Three Dimensional Hydrodynamic Flow-Based Collision Avoidance for UAV Formations Facing Emergent Dynamic Obstacles

TL;DR

The paper addresses collision avoidance for UAV formations in dynamic environments by introducing a 3D hydrodynamics-inspired framework where emergent obstacles are modeled as 3D doublets or ellipsoids to generate local velocity fields via harmonic flow, enabling smooth maneuvers without trajectory replanning. It couples this flow-based avoidance with a Virtual Rigid Body (VRB) formation to preserve formation geometry and trajectory tracking in multi-UAV deployments. Key contributions include the 3D flow model for moving obstacles, the integration with VRB formation using rigidity constraints and Baumgarte stabilization, and comprehensive 3D simulations demonstrating safe, scalable performance under noise. The approach is computationally efficient and well-suited for real-time UAV operations in cluttered, dynamic environments, with future work focusing on higher-fidelity SITL validation (e.g., ROS2-Gazebo).

Abstract

This paper presents a three-dimensional, hydrodynamics-inspired collision avoidance framework for uncrewed aerial vehicle (UAV) formations operating in dynamic environments. When moving obstacles enter a UAV's sensing region, they are modeled as three dimensional doublets or ellipsoids that generate local velocity fields, guiding nearby UAVs to execute smooth, collision-free maneuvers without trajectory discontinuities or explicit trajectory replanning. This flow-based approach enables real-time operation and interpretable behavior by leveraging the nature of fluid flow around obstacles via the harmonic properties of Laplace's equation, inherently avoiding local minima common in traditional potential field methods. To establish and maintain coordination among the UAVs, a Virtual Rigid Body (VRB) formation strategy is integrated, ensuring that formation geometry and trajectory tracking are preserved. Simulation results demonstrate the feasibility and scalability of the method for both individual and multi-UAV scenarios with multiple formation geometries encountering moving obstacles. The proposed approach achieves safe, smooth, and computationally efficient avoidance maneuvers suitable for real-time and practical applications.

Figures (15)

• Figure 1: Illustration of relationship between point source and point sink
• Figure 2: Relationship between the inertial coordinate, doublet, and fluid particle
• Figure 3: Illustration of the relationship between agents
• Figure 4: Illustration of the agent slot allocation method
• Figure 5: Quadcopter diagram