Probabilistic QoS Metric Forecasting in Delay-Tolerant Networks Using Conditional Diffusion Models on Latent Dynamics
Enming Zhang, Zheng Liu, Yu Xiang, Yanwen Qu
TL;DR
This paper addresses QoS metric forecasting in Delay-Tolerant Networks by reframing the task as probabilistic multivariate time-series forecasting and introducing a conditional diffusion framework on latent temporal dynamics. The proposed DiffTCN framework integrates diffusion models with contextual subsequences via latent dynamics captured by temporal convolutional networks and Transformer-based embeddings to handle non-stationary and multi-modal data. Empirical results on a public latency dataset show that DiffTCN improves MSE, MAE, and CRPS over strong baselines, while providing more reliable forecast distributions. This work enables risk-aware QoS management in DTNs and opens avenues for handling out-of-distribution scenarios in dynamic networking environments.
Abstract
Active QoS metric prediction, commonly employed in the maintenance and operation of DTN, could enhance network performance regarding latency, throughput, energy consumption, and dependability. Naturally formulated as a multivariate time series forecasting problem, it attracts substantial research efforts. Traditional mean regression methods for time series forecasting cannot capture the data complexity adequately, resulting in deteriorated performance in operational tasks in DTNs such as routing. This paper formulates the prediction of QoS metrics in DTN as a probabilistic forecasting problem on multivariate time series, where one could quantify the uncertainty of forecasts by characterizing the distribution of these samples. The proposed approach hires diffusion models and incorporates the latent temporal dynamics of non-stationary and multi-mode data into them. Extensive experiments demonstrate the efficacy of the proposed approach by showing that it outperforms the popular probabilistic time series forecasting methods.