AFSACPSO UAV path-planning benchmark results and continuous-clearance diagnostics

This dataset supports the article "Soft Actor-Critic with Attractor-Field PSO Parameter Adaptation for UAV Path Planning" by Viet-Anh Ngo, Thanh-Hung Nguyen and Duc-Toan Nguyen, accepted for publication in Swarm and Evolutionary Computation. The manuscript reference is SWEVO-D-26-01025 and the production reference is SWEVO 102599. The catalog contains 3,720 run-level group records, 348 pairwise statistical comparisons, 23,520 saved waypoint rows representing 3,360 distinct complete paths, terrain grids, full-precision summaries, plot traces and source/checkpoint identities. Five comparison groups cover external RL methods, classical or adaptive PSO methods, matched control interfaces, spread sensitivity and paired geometric-signal ablations. Each configuration has 30 runs per scenario. Shared AFSACPSO reference cohorts and the separate paired signal reference are identified in the documentation. Continuous terrain-clearance diagnostics are reported separately from the sampled finite-penalty objective. The retained Scenario 4 goal lies below the prescribed 8-unit terrain-clearance margin. The saved paths and diagnostics include the observed clearance violations. This version preserves the numerical evidence corresponding to the accepted manuscript. AFSACPSO_Result_Catalog.zip is the complete curated catalog. README.txt and DATA_DICTIONARY.txt are also provided separately for browsing. The clean supplementary appendix with Tables A1-A3, B1-B3 and C1-C5 is supplied separately to the journal. Large native result bundles and frozen checkpoint binaries remain subject to the article's corresponding-author-on-request availability statement.