dodox/fancy_gym
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

updates according to changes request

Browse Source
This commit is contained in:
hongyi.zhou
2023-01-27 17:50:14 +01:00
parent 67f684cf14
commit 344c11d67a
5 changed files with 103 additions and 78 deletions
Download Patch File Download Diff File Expand all files Collapse all files
fancy_gym/black_box/black_box_wrapper.py
+39 -39
View File
@@ -159,52 +159,52 @@ class BlackBoxWrapper(gym.ObservationWrapper):
infos = dict()
done = False
if traj_is_valid is False:
if not traj_is_valid:
obs, trajectory_return, done, infos = self.env.invalid_traj_callback(action, position, velocity,
self.return_context_observation)
return self.observation(obs), trajectory_return, done, infos
else:
self.plan_steps += 1
for t, (pos, vel) in enumerate(zip(position, velocity)):
step_action = self.tracking_controller.get_action(pos, vel, self.current_pos, self.current_vel)
c_action = np.clip(step_action, self.env.action_space.low, self.env.action_space.high)
obs, c_reward, done, info = self.env.step(c_action)
rewards[t] = c_reward
if self.verbose >= 2:
actions[t, :] = c_action
observations[t, :] = obs
for k, v in info.items():
elems = infos.get(k, [None] * trajectory_length)
elems[t] = v
infos[k] = elems
if self.render_kwargs:
self.env.render(**self.render_kwargs)
if done or (self.replanning_schedule(self.current_pos, self.current_vel, obs, c_action,
t + 1 + self.current_traj_steps)
and self.plan_steps < self.max_planning_times):
self.condition_pos = pos if self.condition_on_desired else None
self.condition_vel = vel if self.condition_on_desired else None
break
infos.update({k: v[:t+1] for k, v in infos.items()})
self.current_traj_steps += t + 1
self.plan_steps += 1
for t, (pos, vel) in enumerate(zip(position, velocity)):
step_action = self.tracking_controller.get_action(pos, vel, self.current_pos, self.current_vel)
c_action = np.clip(step_action, self.env.action_space.low, self.env.action_space.high)
obs, c_reward, done, info = self.env.step(c_action)
rewards[t] = c_reward
if self.verbose >= 2:
infos['positions'] = position
infos['velocities'] = velocity
infos['step_actions'] = actions[:t + 1]
infos['step_observations'] = observations[:t + 1]
infos['step_rewards'] = rewards[:t + 1]
actions[t, :] = c_action
observations[t, :] = obs
infos['trajectory_length'] = t + 1
trajectory_return = self.reward_aggregation(rewards[:t + 1])
return self.observation(obs), trajectory_return, done, infos
for k, v in info.items():
elems = infos.get(k, [None] * trajectory_length)
elems[t] = v
infos[k] = elems
if self.render_kwargs:
self.env.render(**self.render_kwargs)
if done or (self.replanning_schedule(self.current_pos, self.current_vel, obs, c_action,
t + 1 + self.current_traj_steps)
and self.plan_steps < self.max_planning_times):
self.condition_pos = pos if self.condition_on_desired else None
self.condition_vel = vel if self.condition_on_desired else None
break
infos.update({k: v[:t+1] for k, v in infos.items()})
self.current_traj_steps += t + 1
if self.verbose >= 2:
infos['positions'] = position
infos['velocities'] = velocity
infos['step_actions'] = actions[:t + 1]
infos['step_observations'] = observations[:t + 1]
infos['step_rewards'] = rewards[:t + 1]
infos['trajectory_length'] = t + 1
trajectory_return = self.reward_aggregation(rewards[:t + 1])
return self.observation(obs), trajectory_return, done, infos
def render(self, **kwargs):
"""Only set render options here, such that they can be used during the rollout.
fancy_gym/black_box/raw_interface_wrapper.py
+28
View File
@@ -56,12 +56,30 @@ class RawInterfaceWrapper(gym.Wrapper):
-> Tuple[bool, np.ndarray, np.ndarray]:
"""
Used to preprocess the action and check if the desired trajectory is valid.
Args:
action: a vector instance of the whole action space, includes traj_gen parameters and additional parameters if
specified, else only traj_gen parameters
pos_traj: a vector instance of the raw position trajectory
vel_traj: a vector instance of the raw velocity trajectory
Returns:
validity flag: bool, True if the raw trajectory is valid, False if not
pos_traj: a vector instance of the preprocessed position trajectory
vel_traj: a vector instance of the preprocessed velocity trajectory
"""
return True, pos_traj, vel_traj
def set_episode_arguments(self, action, pos_traj, vel_traj):
"""
Used to set the arguments for env that valid for the whole episode
deprecated, replaced by preprocessing_and_validity_callback
Args:
action: a vector instance of the whole action space, includes traj_gen parameters and additional parameters if
specified, else only traj_gen parameters
pos_traj: a vector instance of the raw position trajectory
vel_traj: a vector instance of the raw velocity trajectory
Returns:
pos_traj: a vector instance of the preprocessed position trajectory
vel_traj: a vector instance of the preprocessed velocity trajectory
"""
return pos_traj, vel_traj
@@ -82,5 +100,15 @@ class RawInterfaceWrapper(gym.Wrapper):
def invalid_traj_callback(self, action: np.ndarray, pos_traj: np.ndarray, vel_traj: np.ndarray) -> Tuple[np.ndarray, float, bool, dict]:
"""
Used to return a artificial return from the env if the desired trajectory is invalid.
Args:
action: a vector instance of the whole action space, includes traj_gen parameters and additional parameters if
specified, else only traj_gen parameters
pos_traj: a vector instance of the raw position trajectory
vel_traj: a vector instance of the raw velocity trajectory
Returns:
obs: artificial observation if the trajectory is invalid, by default a zero vector
reward: artificial reward if the trajectory is invalid, by default 0
done: artificial done if the trajectory is invalid, by default True
info: artificial info if the trajectory is invalid, by default empty dict
"""
return np.zeros(1), 0, True, {}