Synthetic Action Costs for Planning

该数据集包含了针对不同规划问题生成的特征与动作成本配对数据，其中特征采用多变量高斯分布进行生成，而动作成本则通过合成生成过程得出。此外，数据集还包括了从rovers领域和运输问题中生成的实例，以及为评估决策焦点学习方法有效性而创造的合成动作成本。该数据集规模分为小型和大型实例，以供实验使用。其任务是对带有动作成本预测的自动规划。

This dataset comprises paired feature and action cost data generated for diverse planning problems. The features are sampled from a multivariate Gaussian distribution, while the action costs are derived through a synthetic generation process. Additionally, the dataset includes instances generated from the Rovers planning domain and transportation problems, as well as synthetic action costs constructed to evaluate the effectiveness of decision-focused learning methods. Two instance scales, small and large, are provided for experimental purposes. The associated task is automated planning with action cost prediction.