The Data
Foundation
for Real-World RL
Discover, access, and contribute high-quality real-world reinforcement learning datasets. Accelerate sim-to-real transfer and on-robot policy training with standardized, curated data.
Everything You Need for Real-World RL
A unified platform designed specifically for real-world reinforcement learning data — from collection to training.
Large-Scale Dataset Hosting
Store and serve real-world trajectory datasets from gigabytes to terabytes with optimized delivery pipelines and Alibaba Cloud OSS backend.
Real-World RL Data Support
Native support for the LeRobot format used by Evo-RL. All datasets are directly loadable into modern RL training frameworks without any conversion.
Efficient Data Pipeline
Stream datasets directly into your RL training scripts. Built-in compatibility with LeRobot, PyTorch DataLoader, and Hugging Face Datasets.
Trajectory Visualization
Explore robot trajectories, joint states, reward signals, and sensor readings with interactive visualization tools in your browser.
Open & Standardized
All datasets follow standardized metadata schemas. Clear licensing (CC BY, MIT, Apache) for academic research and commercial use.
Community Driven
Upload your own real-world datasets, submit benchmarks, and collaborate with the global robot learning research community.
See Evo-RL in Action
Real-world trajectory rollouts from our Evo-RL training framework — showcasing reward-based success and failure cases.
Bridging the Sim-to-Real Gap with Data
Real-world data scarcity is one of the biggest bottlenecks in robot learning research. EvoData provides a unified, standardized, and accessible platform to accelerate on-robot RL by making high-quality real-world trajectory data freely available.
- LeRobot-compatible format — collected directly via Evo-RL
- One-click dataset access from RL training scripts
- Community-driven data curation and benchmarks
- Future: real-time upload, streaming and on-robot data collection tools
Ready to Explore?
Browse our curated collection of real-world RL datasets and start accelerating your research today.