Energy-Efficient and Latency-Constrained Task Offloading in Cooperative Cube Satellite Network

Service latency reduction and quality improvement are achieved by Mobile Edge Computing (MEC) through offloading computation-intensive tasks to network edges. Cube satellites (CubeSats) act as assisted MEC servers for large-scale, sparsely distributed user equipment (UE). Considering UEs’ limited computation and energy capacities, this paper investigates a collaborative mobile edge computing system with multiple CubeSats. We address task offloading to minimize execution delays and energy consumption via CubeSat task allocation management. A Markov decision process and a cooperative multi-agent deep reinforcement learning framework with an actor-critic algorithm are employed. Results demonstrate our method outperforms other optimization approaches.