健身房环境的观察空间能否同时拥有形象和其他价值?

时间:2021-02-22 16:49:43

标签: python python-3.x keras openai-gym keras-rl

所以我有一个生成的迷宫让我们说(为简单起见)它是一个 10 x 10 大小的迷宫,以 200 x 200 RGB 像素表示。所以我知道,如果我只想拥有图像,那么它将是 Box(0.0, 255.0, (200, 200, 3), float32) 但是我应该怎么做才能拥有值(在下面的代码中显示)和图像?另外,在我的 step 函数中,我应该如何返回状态?

class MazeEnv(Env):
    def __init__(self):
        # Possible actions taken: Up, Down, Left, Right
        self.action_space = Discrete(4)

        # low and high values for each states
        # Num    Observation                Min                     Max
        # 0      Agent Position X           -Inf                    Inf
        # 1      Agent Position Y           -Inf                    Inf
        # 2      Agent distance from goal X -Inf                    Inf
        # 3      Agent distance from goal X -Inf                    Inf
        # 4      IMAGE OF MAZE              ?                       ?
        
        high = np.array([np.finfo(np.float64).max,
                         np.finfo(np.float64).max,
                         np.finfo(np.float64).max,
                         np.finfo(np.float64).max],
                        dtype=np.float64)
        self.observation_space = Box(low=-high, high=high, dtype=np.float64)

        # Current state of the agent
        self.stateX = 0
        self.stateY = 0

        # The amount of moves remaining for the agent
        self.movesRemaining = 120

0 个答案:

没有答案