S³F · Safety-Critical Aerial Robotics

Real-Time Optimisation-Based Safety-Critical UAV Navigation via a Sliding-Mode Reformulation of Barrier and Lyapunov Constraints

Authors — coming soon (paper under review)
Conference paper · 2026

TL;DR

Barrier-function (CBF/CLF) safe control normally requires solving a constrained quadratic programme at every control update — a real computational burden on embedded UAVs flying through cluttered environments with many simultaneous safety constraints. S³F reformulates that optimisation as a sliding-mode reaching problem: the Fischer–Burmeister function embeds the CLF inequality, HOCBF safety inequalities, and the optimality conditions into a single nonlinear residual, whose zero set becomes an optimality sliding manifold. A dedicated reaching law drives the system onto this manifold — recovering the optimal constrained control action without invoking a QP solver.

−70%
online computation time vs. conventional QP-based safe control
ψ ≥ 0
HOCBF safety measures preserved throughout every flight
3D
cluttered environments, heterogeneous obstacles & safety margins
Real UAV
validated on a physical platform in real time

How it works

1

Nominal tracking

A PD-style controller tracks an A* reference path through the environment — fast, but safety-blind.

2

Safety constraints

CLF convergence, higher-order CBF obstacle avoidance, flight-envelope and actuator limits — all affine in the control input.

3

FB embedding

The Fischer–Burmeister function fuses stationarity, feasibility and complementarity into one nonlinear residual S(z).

4

Sliding manifold

The zero set S(z)=0 is an optimality sliding manifold; a reaching law with proven convergence drives the residual to it.

5

Safe control, no QP

Reaching the manifold recovers the optimal minimally-invasive safe input — at a fraction of the QP cost.

Simulation & flight videos

S³F navigating a cluttered 30 × 30 × 10 m environment with heterogeneous ellipsoidal obstacles, tracking an A* reference while the HOCBF measures ψ₁ and ψ₂ remain non-negative throughout the run.


▶
Video coming soon
Onboard chase view. Close-range flythrough between safety-expanded obstacles; green/red shading marks where the nominal controller alone would have been unsafe.

Contributions

Abstract

Barrier-function-based control provides an attractive framework for safety-critical UAV navigation by incorporating safety requirements directly into online constrained optimisation. However, repeatedly solving the resulting optimisation problem can impose a significant computational burden on real-time robotic systems, particularly in cluttered environments involving multiple safety constraints. This paper proposes a novel reformulation of the barrier–Lyapunov constrained control problem as a sliding-mode control problem for computationally efficient UAV navigation. The Fischer–Burmeister function is employed to incorporate the Control Lyapunov Function (CLF) inequality, Control Barrier Function (CBF) inequality, and associated optimality conditions into a unified nonlinear formulation. The resulting optimality system is represented as a sliding manifold, and a dedicated reaching law is developed to drive the system towards the optimal constrained control solution, with convergence established theoretically. The framework is validated in complex three-dimensional navigation scenarios involving heterogeneous obstacles and safety margins, demonstrating safe navigation while reducing online computation time by up to 70% compared with conventional optimisation methods. Monte Carlo simulations further demonstrate robustness under geometric and dynamic uncertainties, and the approach is implemented on a real UAV platform to evaluate its real-time applicability.

Citation

@inproceedings{s3f2026,
  title     = {Real-Time Optimisation-Based Safety-Critical UAV Navigation
               via a Sliding-Mode Reformulation of Barrier and Lyapunov Constraints},
  author    = {Anonymous},
  booktitle = {Under review},
  year      = {2026},
  note      = {Project page: https://wlaa41.github.io/s3f-safe-uav-navigation/}
}