liveplex/robogate-env

RoboGate / PickPlace-68-v0

Industrial pick-and-place validation benchmark for Vision-Language-Action (VLA) models. 68 scenarios across nominal (20), edge-case (15), adversarial (10), and domain-randomized (23) categories. Franka Panda arm in NVIDIA Isaac Sim 5.1.

Install

pip install robogate-env lerobot>=0.5.1
# or directly from source:
pip install -e git+https://github.com/liveplex-cpu/robogate#subdirectory=web/lerobot-envhub

Load

import robogate_env  # triggers gym registration
import gymnasium as gym

env = gym.make("robogate/PickPlace-68-v0")
obs, info = env.reset(seed=42)
# ... policy loop ...

Scenarios

• nominal x 20
• edge_cases x 15
• adversarial x 10
• domain_randomization x 23

Citation

@misc{liveplex2026robogate,
  title  = {RoboGate: A Validation Gate for Industrial VLA Policy Deployment},
  author = {Liveplex Research Team},
  year   = {2026},
  url    = {https://robogate.io}
}

• Repo: https://github.com/liveplex-cpu/robogate
• Dataset: https://huggingface.co/datasets/liveplex/robogate-failure-dictionary
• EnvHub id: liveplex/robogate-env (version 0.1.0)