Hierarchical planning and control for multi-UAV cooperative encirclement in complex environments

复杂环境下多无人机协同围捕的分层规划与控制

¹ School of Automation, Xi'an University of Posts and Telecommunications, Xi'an 710061, China
² School of Astronautics, Northwestern Polytechnical University, Xi'an 710072, China

Received: 12 October 2025

Abstract

In complex environments, multi-UAV cooperative encirclement tasks often face critical challenges such as low intruder trajectory prediction accuracy, excessive system energy consumption during task allocation, and insufficient obstacle avoidance capability under dynamic conditions. These issues severely limit the efficiency and reliability of encirclement operations. To address these problems, a two-stage cooperative encirclement framework consisting of a gathering phase and a guidance phase is proposed in this paper. In the gathering phase, an interpolation and error-correction polynomial fitting(IEC-PF) method is designed to achieve early interception and significantly improve the accuracy of intruder trajectory prediction. For the task allocation of capture points, an iterative convex optimization-based allocation strategy is developed to minimize total system energy consumption while strictly satisfying the final capture formation constraints. In the guidance phase, UAVs construct an optimal encircling formation based on the assigned capture points. Within the generated obstacle-free convex regions, a nonlinear optimization method is used to compute optimal global paths and formation configuration parameters. An affine formation control strategy based on stress matrices is further introduced to ensure formation stability and reconfiguration during path tracking. Additionally, a deep reinforcement learning(DRL)-based local obstacle avoidance mechanism is integrated to enhance the system's real-time responsiveness to dynamic obstacles. Simulation results demonstrate that the proposed framework substantially improves obstacle avoidance performance in complex environments and enables reliable multi-UAV cooperative capture of intruding targets, as well as guiding them to designated areas.

摘要

在复杂环境下, 多无人机协同围捕入侵目标时, 常面临目标轨迹预测精度欠佳、任务分配过程中系统能耗过高及动态环境下避障能力不足等关键挑战, 严重制约了围捕任务的效率与可靠性。针对上述问题, 提出一种包含聚集阶段与引导阶段的两阶段多无人机协同围捕框架。在聚集阶段, 为实现对入侵目标的提前拦截, 设计插值与误差校正的多项式拟合(IEC-PF)方法, 有效提升目标轨迹预测精度; 针对围捕点分配问题, 提出基于迭代凸优化的任务分配策略, 在严格满足最终围捕编队约束条件的前提下, 实现系统总能耗最小化。在引导阶段, 无人机依据分配的围捕点构建最优包围编队, 基于生成的无障碍凸区域, 通过非线性优化算法求解最优全局路径及编队构型参数; 引入基于应力矩阵的仿射编队控制策略, 保障无人机在路径跟踪过程中编队的稳定与重构; 融入基于深度强化学习(DRL)的局部避障机制, 显著提升系统对动态障碍物的实时响应能力。仿真实验结果表明, 所提框架能够显著增强系统在复杂环境下的避障性能, 有效确保多无人机成功协同围捕入侵目标并将其引导至目标区域。