Adds clip range for JointAction

source/extensions/omni.isaac.lab/config/extension.toml

@@ -1,7 +1,7 @@
 [package]
 
 # Note: Semantic Versioning is used: https://semver.org/
-version = "0.27.28"
+version = "0.27.29"
 
 # Description
 title = "Isaac Lab framework for Robot Learning"

source/extensions/omni.isaac.lab/docs/CHANGELOG.rst

@@ -1,6 +1,15 @@
 Changelog
 ---------
 
+0.27.29 (2024-12-15)
+~~~~~~~~~~~~~~~~~~~~
+
+Added
+^^^^^
+
+* Added action clip to all :class:`omni.isaac.lab.envs.mdp.actions`.
+
+
 0.27.28 (2024-12-14)
 ~~~~~~~~~~~~~~~~~~~~
 

source/extensions/omni.isaac.lab/omni/isaac/lab/envs/mdp/actions/binary_joint_actions.py

@@ -40,9 +40,10 @@ class BinaryJointAction(ActionTerm):
 
     cfg: actions_cfg.BinaryJointActionCfg
     """The configuration of the action term."""
-
     _asset: Articulation
     """The articulation asset on which the action term is applied."""
+    _clip: torch.Tensor
+    """The clip applied to the input action."""
 
     def __init__(self, cfg: actions_cfg.BinaryJointActionCfg, env: ManagerBasedEnv) -> None:
         # initialize the action term
@@ -83,6 +84,17 @@ def __init__(self, cfg: actions_cfg.BinaryJointActionCfg, env: ManagerBasedEnv)
             )
         self._close_command[index_list] = torch.tensor(value_list, device=self.device)
 
+        # parse clip
+        if self.cfg.clip is not None:
+            if isinstance(cfg.clip, dict):
+                self._clip = torch.tensor([[-float("inf"), float("inf")]], device=self.device).repeat(
+                    self.num_envs, self.action_dim, 1
+                )
+                index_list, _, value_list = string_utils.resolve_matching_names_values(self.cfg.clip, self._joint_names)
+                self._clip[:, index_list] = torch.tensor(value_list, device=self.device)
+            else:
+                raise ValueError(f"Unsupported clip type: {type(cfg.clip)}. Supported types are dict.")
+
     """
     Properties.
     """
@@ -115,6 +127,10 @@ def process_actions(self, actions: torch.Tensor):
             binary_mask = actions < 0
         # compute the command
         self._processed_actions = torch.where(binary_mask, self._close_command, self._open_command)
+        if self.cfg.clip is not None:
+            self._processed_actions = torch.clamp(
+                self._processed_actions, min=self._clip[:, :, 0], max=self._clip[:, :, 1]
+            )
 
     def reset(self, env_ids: Sequence[int] | None = None) -> None:
         self._raw_actions[env_ids] = 0.0

source/extensions/omni.isaac.lab/omni/isaac/lab/envs/mdp/actions/joint_actions.py

@@ -50,6 +50,8 @@ class JointAction(ActionTerm):
     """The scaling factor applied to the input action."""
     _offset: torch.Tensor | float
     """The offset applied to the input action."""
+    _clip: torch.Tensor
+    """The clip applied to the input action."""
 
     def __init__(self, cfg: actions_cfg.JointActionCfg, env: ManagerBasedEnv) -> None:
         # initialize the action term
@@ -94,6 +96,16 @@ def __init__(self, cfg: actions_cfg.JointActionCfg, env: ManagerBasedEnv) -> Non
             self._offset[:, index_list] = torch.tensor(value_list, device=self.device)
         else:
             raise ValueError(f"Unsupported offset type: {type(cfg.offset)}. Supported types are float and dict.")
+        # parse clip
+        if self.cfg.clip is not None:
+            if isinstance(cfg.clip, dict):
+                self._clip = torch.tensor([[-float("inf"), float("inf")]], device=self.device).repeat(
+                    self.num_envs, self.action_dim, 1
+                )
+                index_list, _, value_list = string_utils.resolve_matching_names_values(self.cfg.clip, self._joint_names)
+                self._clip[:, index_list] = torch.tensor(value_list, device=self.device)
+            else:
+                raise ValueError(f"Unsupported clip type: {type(cfg.clip)}. Supported types are dict.")
 
     """
     Properties.
@@ -120,6 +132,11 @@ def process_actions(self, actions: torch.Tensor):
         self._raw_actions[:] = actions
         # apply the affine transformations
         self._processed_actions = self._raw_actions * self._scale + self._offset
+        # clip actions
+        if self.cfg.clip is not None:
+            self._processed_actions = torch.clamp(
+                self._processed_actions, min=self._clip[:, :, 0], max=self._clip[:, :, 1]
+            )
 
     def reset(self, env_ids: Sequence[int] | None = None) -> None:
         self._raw_actions[env_ids] = 0.0

source/extensions/omni.isaac.lab/omni/isaac/lab/envs/mdp/actions/joint_actions_to_limits.py

@@ -44,6 +44,8 @@ class JointPositionToLimitsAction(ActionTerm):
     """The articulation asset on which the action term is applied."""
     _scale: torch.Tensor | float
     """The scaling factor applied to the input action."""
+    _clip: torch.Tensor
+    """The clip applied to the input action."""
 
     def __init__(self, cfg: actions_cfg.JointPositionToLimitsActionCfg, env: ManagerBasedEnv):
         # initialize the action term
@@ -76,6 +78,16 @@ def __init__(self, cfg: actions_cfg.JointPositionToLimitsActionCfg, env: Manager
             self._scale[:, index_list] = torch.tensor(value_list, device=self.device)
         else:
             raise ValueError(f"Unsupported scale type: {type(cfg.scale)}. Supported types are float and dict.")
+        # parse clip
+        if self.cfg.clip is not None:
+            if isinstance(cfg.clip, dict):
+                self._clip = torch.tensor([[-float("inf"), float("inf")]], device=self.device).repeat(
+                    self.num_envs, self.action_dim, 1
+                )
+                index_list, _, value_list = string_utils.resolve_matching_names_values(self.cfg.clip, self._joint_names)
+                self._clip[:, index_list] = torch.tensor(value_list, device=self.device)
+            else:
+                raise ValueError(f"Unsupported clip type: {type(cfg.clip)}. Supported types are dict.")
 
     """
     Properties.
@@ -102,6 +114,10 @@ def process_actions(self, actions: torch.Tensor):
         self._raw_actions[:] = actions
         # apply affine transformations
         self._processed_actions = self._raw_actions * self._scale
+        if self.cfg.clip is not None:
+            self._processed_actions = torch.clamp(
+                self._processed_actions, min=self._clip[:, :, 0], max=self._clip[:, :, 1]
+            )
         # rescale the position targets if configured
         # this is useful when the input actions are in the range [-1, 1]
         if self.cfg.rescale_to_limits:

source/extensions/omni.isaac.lab/omni/isaac/lab/envs/mdp/actions/non_holonomic_actions.py

@@ -11,6 +11,7 @@
 
 import omni.log
 
+import omni.isaac.lab.utils.string as string_utils
 from omni.isaac.lab.assets.articulation import Articulation
 from omni.isaac.lab.managers.action_manager import ActionTerm
 from omni.isaac.lab.utils.math import euler_xyz_from_quat
@@ -59,6 +60,8 @@ class NonHolonomicAction(ActionTerm):
     """The scaling factor applied to the input action. Shape is (1, 2)."""
     _offset: torch.Tensor
     """The offset applied to the input action. Shape is (1, 2)."""
+    _clip: torch.Tensor
+    """The clip applied to the input action."""
 
     def __init__(self, cfg: actions_cfg.NonHolonomicActionCfg, env: ManagerBasedEnv):
         # initialize the action term
@@ -104,6 +107,16 @@ def __init__(self, cfg: actions_cfg.NonHolonomicActionCfg, env: ManagerBasedEnv)
         # save the scale and offset as tensors
         self._scale = torch.tensor(self.cfg.scale, device=self.device).unsqueeze(0)
         self._offset = torch.tensor(self.cfg.offset, device=self.device).unsqueeze(0)
+        # parse clip
+        if self.cfg.clip is not None:
+            if isinstance(cfg.clip, dict):
+                self._clip = torch.tensor([[-float("inf"), float("inf")]], device=self.device).repeat(
+                    self.num_envs, self.action_dim, 1
+                )
+                index_list, _, value_list = string_utils.resolve_matching_names_values(self.cfg.clip, self._joint_names)
+                self._clip[:, index_list] = torch.tensor(value_list, device=self.device)
+            else:
+                raise ValueError(f"Unsupported clip type: {type(cfg.clip)}. Supported types are dict.")
 
     """
     Properties.
@@ -129,6 +142,11 @@ def process_actions(self, actions):
         # store the raw actions
         self._raw_actions[:] = actions
         self._processed_actions = self.raw_actions * self._scale + self._offset
+        # clip actions
+        if self.cfg.clip is not None:
+            self._processed_actions = torch.clamp(
+                self._processed_actions, min=self._clip[:, :, 0], max=self._clip[:, :, 1]
+            )
 
     def apply_actions(self):
         # obtain current heading

source/extensions/omni.isaac.lab/omni/isaac/lab/envs/mdp/actions/task_space_actions.py

@@ -12,6 +12,7 @@
 import omni.log
 
 import omni.isaac.lab.utils.math as math_utils
+import omni.isaac.lab.utils.string as string_utils
 from omni.isaac.lab.assets.articulation import Articulation
 from omni.isaac.lab.controllers.differential_ik import DifferentialIKController
 from omni.isaac.lab.managers.action_manager import ActionTerm
@@ -42,6 +43,8 @@ class DifferentialInverseKinematicsAction(ActionTerm):
     """The articulation asset on which the action term is applied."""
     _scale: torch.Tensor
     """The scaling factor applied to the input action. Shape is (1, action_dim)."""
+    _clip: torch.Tensor
+    """The clip applied to the input action."""
 
     def __init__(self, cfg: actions_cfg.DifferentialInverseKinematicsActionCfg, env: ManagerBasedEnv):
         # initialize the action term
@@ -101,6 +104,17 @@ def __init__(self, cfg: actions_cfg.DifferentialInverseKinematicsActionCfg, env:
         else:
             self._offset_pos, self._offset_rot = None, None
 
+        # parse clip
+        if self.cfg.clip is not None:
+            if isinstance(cfg.clip, dict):
+                self._clip = torch.tensor([[-float("inf"), float("inf")]], device=self.device).repeat(
+                    self.num_envs, self.action_dim, 1
+                )
+                index_list, _, value_list = string_utils.resolve_matching_names_values(self.cfg.clip, self._joint_names)
+                self._clip[:, index_list] = torch.tensor(value_list, device=self.device)
+            else:
+                raise ValueError(f"Unsupported clip type: {type(cfg.clip)}. Supported types are dict.")
+
     """
     Properties.
     """
@@ -138,6 +152,10 @@ def process_actions(self, actions: torch.Tensor):
         # store the raw actions
         self._raw_actions[:] = actions
         self._processed_actions[:] = self.raw_actions * self._scale
+        if self.cfg.clip is not None:
+            self._processed_actions = torch.clamp(
+                self._processed_actions, min=self._clip[:, :, 0], max=self._clip[:, :, 1]
+            )
         # obtain quantities from simulation
         ee_pos_curr, ee_quat_curr = self._compute_frame_pose()
         # set command into controller

source/extensions/omni.isaac.lab/omni/isaac/lab/managers/manager_term_cfg.py

@@ -93,6 +93,9 @@ class for more details.
     debug_vis: bool = False
     """Whether to visualize debug information. Defaults to False."""
 
+    clip: dict[str, tuple] | None = None
+    """Clip range for the action (dict of regex expressions). Defaults to None."""
+
 
 ##
 # Command manager.