rtgym

Real-Time Gym (rtgym) is a framework designed for implementing Gymnasium environments with real-time constraints, which is particularly useful for applications in robotics or real-time video games. It allows reinforcement learning algorithms to operate within environments that require precise control of time steps, such as controlling robot arms or interacting with simulated environments. The system is built to manage the timing of actions and observations transparently to the user, ensuring that time steps are elastically constrained to a specified duration. This enables real-time Markov Decision Processes (MDPs) while accommodating delays in action execution or observation retrieval.