Featured Application
Generates regional candidate centerlines on specified static building models for offline screening of length, obstacle proximity and computational cost before vehicle-specific trajectory and flight validation.
Abstract
Clearance-aware UAV route pre-planning balances length, obstacle proximity, and computation on a specified geometric model. We examine a finite-influence quadratic clearance density, accumulated per unit of physical length. Endpoint averaging (CAWH-E) and interior-edge quadrature (CAWH-I) share the graph and search parameters. A Shanghai comparison finds a 0.461 m mean sampled-clearance increase for I at 42.74 times the search cost. A separate Delft measurement-source experiment records 3978 outcomes across 221 tasks, six methods, and three repeats. A revision-stage saved-record analysis covers all 221 tasks completed by five methods: I takes longer than E on every task, with a median per-task time ratio of 3.41; all four quality median paired differences are zero. A separately frozen 50-task comparison adds 450 measurements against a paper-guided independent three-dimensional CCS-Theta adaptation, which succeeds in 112/150 slots. Its I comparison contains 37 paired tasks, with a median normalized-exposure difference of −0.0299. Lower exposure and greater sampled clearance accompany longer routes. The evidence quantifies model-dependent quality–cost trade-offs and limited benefit from interior quadrature. Vehicle dynamics, tracking, and flight remain unvalidated.