env_goal_size=env.goal_size)
# parameters
EPISODES = 5000
for i in range(EPISODES):
state = env.reset()
done = False
while not done:
env.show()
print()
# agent 1 decides its action
actionME = "choose your action here"
# agent 2 decides its action
actionOP = agentOP.get_action(state)
# perform actions on the environment
done, reward_l, reward_r, state_, actions = env.step(
actionME, actionOP)
# training process of agent 1
""" do some training here """
# training process of agent 2
agentOP.adjust(done, reward_r, i)
state = state_
for i in range(EPISODES):
state = env.reset()
stat.set_initial_ball(state[4])
rewardL = 0
rewardR = 0
done = False
while not done:
# agent 1 decides its action
actionL = random.randint(0, env.act_dim-1)
# agent 2 decides its action
actionR = random.randint(0, env.act_dim-1)
# perform actions on the environment
done, reward_l, reward_r, state_, actions = env.step(actionL, actionR)
state = state_
rewardL += reward_l
rewardR += reward_r
if done:
stat.add_stat(rewardL, rewardR)
print(*stat.get_moving_avg(),
file=open('log_files/baseline_reward.txt', 'a'))
# save stats
with open('stats/baseline.pkl', 'wb') as output:
pickle.dump(stat, output)
from env import SoccerEnv
from agents.common.training_opponent import StationaryOpponent, RandomSwitchOpponent, RLBasedOpponent
TOP = 0
TOP_RIGHT = 1
RIGHT = 2
BOTTOM_RIGHT = 3
BOTTOM = 4
BOTTOM_LEFT = 5
LEFT = 6
TOP_LEFT = 7
env = SoccerEnv()
agentOP = StationaryOpponent(env_width=env.width, env_height=env.height, env_goal_size=env.goal_size)
state = env.reset()
# loop
env.show()
actionOP = agentOP.get_action(state)
print(actionOP)
done, reward_l, reward_r, state, actions = env.step("type action here!", actionOP)
agentOP.adjust(done, reward_r, i)