Intelligent Multi-Connectivity and Resource Allocation Framework for LEO Satellite Networks

December 8, 2025·
Madyan Alsenwi
,
Eva Lagunas
,
Jorge Querol
,
Yan Kyaw Tun
,
Symeon Chatzinotas
· 2 min read
Abstract
Low Earth Orbit (LEO) satellite networks are expected to play a key role in extending global 6G coverage. However, their dynamic topology and limited radio resources present significant challenges to reliable and efficient user access. One promising solution is multi-connectivity, where a user connects to multiple visible satellites simultaneously to improve link robustness and throughput. In particular, the efficient operation of multi-connectivity in LEO networks depends on making coordinated decisions about satellite-user links and how available spectrum resources are distributed. In this paper, we formulate an optimization problem to jointly optimize user association with multi-connectivity options and resource block (RB) allocation, considering the dynamic and uncertain nature of LEO satellite networks. To solve this problem, we propose a deep reinforcement learning (DRL) framework that learns to make decisions based on global network observations. Specifically, a DRL agent, deployed at a terrestrial central controller with global network visibility, learns to allocate RBs and select multi-connectivity links to maximize network throughput and ensure user Quality-of-Service (QoS). The proposed approach leverages inter-satellite communication links (ISLs) to forward network state information from satellites to the central ground agent and to relay the agent’s decisions back to satellites that do not have direct connectivity to the ground Gateway (GW) stations. Simulation results confirm the effectiveness of the proposed framework in improving data rates in dynamic LEO environments.
Type
Publication
IEEE Globecom Workshops (GC Wkshps)
publication

Abstract & Overview

Low Earth Orbit (LEO) satellite networks are expected to play a key role in extending global 6G coverage. However, their dynamic topology and limited radio resources present significant challenges to reliable and efficient user access. One promising solution is multi-connectivity, where a user connects to multiple visible satellites simultaneously to improve link robustness and throughput. In particular, the efficient operation of multi-connectivity in LEO networks depends on making coordinated decisions about satellite-user links and how available spectrum resources are distributed. In this paper, we formulate an optimization problem to jointly optimize user association with multi-connectivity options and resource block (RB) allocation, considering the dynamic and uncertain nature of LEO satellite networks. To solve this problem, we propose a deep reinforcement learning (DRL) framework that learns to make decisions based on global network observations. Specifically, a DRL agent, deployed at a terrestrial central controller with global network visibility, learns to allocate RBs and select multi-connectivity links to maximize network throughput and ensure user Quality-of-Service (QoS). The proposed approach leverages inter-satellite communication links (ISLs) to forward network state information from satellites to the central ground agent and to relay the agent’s decisions back to satellites that do not have direct connectivity to the ground Gateway (GW) stations. Simulation results confirm the effectiveness of the proposed framework in improving data rates in dynamic LEO environments.

Publication Metadata

Satellite Communications