GAMMA = 0.99 # reward discount
TARGET_REPLACE_ITER = 999 # target update frequency
DETECTOR_NUM = 0
FIGHTER_NUM = 10
COURSE_NUM = 16
ATTACK_IND_NUM = (DETECTOR_NUM + FIGHTER_NUM) * 2 + 1 # long missile attack + short missile attack + no attack
ACTION_NUM = COURSE_NUM * ATTACK_IND_NUM
LEARN_INTERVAL = TARGET_REPLACE_ITER
BETAS = (0.9, 0.999)
EPS_clip = 0.2
K_epochs = 4
max_timesteps = 300
if __name__ == "__main__":
# create blue agent
blue_agent = Agent()
# get agent obs type
red_agent_obs_ind = 'ppo'
blue_agent_obs_ind = blue_agent.get_obs_ind()
# make env
env = Environment(MAP_PATH, red_agent_obs_ind, blue_agent_obs_ind, render=RENDER)
# get map info
size_x, size_y = env.get_map_size()
red_detector_num, red_fighter_num, blue_detector_num, blue_fighter_num = env.get_unit_num()
# set map info to blue agent
blue_agent.set_map_info(size_x, size_y, blue_detector_num, blue_fighter_num)
red_detector_action = []
fighter_model = ppo.PPOFighter(ACTION_NUM,LR,BETAS,GAMMA,EPS_clip,K_epochs,MAX_EPOCH,LEARN_INTERVAL,max_timesteps)
fighter_model.policy.load_state_dict(torch.load('model/ppo/model_000026500.pkl', map_location='cpu'))
fighter_model.policy_old.load_state_dict(torch.load('model/ppo/model_000026500.pkl', map_location='cpu'))