Lecture 2: Playing OpenAI GYM Games by Sung Kim

link: https://www.youtube.com/watch?v=xgoO54qN4lY&list=PLlMkM4tgfjnKsCWav-Z2F-MMFRx-2gMGG&index=2

Frozen Lake World

S: start(시작점)

F: frozen(얼음이 있는 곳)

H: hole(구멍)

G: goal(도착점)

• (1) Agent가 특정 Action을 취함
• (2) Env가 Action에 대한 State와 Reward를 제공

 

OpenAi Gym

link: https://gym.openai.com/

Gym Documentation

env = gym.make("FrozenLake-v0") # 환경 생성(이름)
observation = env.reset() # 환경 초기화
for _ in range(1000):
  env.render() # 환경을 화면으로 출력
  action = env.action_space.sample() # your agent here (this takes random actions)
  observation, reward, done, info = env.step(action) # action의 결과 반환 done: true/false

import gym
from gym.envs.registration import register
import sys, tty, termios


class _Getch:
  def __call__(self):
    fd = sys.stdin.fileno()
    old_settings = termios.tcgetattr(fd)
    try:
      tty.setraw(sys.stdin.fileno())
      ch = sys.stdin.read(3)
    finally:
      termios.tcsetattr(fd, termios.TCSADRAIN, old_settings)
    return ch


inkey = _Getch()

# MACROS
LEFT = 0
DOWN = 1
RIGHT = 2
UP = 3

# Key mapping
arrow_keys = {
    '\x1b[A': UP,
    '\x1b[B': DOWN,
    '\x1b[C': RIGHT,
    '\x1b[D': LEFT
}

# Register Frozen with is_slippery False
register(
    id='FrozenLake-v3',
    entry_point='gym.envs.toy_text:FrozenLakeEnv',
    kwargs={'map_name': '4x4', 'is_slippery': False}
)

env = gym.make('FrozenLake-v3')
state = env.reset()
env.render()  # Show the inital board

while True:
    # Choose an action from keyboard
    key = inkey()
    if key not in arrow_keys.keys():
        print("Game aborted!")
        break

    action = arrow_keys[key]
    state, reward, done, info = env.step(action)
    env.render()  # Show the board after action
    print("State: ", state, "Action: ", action, "Reward: ", reward, "Info: ", info)

    if done:
        print("Finished with reward", reward)
        break