Lightweight Distributed Spectrum Access for Integrated Satellite–Terrestrial IoT
Ziming Liu, Conghui Lu, Afeng YangABSTRACT
Satellite and terrestrial networks are commonly interconnected in 6G networks, where multi‐dimensional interference and spectrum fragmentation can occur, making traditional static allocation and simplified 2D modeling less suitable. In this dynamic 3D landscape, the available spectrum management approaches do not account for rapid topology changes between LEO satellites and the large, heterogeneous requirements generated by the terrestrial IoT devices. We introduce a new 3D system model and a new spectrum access framework, Att‐DoubleESN, featuring attention‐oriented computation and echo state network (ESN) in a distributed double‐deep Q‐network (DDQN) setting. Our implementation with low computational overhead results in at least 70% lower secondary user–primary user (SU‐PU) collision rates than baseline DRQN and advanced DEQN. Extensive simulations demonstrate that our framework strikes an optimal balance between QoS and intelligent and resilient spectrum management for satellite‐IoT networking architecture in the future, highlighting its potential for deployment on lightweight edge devices.