Probability-Based SCMA Random Access Scheme for Satellite Machine-Type Communication

Satellite-based massive machine-type communication (mMTC) is expected to play a crucial role in future communication systems by enabling the integration of numerous remote and critical machine devices into a unified network. However, due to the limited spectrum resources in the satellite uplink, signals transmitted by various terminals are prone to interfere with each other, leading to communication failures. To further improve system throughput, sparse code multiple access (SCMA)-based techniques are essential for optimizing resource utilization. However, codebook collisions caused by interference signals can severely impact system performance and lead to retransmission failures. Several resource allocation protocols have been proposed to address this issue, with recent advancements leveraging deep reinforcement learning methods. In this study, we compare the epsilon-greedy Q-learning scheme with the argmax- Q-learning scheme and demonstrate that randomly selecting codebooks can enhance system performance. We theoretically demonstrate the existence of an optimal probability that maximizes the overall system access performance. Furthermore, we propose a Proximal Policy Optimization (PPO)-based random access (RA) scheme, where devices select codebooks based on calculated probabilities. Our results show that employing random access for resource block selection, combined with the PPO-based scheme, effectively reduces interference, enhances system reliability, and achieves superior system performance compared to existing approaches.