = GreedyAgent()
tree = None
if playout['method'] == 'greedy':
a = 3
if a <= 0 or a >= 6:
return
if human_agent_action == a:
human_agent_action = 0
print("Press keys (arrow up) (arrow down) to take actions (2 and 3).")
print("No keys pressed is taking action 0 (stay in the same place)")
if __name__ == '__main__':
# You can set the level to logger.DEBUG or logger.WARN if you
# want to change the amount of output.
logger.set_level(logger.INFO)
env = PongGame()
outdir = '/tmp/random-agent-results'
ACTIONS = env.action_space.n
SKIP_CONTROL = 0 # Use previous control decision SKIP_CONTROL times, that's how you
# can test what skip is still usable.
env = wrappers.Monitor(env, directory=outdir, force=True)
env.seed(0)
env.render()
env.unwrapped.viewer.window.on_key_press = key_press
env.unwrapped.viewer.window.on_key_release = key_release
episode_count = 100
reward = 0
done = False