RailSAC v1.0: Reproducible simulation package for SAC-based dynamic UAV path planning in railway corridor inspection

This archive contains the source code, trained Stable-Baselines3 models, raw episode-level CSV files, representative trajectories, training logs, summary tables, and figure-generation scripts used for the RailSAC railway-inspection UAV path-planning study. The final main experiments compare SAC, TD3, DDPG, A*, and RRT* in constrained railway-corridor scenarios with static obstacles, moving trains, and wind disturbance. Main deep reinforcement learning models were trained for 100,000 environment steps over five independent seeds