safegym 0.12

Gymnasium Compatible Safe RL Environments

SafeGym

Implementation of satellite environments and other environments to explore SafeRL

SafeGym

SafeGym is a Gymnasium environment coupled with tools aimed at facilitating Safe Reinforcement Learning (SafeRL) research and development.

Features

• Customizable Environment: Create a variety of satellite chasing scenarios with customizable starting states and noise.
• Underactuated and Fully Actuated Dynamics: Simulate real-world control dynamics with options for both underactuated and fully actuated control systems.
• Reward Shaping: Built-in reward shaping functionality to guide the learning process towards safe and effective solutions.
• Truncation and Termination: Control the episode flow with truncation and termination conditions to manage the agent's learning experience.
• Visualization Tools: Render the environment and visualize the agent's interactions and performance over time.
• Testing Utilities: Validate the environment and control algorithms with a set of provided testing functions.

Getting Started

Clone the repository to your local machine:

git clone https://github.com/spbisc97/SafeGym.git
cd SafeGym

Ensure you have the necessary dependencies installed. The primary dependencies include:

• gymnasium
• numpy
• matplotlib

Usage

Install

pip install -e .

Run a simple experiment

import safegym
import gymnasium as gym

env = gym.make('Satellite-SE',render_mode="human")
observation,info = env.reset()
DONE=False

while not DONE
    action = env.action_space.sample()
    observation, reward, term, trunc, info = env.step(action)
    DONE = term or trunc

env.close()