Learning-Based Predictive Multi-Metric Routing Optimization for UAV Swarm Networks

In recent years, Unmanned Aerial Vehicle (UAV) swarm networks have become increasingly important across various scenarios such as military, industry, and agriculture. They not only improve the efficiency and flexibility of task execution, but also provide new solutions for communication in complex environments. However, the high mobility of UAVs causes the network topology to change frequently and the links to be easily damaged, which poses a severe challenge to providing reliable low-latency communication services. Existing routing strategies are difficult to meet the performance requirements in high QoS demand scenarios. To solve these problems, this paper proposes a centralized source routing method, Adaptive Predictive Multi-Metric (APMM) routing approach. By introducing a generative diffusion model to anticipate the future state of the network, it dynamically evaluates link stability and improves data transmission reliability. Based on multi-dimensional performance indicators, an ant colony system is used to optimize routing decisions and maintain link connection status to achieve high-quality communication services. Experimental results demonstrate that APMM is superior to existing methods, showing significant performance improvement.