Upgrade dexsim to v0.3.8 (#53)

Author: yuecideng

README.md

@@ -4,7 +4,7 @@
 
 [![License](https://img.shields.io/github/license/DexForce/EmbodiChain)](LICENSE)
 [![GitHub Pages](https://img.shields.io/badge/GitHub%20Pages-docs-blue?logo=github&logoColor=white)](https://dexforce.github.io/EmbodiChain/introduction.html)
-[![Python](https://img.shields.io/badge/python-3.10-blue.svg)](https://docs.python.org/3/whatsnew/3.10.html)
+[![Python](https://img.shields.io/badge/python-3.10%20|%203.11-blue.svg)](https://docs.python.org/3/whatsnew/3.10.html)
 [![Version](https://img.shields.io/badge/version-0.0.1-blue.svg)](https://github.com/DexForce/EmbodiChain/releases)
 
 ---

docs/source/features/workspace_analyzer/workspace_analyzer.md

@@ -25,7 +25,6 @@ from embodichain.lab.sim.utility.workspace_analyzer import (
 
 # Setup simulation
 sim = SimulationManager(SimulationManagerCfg(headless=False, sim_device="cpu"))
-sim.build_multiple_arenas(1)
 
 # Add robot
 robot = sim.add_robot(DexforceW1Cfg.from_dict({
@@ -169,7 +168,6 @@ from embodichain.lab.sim.utility.workspace_analyzer.configs import Visualization
 
 # Setup simulation
 sim = SimulationManager(SimulationManagerCfg(headless=False, sim_device="cpu"))
-sim.build_multiple_arenas(1)
 
 # Add robot
 robot = sim.add_robot(DexforceW1Cfg.from_dict({

docs/source/quick_start/install.md

@@ -4,22 +4,21 @@
 
 The following minimum system requirements are recommended to run EmbodiChain reliably. These are the tested configurations during development — other Linux distributions and versions may work but are not officially supported.
 
-- Operating System: Linux (x86_64)
-    - Recommended distributions: Ubuntu 20.04 LTS or Ubuntu 22.04 LTS
+- Operating System: 
+    - Linux (x86_64): Ubuntu 20.04+
 
 - NVIDIA GPU and drivers:
-    - Hardware: NVIDIA GPU with compute capability 7.0 or higher (e.g., RTX 20 series, RTX 30 series, A100, etc.)
-    - NVIDIA driver: 535 or higher (recommended 570)
-    - CUDA Toolkit: any of 11.8 — 12.8 (we test primarily with 11.8 and 12.x)
+    - Hardware: NVIDIA GPU with compute capability 7.0 or higher
+    - NVIDIA Driver: 535 or higher (recommended 570)
+
 
 - Python:
-    - Supported Python versions:
-        - Python 3.10
-    - Use a virtual environment (venv, virtualenv, or conda) to isolate dependencies
+    - 3.10
+    - 3.11
 
 Notes:
 
-- Ensure your NVIDIA driver and CUDA toolkit versions are compatible with your chosen PyTorch wheel.
+- Ensure your NVIDIA driver is compatible with your chosen PyTorch wheel.
 - We recommend installing PyTorch from the official PyTorch instructions for your CUDA version: https://pytorch.org/get-started/locally/
 
 ---

docs/source/resources/robot/cobotmagic.md

@@ -39,9 +39,8 @@ CobotMagic is a versatile dual-arm collaborative robot developed by AgileX Robot
 from embodichain.lab.sim import SimulationManager, SimulationManagerCfg
 from embodichain.lab.sim.robots import CobotMagicCfg
 
-config = SimulationManagerCfg(headless=False, sim_device="cpu")
+config = SimulationManagerCfg(headless=False, sim_device="cpu", num_envs=2)
 sim = SimulationManager(config)
-sim.build_multiple_arenas(2)  # Supports parallel simulation in multiple arenas
 sim.set_manual_update(False)
 
 robot = sim.add_robot(cfg=CobotMagicCfg().from_dict({}))

docs/source/tutorial/create_scene.rst

@@ -31,15 +31,13 @@ Command-line arguments are parsed using ``argparse`` to allow for easy customiza
 .. literalinclude:: ../../../scripts/tutorials/sim/create_scene.py
    :language: python
    :start-at: # Parse command line arguments
-   :end-at: sim.build_multiple_arenas(args.num_envs, space=3.0)
+   :end-at: sim = SimulationManager(sim_cfg)
 
 There are two kinds of physics mode in :class:`SimulationManager`:
 
 - `manual`: The physics updates only when the user calls the :meth:`SimulationManager.update` function. This mode is used for robot learning tasks where precise control over simulation steps is required. Enabled by setting :meth:`SimulationManager.set_manual_update` to True.
 - `auto`: The physics updates in a standalone thread, which enable asynchronous rendering and physics stepping. This mode is suitable for visualizations and demos for digital twins applications. This is the default mode.
 
-If `num_envs` is greater than 1, :meth:`SimulationManager.build_multiple_arenas` should be used to create multiple simulation arenas.
-
 Adding objects to the scene
 ---------------------------
 

docs/source/tutorial/create_softbody.rst

@@ -32,8 +32,6 @@ The first step is to configure the simulation environment. This is done using th
    :start-at: # Configure the simulation
    :end-at:     print("[INFO]: Scene setup complete!")
 
-If ``num_envs`` is greater than 1, :meth:`SimulationManager.build_multiple_arenas` should be used to create multiple simulation arenas.
-
 Adding a soft body to the scene
 -------------------------------
 

docs/source/tutorial/rl.rst

@@ -182,7 +182,7 @@ To start training, run:
 
 .. code-block:: bash
 
-   python embodichain/agents/rl/train.py --config configs/agents/rl/push_cube/train_config.json
+   python -m embodichain.agents.rl.train --config configs/agents/rl/push_cube/train_config.json
 
 Outputs
 -------

embodichain/lab/gym/envs/base_env.py

@@ -109,6 +109,7 @@ def __init__(
             self.sim_cfg = SimulationManagerCfg(headless=True)
         else:
             self.sim_cfg = self.cfg.sim_cfg
+            self.sim_cfg.num_envs = self.num_envs
 
         if self.cfg.seed is not None:
             self.cfg.seed = set_seed(self.cfg.seed)
@@ -208,8 +209,6 @@ def _setup_scene(self, **kwargs):
         logger.log_info(
             f"Initializing {self.num_envs} environments on {self.sim_cfg.sim_device}."
         )
-        if self.num_envs > 1:
-            self.sim.build_multiple_arenas(self.num_envs)
 
         self.robot = self._setup_robot(**kwargs)
         if self.robot is None:

embodichain/lab/sim/objects/articulation.py

@@ -797,7 +797,6 @@ def set_local_pose(
             # we should keep `pose_` life cycle to the end of the function.
             pose_ = torch.cat((quat, xyz), dim=-1)
             indices = self.body_data.gpu_indices[local_env_ids]
-            torch.cuda.synchronize(self.device)
             self._ps.gpu_apply_root_data(
                 data=pose_,
                 gpu_indices=indices,
@@ -981,7 +980,6 @@ def set_qpos(
             indices = self.body_data.gpu_indices[local_env_ids]
             qpos_set = self.body_data._qpos[local_env_ids]
             qpos_set[:, local_joint_ids] = qpos
-            torch.cuda.synchronize(self.device)
             self._ps.gpu_apply_joint_data(
                 data=qpos_set,
                 gpu_indices=indices,
@@ -1045,7 +1043,6 @@ def set_qvel(
                 self.body_data.qvel
                 qvel_set = self.body_data._qvel[local_env_ids]
                 qvel_set[:, joint_ids] = qvel
-            torch.cuda.synchronize(self.device)
             self._ps.gpu_apply_joint_data(
                 data=qvel_set,
                 gpu_indices=indices,
@@ -1086,7 +1083,6 @@ def set_qf(
                 self.body_data.qf
                 qf_set = self.body_data._qf[local_env_ids]
                 qf_set[:, joint_ids] = qf
-            torch.cuda.synchronize(self.device)
             self._ps.gpu_apply_joint_data(
                 data=qf_set,
                 gpu_indices=indices,
@@ -1167,13 +1163,11 @@ def clear_dynamics(self, env_ids: Sequence[int] | None = None) -> None:
                 (len(local_env_ids), self.dof), dtype=torch.float32, device=self.device
             )
             indices = self.body_data.gpu_indices[local_env_ids]
-            torch.cuda.synchronize(self.device)
             self._ps.gpu_apply_joint_data(
                 data=zeros,
                 gpu_indices=indices,
                 data_type=ArticulationGPUAPIWriteType.JOINT_VELOCITY,
             )
-            torch.cuda.synchronize(self.device)
             self._ps.gpu_apply_joint_data(
                 data=zeros,
                 gpu_indices=indices,

embodichain/lab/sim/robots/cobotmagic.py

@@ -207,9 +207,8 @@ def build_pk_serial_chain(
 
     torch.set_printoptions(precision=5, sci_mode=False)
 
-    config = SimulationManagerCfg(headless=False, sim_device="cuda")
+    config = SimulationManagerCfg(headless=False, sim_device="cuda", num_envs=2)
     sim = SimulationManager(config)
-    sim.build_multiple_arenas(2)
 
     config = {
         "init_pos": [0.0, 0.0, 1.0],