car_on_hill.py 文件源码

def step(self, action):
        action = self._discrete_actions[action[0]]
        sa = np.append(self._state, action)
        new_state = odeint(self._dpds, sa, [0, self._dt])

        self._state = new_state[-1, :-1]

        if self._state[0] < -self.max_pos or \
                np.abs(self._state[1]) > self.max_velocity:
            reward = -1
            absorbing = True
        elif self._state[0] > self.max_pos and \
                np.abs(self._state[1]) <= self.max_velocity:
            reward = 1
            absorbing = True
        else:
            reward = 0
            absorbing = False

        return self._state, reward, absorbing, {}