A Dynamic Load Balancing Access Control Scheme for Massive IoT in Space-Terrestrial Integrated Network

With the growth of Internet of Things (IoT) application demand, the Space Terrestrial Integrated Network (STIN) is deemed a promising solution to provide global seamless access services for massive IoT (mIoT) devices. However, under dynamic service traffic and channel resource constraints, optimizing lightweight Random Access (RA) control for low-cost mIoT to improve system throughput and resource utilization is a significant challenge. To address this challenge, we first design a centralized access control architecture deployed on the cloud of STIN, tailored for low-cost mIoT devices and resource-constrained Low Earth Orbit (LEO) satellites. In this architecture, a lightweight control protocol based on the broadcast model is presented to minimize interaction control. Then, with the goal of achieving load balancing, we propose a Space-Time Hybrid Access Control (STHAC) scheme to optimize the RA procedure. This scheme jointly considers three temporally coupled control actions. In the space dimension, Network Selection (NS) is executed. In the time dimension, Access Class Barring (ACB) and Back-Off (BO) are executed. An RA control optimization problem is formulated to optimize the number of success accesses and access delay. To tackle this NP-hard optimization problem, we present a Deep Deterministic Policy Gradient (DDPG) algorithm with Long Short-Term Memory (LSTM) to maximize a long-term cumulative reward. This intelligent algorithm dynamically adjusts RA control actions in an online manner according to the time varying environments. Finally, compared with benchmark problems, simulation results verify the effectiveness of proposed scheme, and demonstrate its superiority in terms of both average success accesses and access delay.