Strategy DTMCs from Q-learning Algorithms

该数据集包含了在随机环境中，特别是OpenAI Gym的冰湖环境里，通过Q-learning算法生成的最优和次优策略的策略DTMCs。该数据集不仅包括正确的LTL任务，也包括错误的LTL任务，每个任务生成了10个正面的最优策略DTMCs和10个负面的最优策略DTMCs，这有助于评估PriTL学习算法。数据集的规模达到了累计状态空间的10^3数量级，其任务是推断策略DTMCs中的简洁LTL规格说明。

This dataset contains strategy DTMCs of optimal and suboptimal policies generated via the Q-learning algorithm in random environments, especially the Frozen Lake environment from OpenAI Gym. It includes both correct and incorrect LTL tasks, with 10 positive optimal policy DTMCs and 10 negative optimal policy DTMCs generated for each task, which facilitates the evaluation of the PriTL learning algorithm. The cumulative state space of the dataset is on the order of 10^3, and its task is to infer concise LTL specifications from the strategy DTMCs.