AUV-PathRL: Trajectory Dataset for Reinforcement Learning in Autonomous Underwater Vehicles

    AUV-PathRL is a trajectory dataset generated by reinforcement learning (RL) models, designed for autonomous underwater vehicle (AUV) path planning and decision-making tasks. This dataset contains high-quality trajectory data collected from RL-trained AUV agents operating in simulated underwater environments. The data includes reinforcement learning information such as states, actions, and reward sequences, as well as path information such as path length and time, making it suitable for training and evaluating various learning-based path planning algorithms.    A key application of AUV-PathRL is training Decision Transformer (DT) models, enabling offline reinforcement learning and trajectory optimization. By leveraging sequential trajectory data, DT can learn to generate optimal navigation paths based on past experiences. This dataset serves as a valuable resource for researchers and engineers developing AI-driven underwater navigation systems, facilitating advancements in AUV autonomy and efficiency.