ASI: Dataset and Artifact Bundle for Spatial Flight-Stability Mapping Using PX4 Flight Logs

This record contains the dataset and artifact bundle associated with the project “ASI: A Closed-Loop Robustness Proxy for Spatial Flight-Stability Mapping Using PX4 Flight Logs.” The release is centered on reproducibility of the conference paper and includes raw PX4 flight logs, processed analysis tables, feature tables, experiment manifests, controller parameter files, selected reproducibility scripts, normalization constants, and publication figures. Together, these files document the full workflow used to study the Airspace Safety Index (ASI), The raw data are provided as PX4 ULog (.ulg) flight-log files organized by environment. The release covers five environments: E1_openfield, E2_nearwall, E3_corridor, E4_nearground, and E5_rooftop_edge. These logs contain the underlying flight records from which hover-window features, drift-based validation metrics, spatial maps, repeatability statistics, and controller-intervention summaries were derived. The set spans baseline hover runs, grid runs, repeatability runs, and disturbance-sweep runs, thereby supporting the principal experimental conditions described in the paper. The processed tabular data are provided mainly as CSV files. These include environment-level summary tables such as final_env_summary_ASIstar.csv, table_env_summary.csv, and grid-only summary variants, as well as feature-level tables such as asi_features_with_asi.csv, asi_features_with_asi_GRIDONLY.csv, asi_features_GRIDONLY_with_ASIstar.csv, and asi_features_GRIDONLY_with_ASIstar_xy.csv. These files hold the derived ASI-related quantities used in analysis, including feature values, normalized quantities, environment tags, drift metrics, and spatially indexed outputs. Additional point-selection files such as E1_openfield_ASIstar_points_used.csv and E3_corridor_ASIstar_points_used.csv identify the data used for specific map-generation steps. The release also includes statistical result files that document the main validation analyses. These include outputs such as dose_response_trend_stats.csv, dose_response_sigma_summary.csv, E3_repeatability_per_cell.csv, weights_groupkfold.csv, weights_permutation_test.csv, asi_bootstrap_ci.txt, pairwise_mwu_holm.csv, and kw_results.txt. These files summarize the dose–response analyses, repeatability behavior, weight-selection behavior, bootstrap confidence intervals, and nonparametric test results used to support the claims in the paper. In addition, normalization constants are provided in text files such as asi_normalization_constants.txt. This archive includes experiment manifests and controller parameter files. Manifest files document the organization of controller-intervention runs, dose–response runs, repeatability runs, and wind-sigma sweeps, while files such as controller_A.params and controller_B.params preserve controller settings used in comparative analysis. A ULog manifest is also provided to make the relationship between logs and derived artifacts explicit. The release contains selected reproducibility scripts written in Python. These scripts provide preprocessing, analysis, and plotting tasks, including dataset assembly, hover-window extraction, ASI computation, statistics, and generation of publication-quality figures. The full code pipeline corresponding to this release is provided separately in a public GitHub repository, and the relevant repository link is provided in this record. The figure assets are included as PNG and PDF files and correspond to the primary visual outputs reported in the paper. These include the closed-loop proxy validation plots, the dose–response figure, the representative spatial maps for selected environments, and the E3 repeatability visualizations. Their inclusion allows users to directly inspect the released outputs and also verify that the processed tables and plotting scripts reproduce the figures reported in the manuscript. Overall this record is linked with ASI conference paper and is a transparent way for reproduction of the principal results reported in the conference paper, including drift-based validation, dose–response disturbance sweeps, controller intervention, and spatial repeatability analyses across multiple environments. The conference paper itself is currently unreleased and is distributed independently. The paper citation will be provided in this record upon publication.