Density-Based Probabilistic Graphical Models for Adaptive Multi-Target Encirclement of AAV Swarm

Abstract

Multi-target encirclement with unmanned aerial vehicle (UAV) swarms is critical for military and civilian applications such as surveillance and disaster response. Existing methods face limitations in adaptability, primarily due to their reliance on predefined formations, excessive communication requirements, and poor scalability. To address these challenges, we propose a density-based probabilistic graphical model (DB-PGM) to achieve adaptive multi-target encirclement. DB-PGM models the cooperative encirclement of multiple targets through a scalable perception-decision graph structure, where node probabilities are dynamically estimated using a local spatial density-driven parameter estimation method. By integrating Maximum A Posteriori inference with real-time perception updates, the system adaptive adjusts encirclement strategies without predefined formations. This approach enables UAVs to achieve adaptive multi-target encirclement based on evolving spatial density distributions and target movements, ensuring adaptability in complex and dynamic environments. Simulations and real-world experiments validate the effectiveness, adaptability, and robustness of the DB-PGM approach, demonstrating its ability to successfully encircle multiple evasive targets in cluttered environments.