In future 5G systems, it is envisioned that the physical resources of a single network will be dynamically shared between the virtual end-to-end networks called “slices” and the network is “sliced”. The dynamic sharing of resources can bring about pooling gains, but different slices can easily influence each other. Focusing on slicing the radio access network, a slice management entity is required to steer the radio resource management (RRM) so that all of the slices are satisfied and negative inter-slice influences are minimized. The steering of RRM can be done by adjusting slice-specific control parameters in scheduler and admission controller mechanisms. We use a model-free reinforcement learning (RL) framework and train an agent as a slice manager. Simulation results show that such agents are capable of relatively quickly learning how to steer the RRM. Furthermore, a hybrid method of Jacobian-matrix approximation with RL approach has been devised and shown to be a practical and efficient solution.


    Access

    Check access

    Check availability in my library

    Order at Subito €


    Export, share and cite



    Title :

    Slice Management in Radio Access Network via Deep Reinforcement Learning


    Contributors:


    Publication date :

    2020-05-01


    Size :

    332070 byte





    Type of media :

    Conference paper


    Type of material :

    Electronic Resource


    Language :

    English




    Slice-aware Open Radio Access Network planning and dimensioning

    Foroughi, Parisa / Martins, Philippe / Nivaggioli, Patrice et al. | IEEE | 2022



    Cognitive Radio Network Throughput Maximization with Deep Reinforcement Learning

    Ong, Kevin Shen Hoong / Zhang, Yang / Niyato, Dusit | IEEE | 2019


    Deep Reinforcement Learning

    Huang, Xiaowei / Jin, Gaojie / Ruan, Wenjie | Springer Verlag | 2012