Analysis of Uplink Transmission Scheduling Strategies for LoRa-Based Direct-to-Satellite IoT Networks Using Deep Reinforcement Learning

Recently, the LoRa-based direct-to-satellite Internet of Things (IoT) network has attracted more attention than the terrestrial IoT network with limited coverage due to its global coverage and low-power access to massive devices. However, the uplink packet collision problem of massive devices constrains its transmission efficiency, so how to design uplink scheduling is crucial for the LoRa-based direct-to-satellite IoT network. Considering that the scheduling strategy based on deep reinforcement learning can quickly respond to dynamic events through learning, a proximal policy optimization algorithm based on a self-attention mechanism, namely SAPPO, is proposed, in which the time division multiple access mechanism is used to avoid packet collision and loss. Compared with existing methods, the proposed SAPPO algorithm can effectively improve the uplink transmission efficiency, promote fairness among devices, and improve the priority processing capability.