Satellite Edge Computing for Mobile Multimedia Communications: A Multi-agent Federated Reinforcement Learning Approach

• DOI: 10.1145/3715146
• Publication Date: 2025-02-03T13:16:02Z
• Authors: Weiwei Jiang; Yafeng Zhan; Xin Fang
• Keywords: Mobile agent; Communications satellite

The rapid expansion of satellite mega-constellations has highlighted the potential edge computing as a promising solution for mobile multimedia communications. While reinforcement learning been explored in communication systems, significant challenges remain, including high latency and limited resources. This study addresses these by focusing on joint optimization communication, computing, caching resources to support applications. A mixed-integer nonlinear programming (MINLP) problem is formulated with objective minimizing total delay experienced users, subject multidimensional resource capacity constraints, which NP-hard computationally intractable solve polynomial time. To address this complexity, we propose multi-agent federated (MAFRL) approach an efficient solution. In framework, each operates autonomous agent equipped actor-critic network structure. proposed MAFRL method demonstrates superior performance, achieving lower delays compared all baseline approaches. It effectively optimizes delay-sensitive communications improving task offloading ratios. best authors’ knowledge, first introduce MAFRL-based allocation marking contribution field.