Update roadmap

docs/source/index.rst

@@ -29,7 +29,6 @@ Table of Contents
 
    overview/sim/index
    overview/gym/index
-   overview/vla/index
    overview/rl/index
 
 .. toctree::

docs/source/overview/sim/sim_soft_object.md

@@ -23,7 +23,7 @@ Soft bodies require both voxelization and physical attributes.
 
 | Parameter | Type | Default | Description |
 | :--- | :--- | :--- | :--- |
-| `triangle_remesh_resolution` | `int` | `8` | Resolution to remesh the softbody mesh before building physx collision mesh. |
+| `triangle_remesh_resolution` | `int` | `8` | Resolution to remesh the softbody mesh before building physics collision mesh. |
 | `triangle_simplify_target` | `int` | `0` | Simplify mesh faces to target value. |
 | `simulation_mesh_resolution` | `int` | `8` | Resolution to build simulation voxelize textra mesh. |
 | `simulation_mesh_output_obj` | `bool` | `False` | Whether to output the simulation mesh as an obj file for debugging. |

docs/source/resources/roadmap.md

@@ -1,34 +1,36 @@
 # Roadmap
 
-Currently, EmbodiChain is under active development. Our roadmap is as follows:
+Currently, EmbodiChain is under active development. Our roadmap includes the following planned features and enhancements:
 
 - Simulation:
     - Rendering:
         - Improve ray-tracing backend performance and fix some konwn issues.
         - Add a high performance Hybrid rendering backend for better visual quality and speed trade-off.
         - Support a more efficient real-time denoiser.
         - Add a new rasterization backend for basic rendering tasks.
-        - Support 3DGS rendering mode (If we have enough bandwidth).
     - Physics:
         - Improve GPU physics throughput.
-        - Improve soft body simulation stability and add more examples and tasks.
-        - We are also exploring how to integrate [newton physics](https://github.com/newton-physics/newton) into EmbodiChain as an alternative physics backend.
+        - We are working on research and development of next-generation physics backend, supporting high-accuracy simulation, differentiable dynamics, and neural physical models for end-to-end AI integration.
     - Sensors:
-        - Add contact and force sensors with examples.
+        - Add more physical sensors (eg, force sensor) with examples.
     - Motion Generation:
-        - Add more advanced motion generation methods and examples.
+        - Add more advanced motion generation methods with examples.
     - Useful Tools:
         - We are working on USD support for EmbodiChain to enable better asset management and interoperability.
-        - We will release a simple Real2Sim pipeline, which enables automatic task generation from real-world data.
     - Robots Integration:
         - Add support for more robot models (eg: LeRobot, Unitree H1/G1, etc).
 
-- Agents:
-    - Add more Reinforcement Learning examples and environments.
-    - We will release a Modular VLA framework for fast prototyping and training of embodied agents.
-    - We will release a simple online data streaming pipeline for Imitation Learning.
+- Data Pipeline Coming Soon:
+    - We will release a Real2Sim pipeline, which enables automatic data generation and scaling from real-world seeding priors.
+    - We will release an agentic skill generation framework for automated expert trajectory generation.
+    - Add assets and scenes generator and the integration with data pipeline.
 
-- Tasks:
-    - We will release a set of Real2Sim tasks as examples for EmbodiChain.
-    - We will release a set of tableware manipulation tasks for demonstration of data generation pipeline.
+- Models & Training Infrastructure Coming Soon:
+    - We will release a modular VLA framework for fast prototyping and training of embodied agents.
+    - Add online data streaming pipeline for model training.
+
+- Embodied Tasks Coming Soon:
+    - Add more benchmark tasks for EmbodiChain.
+    - Add more tasks with reinforcement learning support.
+    - Add a set of manipulation tasks for demonstration of data generation pipeline.
 

embodichain/lab/gym/utils/gym_utils.py

@@ -126,12 +126,12 @@ def batch(*args: Tuple[Union[np.ndarray, Dict]]):
 
 def to_tensor(array: Array, device: Device | None = None):
     """
-    Maps any given sequence to a torch tensor on the CPU/GPU. If physx gpu is not enabled then we use CPU, otherwise GPU, unless specified
+    Maps any given sequence to a torch tensor on the CPU/GPU. If physics gpu is not enabled then we use CPU, otherwise GPU, unless specified
     by the device argument
 
     Args:
         array: The data to map to a tensor
-        device: The device to put the tensor on. By default this is None and to_tensor will put the device on the GPU if physx is enabled
+        device: The device to put the tensor on. By default this is None and to_tensor will put the device on the GPU if physics is enabled
             and CPU otherwise
 
     """

embodichain/lab/sim/cfg.py

@@ -253,7 +253,7 @@ def from_dict(
 class SoftbodyVoxelAttributesCfg:
     # voxel config
     triangle_remesh_resolution: int = 8
-    """Resolution to remesh the softbody mesh before building physx collision mesh."""
+    """Resolution to remesh the softbody mesh before building physics collision mesh."""
 
     triangle_simplify_target: int = 0
     """Simplify mesh faces to target value. Do nothing if this value is zero."""

embodichain/lab/sim/objects/soft_object.py

@@ -340,7 +340,7 @@ def reset(self, env_ids: Sequence[int] | None = None) -> None:
         local_env_ids = self._all_indices if env_ids is None else env_ids
         num_instances = len(local_env_ids)
 
-        # TODO: set attr for soft body after loading in physx scene
+        # TODO: set attr for soft body after loading in physics scene.
 
         # rest soft body to init_pos
         pos = torch.as_tensor(