Safe Low-Altitude Navigation in Steep Terrain with Fixed-Wing Aerial Vehicles

TL;DR

This work tackles safe, low-altitude navigation of fixed-wing sUAS in steep terrain under regulatory altitude limits by introducing a safety-centric planning framework based on circular periodic (loiter) paths. By formalizing inevitable collision states (ICS) and using safe periodic sets, the approach enables efficient, yaw-agnostic goal selection and real-time planning via RRT* in the Dubins airplane space, with collision checks reduced to simple geometric tests against offset collision surfaces $D^-$ and $D^+$. The method is validated through simulations across rugged alpine terrains and a real-world flight test in the Swiss Alps, showing robust onboard replanning, safe tracking, and adherence to European altitude regulations ($50$–$120\,\mathrm{m}$). Key contributions include an efficient ICS evaluation via circular loiters, precomputation of valid loiter regions, and an open-source Dubins airplane RRT* planner implementation for fixed-wing sUAS in altitude-constrained environments.

Abstract

Fixed-wing aerial vehicles provide an efficient way to navigate long distances or cover large areas for environmental monitoring applications. By design, they also require large open spaces due to limited maneuverability. However, strict regulatory and safety altitude limits constrain the available space. Especially in complex, confined, or steep terrain, ensuring the vehicle does not enter an inevitable collision state(ICS) can be challenging. In this work, we propose a strategy to find safe paths that do not enter an ICS while navigating within tight altitude constraints. The method uses periodic paths to efficiently classify ICSs. A sampling-based planner creates collision-free and kinematically feasible paths that begin and end in safe periodic (circular) paths. We show that, in realistic terrain, using circular periodic paths can simplify the goal selection process by making it yaw agnostic and constraining yaw. We demonstrate our approach by dynamically planning safe paths in real-time while navigating steep terrain on a flight test in complex alpine terrain.

Figures (10)

• Figure 1: Example of a planned path for a fixed-wing aerial vehicle above complex terrain with an altitude limit of 120 m above the ground (red). The 2D binary projection below the terrain shows the Valid Loiter Positions proposed by this paper. Blue regions contain no kinematically feasible safe periodic states. Paths planned using the safe periodic states proposed in this paper (cyan) compared to the conventional goal position and heading (magenta) are shown.
• Figure 2: State space of Dubins airplane model.
• Figure 3: Visualization of the terrain and maximum(120m), minimum distance(50m) offset collision surfaces of the example of Fig. \ref{['fig:alpine_planner']} (top). Valid loiter positions according to Eq. \ref{['eq:valid_position']} are marked as green and invalid regions are marked as blue (bottom)
• Figure 4: Example of the unidirectional and bidirectional start and goal samples
• Figure 5: Overview of the proposed method. The operator commands a 2D goal position $c_{xy}$, in which the safety is evaluated using the offset collision surface $\mathcal{H}^+, \mathcal{H}^-$. If the commanded goal position is safe, a planning problem is solved for path $\eta*$ using the start and goal circular path and the collision surface $\mathcal{D}^+, \mathcal{D}^-$

Theorems & Definitions (4)

• Definition 3.1: Offset Collision Surface
• Definition 4.1: Inevitable Collision Statefraichard_inevitable_2004
• Corollary 4.1: Safe Periodic Path
• Definition 4.2: Valid Loiter Position