Design of intelligent control algorithm for low-altitude aircraft cluster based on virtual reality technology

The traditional cluster control algorithm has some problems in dynamic environment, such as collaborative efficiency, real-time decision-making and insufficient human-computer interaction experience. Virtual reality (VR) technology provides a new way to solve these problems with its immersive interaction, 3D visualization and real-time simulation capabilities. In this paper, double-loop control architecture is adopted, and virtual domain and physical domain are integrated to realize cluster control. Virtual domain builds 3D dynamic environment based on Unity/Unreal, integrates meteorology, obstacles and task parameters, and carries out task preview and strategy generation through VR interface. The physical domain is composed of distributed aircraft nodes, equipped with edge computing unit, which receives the optimization parameters of the virtual domain and performs control, and sends back the sensing data in real time with the help of 5G/6G communication. In the aspect of algorithm design, the improved RRT* algorithm guided by potential field is used for path planning, which integrates the target gravity, obstacle repulsion and formation to maintain the potential field, and the weight parameters are dynamically adjusted through VR preview. A delay-resistant consistency protocol is designed, and the communication delay compensation term and the cost function generated by VR are combined to improve the response speed and coordination accuracy of the cluster. The experiment was carried out based on Gazebo and ROS platforms, and six virtual drones were used for testing. The results show that the VR-Consensus algorithm proposed in this paper is superior to the traditional consensus algorithm and the classic RRT* algorithm in path planning performance. The average task completion time is shortened to 63.2s, the collision probability is reduced to 10.6% and the path deviation rate is only 4.7%. Under different interference intensities, the algorithm shows excellent formation keeping ability and anti-jamming performance. The root mean square error (RMSE) of formation keeping is 0.51 m, the packet loss rate (PLR) is 11.4% and the recovery time is only 210ms under moderate interference conditions. In addition, VR preview has a significant effect on parameter optimization. With the increase of preview times, the number of previews and actual collisions decreases significantly.