Scikit-Decide AI Framework for RL, Auto Planning and Scheduling (airbus.github.io)

The Scikit-Decide AI framework has been launched, aiming to enhance Reinforcement Learning (RL), automated planning, and scheduling. Developed at Airbus AI Research with contributions from projects like ANITI and TUPLES, this open-source framework allows users to describe their decision-making problems and automatically match them with suitable solvers. Its significance lies in its extensive catalog of supported domains and solvers, which fosters community involvement and innovation in AI problem-solving. Scikit-Decide boasts versatile features, including compatibility with Gym environments for RL and various planning languages such as PDDL and RDDL. It supports state-of-the-art solvers and algorithms, including RL frameworks like Ray's RLlib and Stable-Baselines3, and offers advanced capabilities like action masking and graph observation adaptations via GNNs. With a wide range of built-in search and planning solvers, such as A*, Monte Carlo Tree Search, and hybrid approaches, it addresses diverse problem-solving scenarios, from flight planning to scheduling. This framework not only streamlines AI development but also opens avenues for research and application across multiple domains.