Supports composite observation space with proper min max in manager based env (isaac-sim#2811)

# Description
This PR supports enables ManagerBasedRLEnv to properly support
gymnaisum's composite observation gym.spaces.Dict at term-level with
proper min, max specification. The benefit is that this will give
rl-library a clearer signs how to pre-process the obseravtion data.

Before:
All terms are assigned with `gym.Spaces.Box(min=-np.inf, max=np.inf)`,
one problem with this is that from rl-library side, is that some
rl-libraries network construction depends heavily on observations space
details. RL-library (e.g. sb3) looks at `gym.Spaces.Box.min` and
`gym.Spaces.Box.max` to determine if they need to normalize it at
library side.

After:
this PR utilizes on obs_terms's clip to determine if gym.Spaces.Box
should be stricter. For example, environment obs_term returning
gym.Spaces.Box(min=0, max=255) will be automatically detected by sb3,
and a proper scalling be applied automatically. whereas
gym.Spaces.Box(min=-inf, max=inf) will fail.

while no special case is treated for gym.Spaces.Box(min=-1, max=1), but
this will makes the rl-library easy to figure out that this term is
already clipped, not more transformation needed.


## 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 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
- [x] 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>

Author: ooctipus

source/isaaclab/config/extension.toml

@@ -1,7 +1,7 @@
 [package]
 
 # Note: Semantic Versioning is used: https://semver.org/
-version = "0.40.22"
+version = "0.40.23"
 
 # Description
 title = "Isaac Lab framework for Robot Learning"

source/isaaclab/docs/CHANGELOG.rst

@@ -1,6 +1,17 @@
 Changelog
 ---------
 
+0.40.23 (2025-06-29)
+~~~~~~~~~~~~~~~~~~~~
+
+Added
+^^^^^
+
+* Added MangerBasedRLEnv support for composite gym observation spaces.
+* A test for the composite gym observation spaces in ManagerBasedRLEnv is added to ensure that the observation spaces
+  are correctly configured base on the clip.
+
+
 0.40.22 (2025-07-11)
 ~~~~~~~~~~~~~~~~~~~~
 

source/isaaclab/isaaclab/envs/manager_based_rl_env.py

@@ -333,10 +333,13 @@ def _configure_gym_env_spaces(self):
             if has_concatenated_obs:
                 self.single_observation_space[group_name] = gym.spaces.Box(low=-np.inf, high=np.inf, shape=group_dim)
             else:
-                self.single_observation_space[group_name] = gym.spaces.Dict({
-                    term_name: gym.spaces.Box(low=-np.inf, high=np.inf, shape=term_dim)
-                    for term_name, term_dim in zip(group_term_names, group_dim)
-                })
+                group_term_cfgs = self.observation_manager._group_obs_term_cfgs[group_name]
+                for term_name, term_dim, term_cfg in zip(group_term_names, group_dim, group_term_cfgs):
+                    low = -np.inf if term_cfg.clip is None else term_cfg.clip[0]
+                    high = np.inf if term_cfg.clip is None else term_cfg.clip[1]
+                    self.single_observation_space[group_name] = gym.spaces.Dict(
+                        {term_name: gym.spaces.Box(low=low, high=high, shape=term_dim)}
+                    )
         # action space (unbounded since we don't impose any limits)
         action_dim = sum(self.action_manager.action_term_dim)
         self.single_action_space = gym.spaces.Box(low=-np.inf, high=np.inf, shape=(action_dim,))

source/isaaclab/test/envs/test_manager_based_rl_env_obs_spaces.py

@@ -0,0 +1,57 @@
+# Copyright (c) 2022-2025, The Isaac Lab Project Developers (https://github.com/isaac-sim/IsaacLab/blob/main/CONTRIBUTORS.md).
+# All rights reserved.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+
+"""Test texture randomization in the cartpole scene using pytest."""
+
+from isaaclab.app import AppLauncher
+
+# launch omniverse app
+simulation_app = AppLauncher(headless=True, enable_cameras=True).app
+
+import gymnasium as gym
+import numpy as np
+
+import omni.usd
+import pytest
+
+from isaaclab.envs import ManagerBasedRLEnv
+
+from isaaclab_tasks.manager_based.classic.cartpole.cartpole_camera_env_cfg import (
+    CartpoleDepthCameraEnvCfg,
+    CartpoleRGBCameraEnvCfg,
+)
+from isaaclab_tasks.manager_based.locomotion.velocity.config.anymal_c.rough_env_cfg import AnymalCRoughEnvCfg
+
+
+@pytest.mark.parametrize(
+    "env_cfg_cls",
+    [CartpoleRGBCameraEnvCfg, CartpoleDepthCameraEnvCfg, AnymalCRoughEnvCfg],
+    ids=["RGB", "Depth", "RayCaster"],
+)
+@pytest.mark.parametrize("device", ["cpu", "cuda"])
+def test_obs_space_follows_clip_contraint(env_cfg_cls, device):
+    """Ensure curriculum terms apply correctly after the fallback and replacement."""
+    # new USD stage
+    omni.usd.get_context().new_stage()
+
+    # configure the cartpole env
+    env_cfg = env_cfg_cls()
+    env_cfg.scene.num_envs = 2  # keep num_envs small for testing
+    env_cfg.observations.policy.concatenate_terms = False
+    env_cfg.sim.device = device
+
+    env = ManagerBasedRLEnv(cfg=env_cfg)
+    for group_name, group_space in env.observation_space.spaces.items():
+        for term_name, term_space in group_space.spaces.items():
+            term_cfg = getattr(getattr(env_cfg.observations, group_name), term_name)
+            low = -np.inf if term_cfg.clip is None else term_cfg.clip[0]
+            high = np.inf if term_cfg.clip is None else term_cfg.clip[1]
+            assert isinstance(
+                term_space, gym.spaces.Box
+            ), f"Expected Box space for {term_name} in {group_name}, got {type(term_space)}"
+            assert np.all(term_space.low == low)
+            assert np.all(term_space.high == high)
+
+    env.close()