Helicopter Formation Cooperative Command System Architecture with Fused Unmanned System

Abstract

A helicopter formation cooperative command system architecture integrating unmanned systems is proposed in the context of multi-platform heterogeneous formation combat missions. With the four-layer module of "perception-fusion-decision-execution" as the core, the system constructs an intelligent cooperative command model based on graphical neural network (GCN), graphical attention network (GAT) and multi-intelligence reinforcement learning (MARL), and realizes the cooperative command of multiple sources and heterogeneous formations. We construct an intelligent collaborative command model based on graph neural network (GCN), graph attention network (GAT) and multi-intelligence reinforcement learning (MARL), which realizes the fusion processing of heterogeneous perceptual information from multiple sources and the joint scheduling control of task-path-avoidance. The model design incorporates a lightweight convolutional encoder and a Transformer timing alignment network to improve perceptual robustness, adopts a joint policy search and graph matching mechanism to achieve efficient task allocation, and introduces constrained optimized path generation and RVO dynamic obstacle avoidance strategy to construct a closed-loop control. The system is tested in ROS 2 and Gazebo simulation environments for three types of typical combat missions (patrol, reconnaissance, and proximity strike), and the results show that the system maintains the feature consistency FCI > 0.9, the mission average response time AAL < 50ms, and the trajectory deviation rate MTD < 3.0m in the scenarios of high mission density and communication delay, which verifies the adaptability of the proposed architecture in the complex combat environments. adaptability and stability of the proposed system architecture in the complex combat environment.