关键词: 无人机蜂群, 序列凸近似优化, 雷达散射截面（RCS）, 稀疏阵列

Abstract: Aiming at the challenges faced by Unmanned Aerial Vehicle (UAV) swarms in aggressive maneuvering penetration missions—specifically complex dynamic constraints, high real-time requirements, and the difficulty of optimizing non-convex stealth characteristics—a collaborative stealth control method based on Sequential Convex Approximation (SCA) is proposed. First, the electromagnetic scattering characteristics of sparse arrays under nonlinear dynamic con-straints are analyzed. The mapping relationship between maneuvering flight and dynamic Radar Cross Section (RCS) is revealed, and a collaborative stealth control model under aggressive maneuvering conditions is established. To ad-dress the difficulty of solving this control model, first-order Taylor expansion and convex relaxation techniques are utilized to transform the highly non-convex objective function and collision avoidance constraints into a series of Sec-ond-Order Cone Programming (SOCP) problems. Furthermore, an adaptive trust region strategy is introduced to en-sure algorithm convergence. Simulation results in multi-radar threat scenarios demonstrate that the proposed algo-rithm achieves fast convergence and possesses millisecond-level real-time calculation capabilities. By dynamically reconfiguring the formation configuration, the swarm achieves RCS null formation in threat directions while strictly satisfying maximum maneuvering overload and safety collision avoidance constraints. The mean RCS in the main threat directions is reduced by over 100 dB, significantly enhancing the survivability of the swarm during aggressive maneuvering penetration.

Key words: UAV swarm, Sequential Convex Approximation (SCA), Radar Cross Section (RCS), Sparse array