Rule-Decoupled and Predictive Hierarchical Path Planning for Composite-Wing UAVs in Dynamic Urban Airspace
Abstract Composite-wing unmanned aerial vehicles (UAVs) in dynamic urban airspace must coordinate planning with exclusion rules, moving traffic, and limited turning and braking capabilities. This paper presents Rule-Decoupled and Predictive Hierarchical Path Planning (RDPHPP). An independent interface propagates exclusion geometry, while direction-augmented multi-cost search, Bézier smoothing, and conditional circular fillet recovery construct acceptable references. Prediction and execution share an acceleration-limited integrator to compare continuation, speed yielding, and detours; speed-regime dynamic window control revalidates and executes candidates. On Shenzhen and Guangzhou maps, all 116 successful paths from 60 rule-replacement pairs satisfy exclusion constraints. A fixed-input fillet diagnostic increases reference acceptance from 40/60 to 57/60. Across 540 paired dynamic scenarios, RDPHPP completes 515 missions (95.37%). Its success-rate difference from the generalized velocity obstacle (GVO) adaptation is significant after multiplicity correction, whereas differences from Reactive and No-wait are not significant. In independent controlled execution, violations of the time-growing separation margin occur in 0/256 conditions accepted by finite-deceleration prediction and 27/264 accepted by an instantaneous approximation. These results support rule-configuration propagation, geometric reference construction, and finite-deceleration candidate evaluation, providing a verifiable workflow connecting hierarchical planning stages. The main dynamic study still records 25 collisions, and mission performance incurs flight-distance and modeled energy-consumption costs. The independent benefit of waiting and closed-loop safety on real aircraft require further validation.
Authors
- Ying Li
- Meng Wang
- Hanyu Liu
Institutions
- Guangzhou College of Commerce (CN)
Publication Details
- Journal
- Research Square
- Published
- 2026-11-16
- DOI
- https://doi.org/10.21203/rs.3.rs-11498295/v1
- Primary Topic
- Robotic Path Planning Algorithms
- Type
- preprint