Safety Gym

Safety Gym is an open-source toolkit for research in safe reinforcement learning. It offers a suite of simulated robotic environments where agents must accomplish tasks (navigation, goal reaching, object interaction) while respecting safety constraints and avoiding hazards. Safety Gym enables benchmarking and development of algorithms that balance task performance with risk-awareness, and it integrates with Gym/Gymnasium and common RL libraries for experimentation and training in safety-critical settings.

Features of Safety Gym:

• Collection of safety-focused simulated environments (navigation, point, car, and block tasks).
• Environments include explicit safety constraints and hazard regions to penalize unsafe behaviour.
• Designed for benchmarking safe reinforcement learning algorithms and reward/constraint formulations.
• Integrates with Gym/Gymnasium APIs and standard RL toolkits for easy experimentation.
• Configurable task parameters, reward shaping, and customizable cost functions.
• Provides example scripts and reference implementations for common safety scenarios.
• Dependencies include physics backends (e.g., MuJoCo) for realistic simulation dynamics.
• Open-source and actively used by the research community for safety- and risk-aware RL work.

To verify the working of Safety Gym in your shell:

 $ sudo su $ sudo apt update $ cd /opt/safety-gym-project $ source ~/safetygym-env/bin/activate $ python -c 'import safety_gymnasium; env = safety_gymnasium.make("SafetyPointGoal1-v0"); env.reset(); print("\nSUCCESS: Safety-Gymnasium environment loaded correctly!")' $ pip show safety-gymnasium 

Disclaimer: Safety Gym is an open-source community project and is not affiliated with, endorsed by, or sponsored by OpenAI or any other company unless explicitly stated by the project maintainers. The toolkit is provided "as is," without any warranty, express or implied. Users assume full responsibility for their usage, and the developers and contributors are not liable for any damages, losses, or consequences resulting from the use of Safety Gym. Users should review and comply with licensing terms, third-party dependency requirements (e.g., MuJoCo), and applicable regulations when using Safety Gym.