Adds parameter to specify number of rerenders after environment reset (#3818)

# Description

<!--
Thank you for your interest in sending a pull request. Please make sure
to check the contribution guidelines.

Link:
https://isaac-sim.github.io/IsaacLab/main/source/refs/contributing.html

💡 Please try to keep PRs small and focused. Large PRs are harder to
review and merge.
-->

Adds a new parameter to ManagerBasedEnv and DirectRLEnv to give users
better control over rerender on reset behaviour. The new parameter
`num_rerenders_on_reset` allows users to explicitly define the number of
re-render steps after an env reset. When using DLSS, this allows for the
elimination of artifacts/ghosting that are present after a single
rendering step.

Add a deprecation warning for the old parameter `rerender_on_reset`.
Functionality of old parameter is preserved.

Updates the existing visuomotor envs to use new rerendering API together
with DLAA for high quality rendering.

Fixes # (issue)

Non-DLSS denoising is not supported on aarch64. There are also future
plans from the rendering team to disable use of non-DLSS antialiasing
for all platforms in the future. This causes an issue for visuomotor
envs which suffer from image ghosting/artifacts when using DLSS. The new
rerendering API allows for users of visuomotor envs to enable DLSS/DLAA
while preserving image integrity.

## Type of change

<!-- As you go through the list, delete the ones that are not
applicable. -->

- New feature (non-breaking change which adds functionality)

## Screenshots

Please attach before and after screenshots of the change if applicable.

<!--
Example:

| Before | After |
| ------ | ----- |
| _gif/png before_ | _gif/png after_ |

To upload images to a PR -- simply drag and drop an image while in edit
mode and it should upload the image directly. You can then paste that
source into the above before/after sections.
-->

## Checklist

- [x] I have read and understood the [contribution
guidelines](https://isaac-sim.github.io/IsaacLab/main/source/refs/contributing.html)
- [x] I have run the [`pre-commit` checks](https://pre-commit.com/) with
`./isaaclab.sh --format`
- [ ] I have made corresponding changes to the documentation
- [x] My changes generate no new warnings
- [ ] I have added tests that prove my fix is effective or that my
feature works
- [x] I have updated the changelog and the corresponding version in the
extension's `config/extension.toml` file
- [x] I have added my name to the `CONTRIBUTORS.md` or my name already
exists there

<!--
As you go through the checklist above, you can mark something as done by
putting an x character in it

For example,
- [x] I have done this task
- [ ] I have not done this task
-->

---------
Signed-off-by: Kelly Guo <kellyg@nvidia.com>
Co-authored-by: Kelly Guo <kellyg@nvidia.com>

docs/source/setup/installation/index.rst

@@ -91,8 +91,6 @@ Other notable limitations with respect to Isaac Lab include...
 
 #. Multi-node training may require direct connections between Spark machines or additional network configurations.
 
-#. :ref:`Isaac Lab Mimic <generating-additional-demonstrations>` data generation and policy inference for visuomotor environments are not supported on DGX Spark due to a lack of non-DLSS image denoiser on aarch64.
-
 #. :ref:`Running Cosmos Transfer1 <running-cosmos>` is not currently supported on the DGX Spark.
 
 Troubleshooting

scripts/imitation_learning/robomimic/train.py

@@ -59,6 +59,7 @@ import argparse
 # Third-party imports
 import gymnasium as gym
 import h5py
+import importlib
 import json
 import numpy as np
 import os
@@ -369,7 +370,18 @@ def main(args: argparse.Namespace):
                 f" Please check that the gym registry has the entry point: '{cfg_entry_point_key}'."
             )
 
-        with open(cfg_entry_point_file) as f:
+        # resolve module path if needed
+        if ":" in cfg_entry_point_file:
+            mod_name, file_name = cfg_entry_point_file.split(":")
+            mod = importlib.import_module(mod_name)
+            if mod.__file__ is None:
+                raise ValueError(f"Could not find module file for: '{mod_name}'")
+            mod_path = os.path.dirname(mod.__file__)
+            config_file = os.path.join(mod_path, file_name)
+        else:
+            config_file = cfg_entry_point_file
+
+        with open(config_file) as f:
             ext_cfg = json.load(f)
             config = config_factory(ext_cfg["algo_name"])
         # update config with external json - this will throw errors if

source/isaaclab/config/extension.toml

 [package]
 
 # Note: Semantic Versioning is used: https://semver.org/
-version = "0.47.9"
+version = "0.47.10"
 
 # Description
 title = "Isaac Lab framework for Robot Learning"

source/isaaclab/docs/CHANGELOG.rst

 Changelog
 ---------
 
+
+0.47.10 (2025-11-06)
+~~~~~~~~~~~~~~~~~~~~
+
+Added
+^^^^^
+
+* Added ``num_rerenders_on_reset`` parameter to ManagerBasedEnvCfg and DirectRLEnvCfg to configure the number
+  of render steps to perform after reset. This enables more control over DLSS rendering behavior after reset.
+
+Changed
+^^^^^^^
+
+* Added deprecation warning for ``rerender_on_reset`` parameter in ManagerBasedEnv and DirectRLEnv.
+
+
 0.47.9 (2025-11-05)
 ~~~~~~~~~~~~~~~~~~~
 

source/isaaclab/isaaclab/envs/direct_rl_env.py

@@ -11,6 +11,7 @@ import inspect
 import math
 import numpy as np
 import torch
+import warnings
 import weakref
 from abc import abstractmethod
 from collections.abc import Sequence
@@ -219,6 +220,20 @@ class DirectRLEnv(gym.Env):
         # video matches the simulation
         self.metadata["render_fps"] = 1 / self.step_dt
 
+        # show deprecation message for rerender_on_reset
+        if self.cfg.rerender_on_reset:
+            msg = (
+                "\033[93m\033[1m[DEPRECATION WARNING] DirectRLEnvCfg.rerender_on_reset is deprecated. Use"
+                " DirectRLEnvCfg.num_rerenders_on_reset instead.\033[0m"
+            )
+            warnings.warn(
+                msg,
+                FutureWarning,
+                stacklevel=2,
+            )
+            if self.cfg.num_rerenders_on_reset == 0:
+                self.cfg.num_rerenders_on_reset = 1
+
         # print the environment information
         print("[INFO]: Completed setting up the environment...")
 
@@ -300,8 +315,9 @@ class DirectRLEnv(gym.Env):
         self.sim.forward()
 
         # if sensors are added to the scene, make sure we render to reflect changes in reset
-        if self.sim.has_rtx_sensors() and self.cfg.rerender_on_reset:
-            self.sim.render()
+        if self.sim.has_rtx_sensors() and self.cfg.num_rerenders_on_reset > 0:
+            for _ in range(self.cfg.num_rerenders_on_reset):
+                self.sim.render()
 
         if self.cfg.wait_for_textures and self.sim.has_rtx_sensors():
             while SimulationManager.assets_loading():
@@ -377,8 +393,9 @@ class DirectRLEnv(gym.Env):
         if len(reset_env_ids) > 0:
             self._reset_idx(reset_env_ids)
             # if sensors are added to the scene, make sure we render to reflect changes in reset
-            if self.sim.has_rtx_sensors() and self.cfg.rerender_on_reset:
-                self.sim.render()
+            if self.sim.has_rtx_sensors() and self.cfg.num_rerenders_on_reset > 0:
+                for _ in range(self.cfg.num_rerenders_on_reset):
+                    self.sim.render()
 
         # post-step: step interval event
         if self.cfg.events:

source/isaaclab/isaaclab/envs/direct_rl_env_cfg.py

@@ -222,6 +222,22 @@ class DirectRLEnvCfg:
       to reflect the latest states from the reset. This comes at a cost of performance as an additional render
       step will be performed after each time an environment is reset.
 
+    .. deprecated:: 2.3.1
+        This attribute is deprecated and will be removed in the future. Please use
+        :attr:`num_rerenders_on_reset` instead.
+
+        To get the same behaviour as setting this parameter to ``True`` or ``False``, set
+        :attr:`num_rerenders_on_reset` to 1 or 0, respectively.
+    """
+
+    num_rerenders_on_reset: int = 0
+    """Number of render steps to perform after reset. Defaults to 0, which means no render step will be performed after reset.
+
+    * When this is 0, no render step will be performed after reset. Data collected from sensors after performing reset will be stale and will not reflect the
+      latest states in simulation caused by the reset.
+    * When this is greater than 0, the specified number of extra render steps will be performed to update the sensor data
+      to reflect the latest states from the reset. This comes at a cost of performance as additional render
+      steps will be performed after each time an environment is reset.
     """
 
     wait_for_textures: bool = True

source/isaaclab/isaaclab/envs/manager_based_env.py

@@ -5,6 +5,7 @@
 
 import builtins
 import torch
+import warnings
 from collections.abc import Sequence
 from typing import Any
 
@@ -190,6 +191,20 @@ class ManagerBasedEnv:
         if self.cfg.export_io_descriptors:
             self.export_IO_descriptors()
 
+        # show deprecation message for rerender_on_reset
+        if self.cfg.rerender_on_reset:
+            msg = (
+                "\033[93m\033[1m[DEPRECATION WARNING] ManagerBasedEnvCfg.rerender_on_reset is deprecated. Use"
+                " ManagerBasedEnvCfg.num_rerenders_on_reset instead.\033[0m"
+            )
+            warnings.warn(
+                msg,
+                FutureWarning,
+                stacklevel=2,
+            )
+            if self.cfg.num_rerenders_on_reset == 0:
+                self.cfg.num_rerenders_on_reset = 1
+
     def __del__(self):
         """Cleanup for the environment."""
         self.close()
@@ -353,8 +368,9 @@ class ManagerBasedEnv:
         self.scene.write_data_to_sim()
         self.sim.forward()
         # if sensors are added to the scene, make sure we render to reflect changes in reset
-        if self.sim.has_rtx_sensors() and self.cfg.rerender_on_reset:
-            self.sim.render()
+        if self.sim.has_rtx_sensors() and self.cfg.num_rerenders_on_reset > 0:
+            for _ in range(self.cfg.num_rerenders_on_reset):
+                self.sim.render()
 
         # trigger recorder terms for post-reset calls
         self.recorder_manager.record_post_reset(env_ids)
@@ -413,8 +429,9 @@ class ManagerBasedEnv:
         self.sim.forward()
 
         # if sensors are added to the scene, make sure we render to reflect changes in reset
-        if self.sim.has_rtx_sensors() and self.cfg.rerender_on_reset:
-            self.sim.render()
+        if self.sim.has_rtx_sensors() and self.cfg.num_rerenders_on_reset > 0:
+            for _ in range(self.cfg.num_rerenders_on_reset):
+                self.sim.render()
 
         # trigger recorder terms for post-reset calls
         self.recorder_manager.record_post_reset(env_ids)

source/isaaclab/isaaclab/envs/manager_based_env_cfg.py

@@ -115,6 +115,22 @@ class ManagerBasedEnvCfg:
       to reflect the latest states from the reset. This comes at a cost of performance as an additional render
       step will be performed after each time an environment is reset.
 
+    .. deprecated:: 2.3.1
+        This attribute is deprecated and will be removed in the future. Please use
+        :attr:`num_rerenders_on_reset` instead.
+
+        To get the same behaviour as setting this parameter to ``True`` or ``False``, set
+        :attr:`num_rerenders_on_reset` to 1 or 0, respectively.
+    """
+
+    num_rerenders_on_reset: int = 0
+    """Number of render steps to perform after reset. Defaults to 0, which means no render step will be performed after reset.
+
+    * When this is 0, no render step will be performed after reset. Data collected from sensors after performing reset will be stale and will not reflect the
+      latest states in simulation caused by the reset.
+    * When this is greater than 0, the specified number of extra render steps will be performed to update the sensor data
+      to reflect the latest states from the reset. This comes at a cost of performance as additional render
+      steps will be performed after each time an environment is reset.
     """
 
     wait_for_textures: bool = True

source/isaaclab/isaaclab/envs/manager_based_rl_env.py

@@ -222,8 +222,9 @@ class ManagerBasedRLEnv(ManagerBasedEnv, gym.Env):
             self._reset_idx(reset_env_ids)
 
             # if sensors are added to the scene, make sure we render to reflect changes in reset
-            if self.sim.has_rtx_sensors() and self.cfg.rerender_on_reset:
-                self.sim.render()
+            if self.sim.has_rtx_sensors() and self.cfg.num_rerenders_on_reset > 0:
+                for _ in range(self.cfg.num_rerenders_on_reset):
+                    self.sim.render()
 
             # trigger recorder terms for post-reset calls
             self.recorder_manager.record_post_reset(reset_env_ids)

source/isaaclab_tasks/config/extension.toml

 [package]
 
 # Note: Semantic Versioning is used: https://semver.org/
-version = "0.11.7"
+version = "0.11.8"
 
 # Description
 title = "Isaac Lab Environments"

source/isaaclab_tasks/docs/CHANGELOG.rst

 Changelog
 ---------
 
+
+0.11.8 (2025-11-06)
+~~~~~~~~~~~~~~~~~~~~
+
+Changed
+^^^^^^^
+
+* Changed to use of ``num_rerenders_on_reset`` and ``DLAA`` in visuomotor imitation learning environments.
+
+
 0.11.7 (2025-10-22)
 ~~~~~~~~~~~~~~~~~~~~
 

source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/pick_place/exhaustpipe_gr1t2_base_env_cfg.py

@@ -323,9 +323,9 @@ class ExhaustPipeGR1T2BaseEnvCfg(ManagerBasedRLEnvCfg):
         self.sim.dt = 1 / 100
         self.sim.render_interval = 2
 
-        # # Set settings for camera rendering
-        self.rerender_on_reset = True
-        self.sim.render.antialiasing_mode = "OFF"  # disable dlss
+        # Set settings for camera rendering
+        self.num_rerenders_on_reset = 3
+        self.sim.render.antialiasing_mode = "DLAA"  # Use DLAA for higher quality rendering
 
         # List of image observations in policy observations
         self.image_obs_list = ["robot_pov_cam"]

source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/pick_place/nutpour_gr1t2_base_env_cfg.py

@@ -359,8 +359,8 @@ class NutPourGR1T2BaseEnvCfg(ManagerBasedRLEnvCfg):
         self.sim.render_interval = 2
 
         # Set settings for camera rendering
-        self.rerender_on_reset = True
-        self.sim.render.antialiasing_mode = "OFF"  # disable dlss
+        self.num_rerenders_on_reset = 3
+        self.sim.render.antialiasing_mode = "DLAA"  # Use DLAA for higher quality rendering
 
         # List of image observations in policy observations
         self.image_obs_list = ["robot_pov_cam"]

source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/stack/config/franka/__init__.py

@@ -54,7 +54,7 @@ gym.register(
     entry_point="isaaclab.envs:ManagerBasedRLEnv",
     kwargs={
         "env_cfg_entry_point": f"{__name__}.stack_ik_rel_visuomotor_env_cfg:FrankaCubeStackVisuomotorEnvCfg",
-        "robomimic_bc_cfg_entry_point": f"{agents.__name__}:robomimic/bc_rnn_image_84.json",
+        "robomimic_bc_cfg_entry_point": f"{agents.__name__}:robomimic/bc_rnn_image_200.json",
     },
     disable_env_checker=True,
 )

source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/stack/config/franka/agents/robomimic/bc_rnn_image_84.json → source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/stack/config/franka/agents/robomimic/bc_rnn_image_200.json

@@ -196,8 +196,8 @@
                 },
                 "obs_randomizer_class": "CropRandomizer",
                 "obs_randomizer_kwargs": {
-                    "crop_height": 76,
-                    "crop_width": 76,
+                    "crop_height": 181,
+                    "crop_width": 181,
                     "num_crops": 1,
                     "pos_enc": false
                 }

source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/stack/config/franka/stack_ik_rel_visuomotor_cosmos_env_cfg.py

@@ -153,8 +153,8 @@ class FrankaCubeStackVisuomotorCosmosEnvCfg(stack_ik_rel_visuomotor_env_cfg.Fran
         )
 
         # Set settings for camera rendering
-        self.rerender_on_reset = True
-        self.sim.render.antialiasing_mode = "OFF"  # disable dlss
+        self.num_rerenders_on_reset = 1
+        self.sim.render.antialiasing_mode = "OFF"
 
         # List of image observations in policy observations
         self.image_obs_list = ["table_cam", "wrist_cam"]

source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/stack/config/franka/stack_ik_rel_visuomotor_env_cfg.py

@@ -204,8 +204,8 @@ class FrankaCubeStackVisuomotorEnvCfg(stack_joint_pos_env_cfg.FrankaCubeStackEnv
         self.scene.wrist_cam = CameraCfg(
             prim_path="{ENV_REGEX_NS}/Robot/panda_hand/wrist_cam",
             update_period=0.0,
-            height=84,
-            width=84,
+            height=200,
+            width=200,
             data_types=["rgb", "distance_to_image_plane"],
             spawn=sim_utils.PinholeCameraCfg(
                 focal_length=24.0, focus_distance=400.0, horizontal_aperture=20.955, clipping_range=(0.1, 2)
@@ -219,8 +219,8 @@ class FrankaCubeStackVisuomotorEnvCfg(stack_joint_pos_env_cfg.FrankaCubeStackEnv
         self.scene.table_cam = CameraCfg(
             prim_path="{ENV_REGEX_NS}/table_cam",
             update_period=0.0,
-            height=84,
-            width=84,
+            height=200,
+            width=200,
             data_types=["rgb", "distance_to_image_plane"],
             spawn=sim_utils.PinholeCameraCfg(
                 focal_length=24.0, focus_distance=400.0, horizontal_aperture=20.955, clipping_range=(0.1, 2)
@@ -231,8 +231,8 @@ class FrankaCubeStackVisuomotorEnvCfg(stack_joint_pos_env_cfg.FrankaCubeStackEnv
         )
 
         # Set settings for camera rendering
-        self.rerender_on_reset = True
-        self.sim.render.antialiasing_mode = "OFF"  # disable dlss
+        self.num_rerenders_on_reset = 3
+        self.sim.render.antialiasing_mode = "DLAA"  # Use DLAA for higher quality rendering
 
         # List of image observations in policy observations
         self.image_obs_list = ["table_cam", "wrist_cam"]

source/isaaclab_tasks/isaaclab_tasks/manager_based/manipulation/stack/config/galbot/stack_rmp_rel_env_cfg.py

@@ -271,8 +271,8 @@ class RmpFlowGalbotLeftArmCubeStackVisuomotorEnvCfg(RmpFlowGalbotLeftArmCubeStac
         )
 
         # Set settings for camera rendering
-        self.rerender_on_reset = True
-        self.sim.render.antialiasing_mode = "OFF"  # disable dlss
+        self.num_rerenders_on_reset = 3
+        self.sim.render.antialiasing_mode = "DLAA"  # Use DLAA for higher quality rendering
 
         # List of image observations in policy observations
         self.image_obs_list = ["ego_cam", "left_wrist_cam", "right_wrist_cam"]