Traffic-Aware Resource Management of Beam Hopping in Satellite-Enabled Internet of Things

Beam hopping (BH)-enhanced satellite-enabled Internet of Things (S-IoT) is a significant complement to terrestrial Internet of Things (IoT), and is also a key component of the nonterrestrial network (NTN)-enabled IoT. For BH low-Earth orbit (LEO) satellite IoT, efficient resource management is crucial for improving system performance. The joint allocation of multidimensional resources, such as time, frequency, and power, needs to be investigated urgently, with multiple purposes of maximizing the long-term throughput, minimizing the average delay of real time (RT) services and assuring the fairness. Involving both discrete and continuous variables, the multidimensional resources allocation problem is formulated as a multiobjective mixed integer programming problem. To address this problem, we transform it into two subproblems. First, the power optimization (PO) subproblem is approximated as a convex optimization problem and further solved. Subsequently, the beam scheduling subproblem is modeled as a Markov decision process. Furthermore, an action masking multiobjective double deep Q network (AMM-DDQN) algorithm is proposed based on Chebyshev scaling and action masking strategy. The simulation results demonstrate the convergence of the proposed AMM-DDQN algorithm, which outperforms the baseline methods in terms of multiple performances. Specifically, compared with the greedy with distance limit strategy, TopKDQN without PO method, TopKDQN method, genetic algorithm, and random method, the average delay of RT services of the proposed algorithm is reduced by 22.51%, 10.82%, 4.42%, 34.41%, and 52.13%, respectively, achieving QoS guarantees in BH LEO S-IoT.