Spatiotemporal Correlation Representation based Precise Resource Management in Space-Air-Ground Integrated Network

Abstract

In the space-air-ground integrated network, resources are unevenly distributed and dynamically change over time.Traditional optimization scheduling methods based on static resource representation only optimize scheduling within a single time slot, leading to low resource utilization.Moreover, frequent strategy adjustments are required when the network changes, resulting in significant migration costs.To address this issue, this paper proposed a method for spatiotemporal correlation representation of resources in the space-air-ground integrated network.This method computed node attention representations in both structural neighborhoods and chronological neighborhoods, enabled precise modeling of spatiotemporal resource correlations in the space-air-ground integrated network.Based on this, this paper proposed a resource precise embedding method based on spatiotemporal correlation representation, and introduced the trust region policy optimization method to adaptively learn and optimize algorithm parameters.

Keywords

• space-air-ground integrated network;spatiotemporal attention model;virtual network embedding;time-varying graph