Artificial Intelligence for Telescope Observation Scheduling

Scheduling observations for modern astronomical surveys requires balancing competing scientific goals, variable environmental conditions, and limited telescope resources. Traditional scheduling methods often rely on bespoke, hand-tuned algorithms that can limit observing efficiency and require substantial human oversight. I present an artificial intelligence (AI) framework for optimizing telescope observation scheduling using reinforcement learning (RL). We leverage a large historical dataset from the Dark Energy Camera (DECam) to train and evaluate RL agents capable of balancing multi-objective rewards and dynamically adapting to observing conditions. Preliminary experiments show that AI-based scheduling can reduce manual intervention while maintaining or improving observing efficiency. I discuss the ongoing work to train and deploy these agents on real telescope control systems, thereby illustrating a path toward integrating AI into telescope operations and survey planning.