"3D HARRT Dataset"

"This dataset provides a large-scale collection of generated 3D environments for evaluating sampling-based motion planning in complex and high-dimensional spaces. Each scenario is available in two complementary formats: (i) a raw geometric representation, where environments are defined using parametric box and spherical obstacles along with start\u2013goal pairs in order to enable flexible reconstruction, analysis, and customization; and (ii) a voxelized occupancy grid representation suitable for direct integration with learning-based methods. The environments are designed to mimic dense urban-like layouts, consisting of vertically structured obstacles, narrow corridors, and additional clutter that induce challenging navigation conditions. The dataset captures a wide range of spatial configurations, topology variations, and obstacle densities in order to support reproducible benchmarking. In total, the dataset includes 10,000 testing scenarios, 500 validation scenarios, and 5 unseen test scenarios designed to evaluate generalization. "