alpyperl

ALPypeRL is an open-source library that bridges AnyLogic simulation models with reinforcement learning frameworks adhering to the OpenAI Gymnasium interface, such as Ray RLlib. This tool enables users to train, deploy, and evaluate RL policies in dynamic AnyLogic simulations while benefiting from its rich visualization capabilities. ALPypeRL supports scaling by running multiple AnyLogic models concurrently and offers robust debugging features, allowing users to identify policy and model issues during training. By replicating runs with controlled seeds, it facilitates detailed debugging and ensures consistent experiments for reliable results.

The library comprises two key components: ALPypeRLConnector, a drag-and-drop agent for AnyLogic models, and alpyperl, a Python package for managing RL policies. ALPypeRL simplifies the integration of RL algorithms into AnyLogic, making it a versatile tool for both academic research and industry applications. With dependencies on frameworks like Ray RLlib, TensorFlow, and PyTorch, it supports a wide range of RL experiments while being compatible with AnyLogic’s free Personal Learning Edition for single-instance experiments. Comprehensive documentation and examples make it accessible to new users and seasoned practitioners alike.