def battle(process_number):
# プロセスIDを取得しておく。
pid = os.getpid()
print("battle start, process_number={}, pid={}".format(
process_number, pid))
# Javaへの接続を
gateway = JavaGateway()
addition_app = gateway.entry_point
# ゲーム開始を伝える。
addition_app.start_game(pid)
# List of four agents
agent_list = [
#agents.SimpleAgent(),
#MyAgentO(),
#MyAgentT(),
#MyAgentO(),
#MyAgentT(),
MyAgentT(),
MyAgentT(),
MyAgentT(),
MyAgentT(),
]
env = pommerman.make('PommeTeamCompetition-v0', agent_list)
#env = pommerman.make('PommeTeamCompetition-v1', agent_list)
#env = pommerman.make('PommeTeamCompetitionFast-v0', agent_list)
#env = pommerman.make('PommeTeamCompetitionFast-v1', agent_list)
numAliveLast = 4
state = env.reset()
step = 0
done = False
while not done:
step += 1
actions = env.act(state)
state, reward, done, info = env.step(actions)
print("pid={} step={} actions={}".format(pid, step, actions))
temp = state[0]
temp2 = temp['alive']
numAlive = len(temp2)
if numAlive != numAliveLast:
numAliveLast = numAlive
addition_app.changed()
# 結果を出力する。
print("battle finished, process_number={}, pid={}, reward={}".format(
process_number, pid, reward))
# ゲーム終了を伝える。
addition_app.finish_game(pid, reward[0], reward[1], reward[2], reward[3])
# 不要なハンドルを閉じる。
env.close()
def battle(episode):
pid = os.getpid()
print(episode, ", ", pid)
res = addition_app.start_game(pid)
print(res)
# List of four agents
if False:
agent_list = [
MyAgentT(),
MyAgentT(),
MyAgentT(),
MyAgentT(),
]
agent_list = [
agents.SimpleAgent(),
agents.SimpleAgent(),
agents.SimpleAgent(),
agents.SimpleAgent(),
]
agent_list = [
MyAgentO(),
MyAgentT(),
MyAgentO(),
MyAgentT(),
]
env = pommerman.make('PommeTeamCompetition-v0', agent_list)
state = env.reset()
done = False
step = 0
while not done:
step += 1
actions = env.act(state)
state, reward, done, info = env.step(actions)
# rewards.append(reward)
print('Episode {} finished'.format(episode), reward)
res = addition_app.finish_game(pid, reward[0], reward[1], reward[2],
reward[3])
print(res)
def main(render=False, interactive=False):
# List of four agents
agent_list = [
agents.SimpleAgent(),
agents.SimpleAgent(),
agents.SimpleAgent(),
agents.SimpleAgent(),
]
agent_list = [
MyAgentT(),
MyAgentT(),
MyAgentT(),
MyAgentT(),
]
agent_list = [
MyAgentO(),
MyAgentT(),
MyAgentO(),
MyAgentT(),
]
# Environment of FFA competition
#env = pommerman.make('PommeFFACompetition-v0', agent_list)
env = pommerman.make('PommeTeamCompetition-v0', agent_list)
# Run
rewards = list()
for episode in range(10000000):
state = env.reset()
done = False
step = 0
while not done:
if verbose:
print("Step: ", step)
step += 1
if render:
env.render()
actions = env.act(state)
if verbose:
print(actions[-1])
if forwardModelDebug:
send_env(env, 0)
send_action(actions)
# compute step()
state, reward, done, info = env.step(actions)
if forwardModelDebug:
send_env(env, 1)
if interactive:
sys.stdin.readline()
#rewards.append(reward)
print('Episode {} finished'.format(episode), reward)
rewards = np.array(rewards)
print(np.mean(rewards, axis=0))
env.close()
def __init__(self):
self._agent = MyAgentT()