Considering progressively more flexible satellite payloads and increasingly dynamic and heterogeneous traffic requirements, it is essential to allocate the limited and multidimensional communication resources efficiently. In order to match system throughput and users’ demands, an algorithm based on multi-agent reinforcement learning for multi-beam satellite systems is proposed in this paper, which takes two degrees of freedom, power and frequency, into account. To that end, a downlink model is established firstly. Then a joint frequency assignment and power allocation optimization problem is formulated and a multi-agent deep reinforcement learning algorithm using centralized training and distributed execution is presented, where each beam is modeled as one agent to avoid the curse of dimensionality in the action space. Moreover, a converting multi-discrete actions to discrete (CMD) trick is raised to handle the hierarchical multi-discrete actions. Simulation results show that the proposed approach has better performance compared to the existing algorithms and can achieve real-time frequency assignment and power allocation intelligently.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Joint Frequency Assignment and Power Allocation Based on Multi-Agent Deep Reinforcement Learning for Multi-Beam Satellite Systems


    Contributors:
    Li, Yuanjun (author) / Yang, Dewei (author) / Yang, Haowen (author) / Kuang, Jingming (author)


    Publication date :

    2023-06-01


    Size :

    1314461 byte





    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English




    DEEP REINFORCEMENT LEARNING FOR MULTI-AGENT AUTONOMOUS SATELLITE INSPECTION

    Lei, Henry H. / Shubert, Matt / Damron, Nathan et al. | Springer Verlag | 2024