Method step

dm_control/rl/control.py:99–127

Updates the environment using the action and returns a `TimeStep`.

(self, action)

        observation=observation)

  def step(self, action):
    """Updates the environment using the action and returns a `TimeStep`."""

    if self._reset_next_step:
      return self.reset()

    self._task.before_step(action, self._physics)
    self._physics.step(self._n_sub_steps)
    self._task.after_step(self._physics)

    reward = self._task.get_reward(self._physics)
    observation = self._task.get_observation(self._physics)
    if self._flat_observation:
      observation = flatten_observation(observation)

    self._step_count += 1
    if self._step_count >= self._step_limit:
      discount = 1.0
    else:
      discount = self._task.get_termination(self._physics)

    episode_over = discount is not None

    if episode_over:
      self._reset_next_step = True
      return dm_env.TimeStep(
          dm_env.StepType.LAST, reward, discount, observation)
    else:
      return dm_env.TimeStep(dm_env.StepType.MID, reward, 1.0, observation)

  def action_spec(self):
    """Returns the action specification for this environment."""