Hierarchical MDP Benchmarks

该数据集包含了一系列基准测试，涵盖了网络、作业调度器和机器人等领域，用于评估在分层马尔可夫决策过程（MDP）中提出的抽象精化方法的表现性能。该数据集还包括了具有不同状态数量的分层MDP实例、非平凡分割以及在不确定的宏观MDP和子MDP中的状态/动作。这些基准测试在平坦表示中至少拥有10^7个状态。该数据集的任务是对分层马尔可夫决策过程进行验证和分析。

This dataset comprises a suite of benchmark tests covering networking, job schedulers, robotics and other related domains, aimed at evaluating the performance of the proposed abstraction refinement approach in hierarchical Markov Decision Processes (MDPs). It further includes hierarchical MDP instances with varying state counts, non-trivial partitions, as well as states and actions within uncertain macro-MDPs and sub-MDPs. These benchmarks feature at least 10^7 states when represented in a flat form. The core task of this dataset is to conduct validation and analysis of hierarchical Markov Decision Processes.