CLF-ULP: Cross-Layer Fusion-Based Link Prediction in Dynamic Multiplex UAV Networks

TL;DR

This work addresses the challenge of predicting future connectivity in dynamic multiplex UAV networks with multiple interacting layers. It proposes CLF-ULP, a cross-layer fusion-based model that combines intra-layer GAT embeddings, cross-layer attention fusion, and a shared-parameter LSTM to capture temporal evolution, optimized via a joint loss including intra- and inter-layer embedding consistency. The authors introduce a dynamic multiplex UAV network model, generate datasets under diverse mobility patterns, and demonstrate that CLF-ULP achieves state-of-the-art performance on AUC and AP across scenarios, with robustness to sampling intervals and flight speeds. The framework and datasets offer a strong foundation for global topology forecasting in UAV networks and can be extended to other dynamic multiplex systems.

Abstract

In complex Unmanned Aerial Vehicle (UAV) networks, UAVs can establish dynamic and heterogeneous links with one another for various purposes, such as communication coverage, collective sensing, and task collaboration. These interactions give rise to dynamic multiplex UAV networks, where each layer represents a distinct type of interaction among UAVs. Understanding how such links form and evolve is both of theoretical interest and of practical importance for the control and maintenance of networked UAV systems. In this paper, we first develop a dynamic multiplex network model for UAV networks to characterize their dynamic and heterogeneous link properties. We then propose a cross-layer fusion-based deep learning model, termed CLF-ULP, to predict future inter-UAV links based on historical topology data. CLF-ULP incorporates graph attention networks to extract topological features within each layer and perform a cross-layer attention fusion to capture inter-layer dependencies. Furthermore, a shared-parameter long short-term memory network is employed to model the temporal evolution of each layer. To improve embedding quality and link prediction performance, we develop a joint loss function that considers both intra-layer and inter-layer UAV adjacency. Extensive experiments on simulated UAV datasets under diverse mobility patterns demonstrate that CLF-ULP achieves state-of-the-art performance in predicting links within dynamic multiplex UAV networks.

Figures (7)

• Figure 1: An example of a dynamic multiplex UAV network represented as a sequence of snapshots over time. The network consists of three layers corresponding to different link types among eight shared nodes. Intralayer links connect different nodes within a layer, while interlayer links connect identical nodes across layers.
• Figure 2: The overall framework of the proposed link prediction model, CLF-ULP. The snapshot sequence of each layer is independently fed into a GAT to generate the corresponding intra-layer node embeddings. These embeddings are then aggregated across layers to capture inter-layer interactions through a cross-layer attention fusion mechanism. Subsequently, a shared-parameter LSTM is employed to extract temporal dependencies and produce final node embeddings that incorporate temporal information. Finally, in the decoder module, an MLP takes as input the link features constructed by concatenating the embeddings of node pairs and transforms them into the predicted adjacency matrix.
• Figure 3: AUC and AP vs. sampling interval for different methods across the four mobility models.
• Figure 4: AUC and AP vs. flight speed for different methods across the four mobility models.
• Figure 5: Node embedding visualization for the four mobility models.