This paper addresses the increasing significance of UAVs (Unmanned Aerial
Vehicles) and the emergence of UAV swarms for collaborative operations in
various domains. However, the effectiveness of UAV swarms can be severely
compromised by jamming technology, necessitating robust antijamming strategies.
While existing methods such as frequency hopping and physical path planning
have been explored, there remains a gap in research on path planning for UAV
swarms when the jammer's location is unknown. To address this, a novel
approach, where UAV swarms leverage collective intelligence to predict jamming
areas, evade them, and efficiently reach target destinations, is proposed. This
approach utilizes Graph Convolutional Networks (GCN) to predict the location
and intensity of jamming areas based on information gathered from each UAV. A
multi-agent control algorithm is then employed to disperse the UAV swarm, avoid
jamming, and regroup upon reaching the target. Through simulations, the
effectiveness of the proposed method is demonstrated, showcasing accurate
prediction of jamming areas and successful evasion through obstacle avoidance
algorithms, ultimately achieving the mission objective. Proposed method offers
robustness, scalability, and computational efficiency, making it applicable
across various scenarios where UAV swarms operate in potentially hostile
environments.

通过利用无人机集体智能预测干扰区域，回避干扰并高效地达到目标，本文提出了一种鲁棒、可扩展和计算高效的路径规划方法，利用图卷积网络预测干扰区域的位置和强度，并通过多智能体控制算法实现无人机群集的分散和再聚集。模拟实验表明该方法有效地预测了干扰区域并通过避障算法成功回避干扰，最终实现了任务目标的达成。该方法适用于无人机群集在潜在敌对环境中操作的各种情景。