Connectivity Preserving Decentralized UAV Swarm Navigation in Obstacle-laden Environments without Explicit Communication

TL;DR

This paper tackles decentralized UAV swarm navigation in obstacle-rich environments while preserving sensing network connectivity without inter-vehicle communication. It introduces a control barrier function (CBF) framework that enforces multiple safety and connectivity constraints, including line-of-sight (LOS) maintenance, inter-agent distances, and obstacle avoidance, by formulating derivative-based, distributed constraints and solving a soft-constrained quadratic program (QP) to modify desired inputs. Two solution pathways are provided: an optimization-based method with soft constraints and an efficient approximate method that bypasses numerical optimization, both demonstrated to reduce oscillations and constraint violations relative to repulsive APF-based approaches; extensive simulations and real quadrotor experiments validate performance and feasibility. The work advances practical, scalable, non-communicative coordination for UAV swarms in clutter, offering a viable option for real-time deployment with varying onboard compute capabilities. Overall, the approach enables robust, connectivity-preserving swarm navigation through complex environments with practical implications for search-and-rescue, surveillance, and cooperative exploration.

Abstract

This paper presents a novel control method for a group of UAVs in obstacle-laden environments while preserving sensing network connectivity without data transmission between the UAVs. By leveraging constraints rooted in control barrier functions (CBFs), the proposed method aims to overcome the limitations, such as oscillatory behaviors and frequent constraint violations, of the existing method based on artificial potential fields (APFs). More specifically, the proposed method first determines desired control inputs by considering CBF-based constraints rather than repulsive APFs. The desired inputs are then minimally modified by solving a numerical optimization problem with soft constraints. In addition to the optimization-based method, we present an approximate method without numerical optimization. The effectiveness of the proposed methods is evaluated by extensive simulations to compare the performance of the CBF-based methods with an APF-based approach. Experimental results using real quadrotors are also presented.

Figures (11)

• Figure 1: Example of $\Phi^{\rm col}(z)$ ($d_c = 0.3, d_r = 0.7, d_m = 1.0$).
• Figure 2: Relative position and velocity between robots $i$ and $j$, and illustration of point $q$ on LOS.
• Figure 3: 2D illustration of $D_{ij}$ and $\|\varphi(x_{ki}, x_{ji})\|$.
• Figure 4: Example of $\omega_{m}(z)$ for $\mu_m = 1, \beta_m = 0.1$.
• Figure 5: 2D illustration of deadlock isosceles triangle.