Visibility-aware Cooperative Aerial Tracking with Decentralized LiDAR-based Swarms

TL;DR

The paper addresses the challenge of persistent visibility in cooperative aerial tracking with decentralized LiDAR swarms. It introduces a Spherical Signed Distance Field (SSDF) to model 3-D occlusion, a general Field-of-View alignment cost for heterogeneous LiDARs, and a 3-D swarm distribution cost inspired by Thomson's problem, integrated into a two-stage kinodynamic front-end and SE(3) back-end planner. The approach is implemented in a fully decentralized Swarm-LIO-based system and validated through extensive simulations and real-world experiments, including dynamic joining/leaving of agents and operation in dense outdoor environments. The results demonstrate robust, scalable, and real-time cooperative tracking with high visibility maintenance and flexible sensor configurations, highlighting practical impact for surveillance, cinematography, and industrial inspection tasks.

Abstract

Autonomous aerial tracking with drones offers vast potential for surveillance, cinematography, and industrial inspection applications. While single-drone tracking systems have been extensively studied, swarm-based target tracking remains underexplored, despite its unique advantages of distributed perception, fault-tolerant redundancy, and multidirectional target coverage. To bridge this gap, we propose a novel decentralized LiDAR-based swarm tracking framework that enables visibility-aware, cooperative target tracking in complex environments, while fully harnessing the unique capabilities of swarm systems. To address visibility, we introduce a novel Spherical Signed Distance Field (SSDF)-based metric for 3-D environmental occlusion representation, coupled with an efficient algorithm that enables real-time onboard SSDF updating. A general Field-of-View (FOV) alignment cost supporting heterogeneous LiDAR configurations is proposed for consistent target observation. Swarm coordination is enhanced through cooperative costs that enforce inter-robot safe clearance, prevent mutual occlusions, and notably facilitate 3-D multidirectional target encirclement via a novel electrostatic-potential-inspired distribution metric. These innovations are integrated into a hierarchical planner, combining a kinodynamic front-end searcher with a spatiotemporal $SE(3)$ back-end optimizer to generate collision-free, visibility-optimized trajectories.Deployed on heterogeneous LiDAR swarms, our fully decentralized implementation features collaborative perception, distributed planning, and dynamic swarm reconfigurability. Validated through rigorous real-world experiments in cluttered outdoor environments, the proposed system demonstrates robust cooperative tracking of agile targets (drones, humans) while achieving superior visibility maintenance.

Figures (18)

• Figure 1: A swarm of four autonomous drones is cooperatively tracking a human runner using heterogeneous LiDAR configurations. The LiDAR setup consists of one upward-facing Mid360 LiDAR (marked by blue dashed lines), one downward-facing Mid360 LiDAR (green dashed lines), and two Avia LiDARs (red dashed lines). The swarm forms a 3-D distribution to track the target, with each tracker positioned optimally to suit its FOV settings.
• Figure 2: An overview of our complete decentralized swarm tracking system, including the decentralized swarm localization, shared mapping, collaborative target estimation, onboard control, and motion planning modules. A hierarchical planner is designed to generate optimal trajectories for swarm tracking. All critical operational data is exchanged between swarm members via a UDP-based wireless network.
• Figure 3: (a) An illustration of the spherical discretization and the definition of the visibility map. A map cell is flagged as occluded if obstacles block the LOS. (b) An illustration of the monotonic property used in visibility maps. Along the direction $(0,\phi)$ in the figure, an obstacle blockage is at radius $r_{min}$. Then all grids with radii larger than $r_{min}$ can be directly set as occluded.
• Figure 4: Illustrations of the SSDF definition. (a) On the surface of a unit sphere, an occluded area is shadowed by an obstacle. For two queried directions $\textbf{v}_a$ and $\textbf{v}_b$, the black dashed curves on the sphere indicate the directions' Spherical Signed Distances (SSD) to the closest visibility boundary. (b) A top-down view of figure (a). (c) The corresponding occluded cells on the discretized 2-D $\theta$-$\phi$ grid. The red dashed curves indicate the SSDF values of the queried directions $\textbf{v}_a$ and $\textbf{v}_b$ on the discrete grid.
• Figure 5: An illustration of the incremental SSDF update strategy on a 2-D visibility map. The SSDF is updated layer by layer, starting from the outermost 1-D visibility layer at $r_6$ to the innermost layer at $r_1$. Proceeding to each layer, only the newly visible cells (in red boxes) and value-changed cells (in blue boxes) are identified and updated to compute the SSDF. In this example, only two cells in blue boxes require new value calculations throughout the update.