Open-access physical robotics environment for real-world reinforcement learning benchmarking and research

Benchmark your algorithm in a real world environment for free!

Sign up

Reinforcement learning has made remarkable progress in training agents for games and simulated environments. But the ultimate goal is to learn to train physical agents to operate in the real world. OffWorld Gym provides the platform and the benchmark to figure out the specifics of real-world learning and run your algorithms in the wild!

  • Install the library
  • Book your time to run the experiment
  • Train your agent!

Run the minimal example
Read the paper
Check out documentation
Contribute on GitHub