DQN001.py 文件源码

def __init__(self, panel, x, y, epsilon = 0.99):
        super(Agent, self).__init__(
            x, y, wx.Colour(112,146,190)
        )
        self.dir_Angle = math.pi/4
        self.speed     = 5

        self.pos_x_max, self.pos_y_max = panel.GetSize()

        self.eyes = [ EYE(i) for i in range(0, NUM_EYES)]

        self.actions = [-math.pi/16, -math.pi/8, 0.0, math.pi/8, math.pi/16]
        self.prevActions = np.zeros_like(self.actions)

        # DQN Model
        self.model = Q()
        self.optimizer = optimizers.Adam()
        self.optimizer.setup(self.model)

        self.epsilon = epsilon

        self.eMem = np.array([],dtype = np.float32)
        self.memPos = 0 
        self.memSize = 30000

        self.batch_num = 30
        self.gamma = 0.7
        self.loss = 0.0

        self.State = SState()
        self.prevState = np.ones((1,STATE_DIM))