Planning Tasks

该数据集包含了从诸如积木、仓库、网格、N-谜题、管道-NT、探测车、扫描器、存储、运输和遍历等不同领域选取的各种经典规划任务。每个任务都附带有完整的状态空间以及到达目标的最优成本估计。此外，该数据集还包括了使用具有特定配置和超参数的残差神经网络进行的实现和实验。规模方面，除了两个特定变体之外，多个领域的成本均为单位成本。该数据集的任务是学习经典规划中的启发式函数。

This dataset contains various classical planning tasks selected from diverse domains including blocks, warehouse, grid, N-puzzle, pipe-nt, rover, scanner, storage, transportation, and traversal. Each task is accompanied by a complete state space and an optimal cost estimate for reaching the goal state. Additionally, the dataset includes implementations and experiments carried out using residual neural networks with specific configurations and hyperparameters. In terms of cost scale, most domains use unit action costs, with the exception of two specific variants. The core objective of this dataset is to learn heuristic functions for classical planning.