Frequently Asked Questions#

This page answers common questions about RoboVerse. For simulator-specific issues, see the Troubleshooting section.

General Questions#

What is RoboVerse?#

RoboVerse is a unified platform for scalable and generalizable robot learning. See the project page and paper for details.

What simulators are supported?#

See the Support Matrix for the full list of supported simulators and features.

What robots are available?#

See Robots Dataset for the complete list of available robots.

Installation#

What are the system requirements?#

Minimum Requirements:

  • Ubuntu 20.04+ or macOS 12+

  • Python 3.10+

  • 16GB RAM

  • NVIDIA GPU with CUDA 11.8+ (for GPU features)

Recommended:

  • Ubuntu 22.04

  • Python 3.10

  • 32GB RAM

  • NVIDIA RTX 3090 or better

How do I install RoboVerse?#

# Clone the repository
git clone https://github.com/RoboVerseOrg/RoboVerse.git
cd RoboVerse

# Create conda environment
conda create -n roboverse python=3.10
conda activate roboverse

# Install MetaSim
pip install -e .

See the Installation Guide for detailed instructions.

Can I use Docker?#

Yes! We provide official Docker images:

docker pull roboverseorg/roboverse:latest
docker run -it --gpus all roboverseorg/roboverse:latest

See Docker Guide for more details.

How do I install optional dependencies?#

Different features require additional packages:

# For Isaac Sim support
pip install -e ".[isaacsim]"

# For motion planning
pip install -e ".[curobo]"

# For imitation learning
pip install -e ".[il]"

# For reinforcement learning
pip install -e ".[rl]"

See Advanced Installation for all options.

Usage#

How do I run my first simulation?#

from metasim.cfg.scenario import ScenarioCfg
from metasim.sim.mujoco_handler import MujocoHandler
from roboverse_pack.robots.franka import FrankaCfg

# Configure scenario
scenario = ScenarioCfg(
    robots=[FrankaCfg()],
    num_envs=1,
)

# Create and launch handler
handler = MujocoHandler(scenario)
handler.launch()

# Step simulation
for _ in range(1000):
    handler.step()

See Quick Start Tutorials for more examples.

How do I switch between simulators?#

Simply change the handler class while keeping the same scenario:

from metasim.sim.mujoco_handler import MujocoHandler
from metasim.sim.sapien_handler import SapienHandler
from metasim.sim.isaacsim_handler import IsaacSimHandler

# Same scenario works with any simulator
handler = MujocoHandler(scenario)    # or
handler = SapienHandler(scenario)    # or
handler = IsaacSimHandler(scenario)

How do I run parallel environments?#

scenario = ScenarioCfg(
    robots=[FrankaCfg()],
    num_envs=1024,  # Run 1024 environments in parallel
)
handler = MujocoHandler(scenario)
handler.launch()

GPU-accelerated simulators (Isaac Sim, MJX, Genesis) provide best performance for large-scale parallelism.

How do I train a policy?#

Reinforcement Learning:

python roboverse_learn/rl/train_ppo.py --task pick_cube --robot franka

Imitation Learning:

python roboverse_learn/il/train_dp.py --task pick_cube --data_path /path/to/demos

See RoboVerse Learn for algorithm-specific guides.

Troubleshooting#

I get “libGL error” on Linux#

Install the required OpenGL libraries:

conda install -c conda-forge libstdcxx-ng
# or
sudo apt-get install libgl1-mesa-glx libegl1

MuJoCo shows “egl error”#

Install EGL libraries:

sudo apt-get install libegl1 libgl1-mesa-glx

MJX shows “DNN library initialization failed”#

Install a compatible cuDNN version:

pip install nvidia-cudnn-cu12==9.10.2.21

Isaac Sim fails to start#

Ensure you have:

  1. NVIDIA driver version 525+

  2. Isaac Sim 2023.1.1 or later installed

  3. Sufficient GPU memory (8GB+ recommended)

See Isaac Sim Troubleshooting for more solutions.

How do I report a bug?#

  1. Check existing GitHub Issues

  2. If not found, create a new issue with:

    • System information (OS, Python version, GPU)

    • Steps to reproduce

    • Full error traceback

    • Expected vs actual behavior

Contributing#

How can I contribute?#

We welcome contributions! See our guides:

How do I add a new robot?#

  1. Prepare robot assets (URDF/MJCF/USD)

  2. Create a configuration class in roboverse_pack/robots/

  3. Add tests for the new robot

  4. Submit a pull request

See Contributing New Robot for details.

How do I add a new task?#

  1. Create a task class inheriting from BaseTaskEnv

  2. Define reward function and success criteria

  3. Configure default scenario with appropriate robots/objects

  4. Add to the task registry

See Task System for the task API.

Community#

Where can I get help?#

How do I cite RoboVerse?#

@misc{geng2025roboverse,
    title={RoboVerse: Towards a Unified Platform, Dataset and Benchmark 
           for Scalable and Generalizable Robot Learning},
    author={Haoran Geng and Feishi Wang and Songlin Wei and ...},
    year={2025},
    eprint={2504.18904},
    archivePrefix={arXiv},
}