env = ICRABattleField()
env.seed(args.seed)
losses = []
rewards = []
for i_episode in range(1, args.epoch + 1):
print("Epoch: [{}/{}]".format(i_episode, args.epoch))
# Initialize the environment and state
action = Action()
pos = env.reset()
if args.enemy == "hand":
agent2.reset(pos)
state, reward, done, info = env.step(action)
for t in (range(2*60*30)):
# Other agent
if args.enemy == "hand":
env.set_robot_action(ID_B1, agent2.select_action(state[ID_B1]))
elif args.enemy == "AC":
env.set_robot_action(ID_B1, agent2.select_action(
state[ID_B1], mode="max_probability"))
# Select and perform an action
state_map = agent.preprocess(state[ID_R1])
a_m, a_t = agent.run_AC(state_map)
action = agent.decode_action(a_m, a_t, state[ID_R1], "max_probability")
# Step
next_state, reward, done, info = env.step(action)
tensor_next_state = agent.preprocess(next_state[ID_R1])
# Store the transition in memory
agent.push(state_map, tensor_next_state, [a_m, a_t], [reward])
losses = []
rewards = []
for i_episode in range(1, args.epoch + 1):
print("Epoch: [{}/{}]".format(i_episode, args.epoch))
# Initialize the environment and state
action = Action()
pos = env.reset()
if args.enemy == "hand":
pass
# agent2.reset([7.5, 0.5])
state, reward, done, info = env.step([action, action])
for t in (range(2 * 60 * 30)):
# Other agent
if args.enemy == 'person':
env.set_robot_action(ID_B1, env.action[0])
if args.enemy == "hand":
env.set_robot_action(ID_B1, agent2.select_action(state[ID_B1]))
env.set_robot_action(ID_B2, aux_agent2.select_action(state[ID_B2]))
elif args.enemy == "AC":
env.set_robot_action(
ID_B1,
agent2.select_action(state[ID_B1], mode="max_probability"))
env.set_robot_action(
ID_B2,
aux_agent2.select_action(state[ID_B2], mode="max_probability"))
# Select and perform an action
# 新增红方
state_map = aux_agent.preprocess(state[ID_R2])
a_m, a_t = aux_agent.run_AC(state_map)