| 大会名称 |
| 2020年 ソサイエティ大会 |
| 大会コ－ド |
| 2020S |
| 開催年 |
|
2020
|
| 発行日 |
|
2020/9/1
|
| セッション番号 |
|
B-5A
|
| セッション名 |
|
無線通信システムA
|
| 講演日 |
|
2020/9/15
|
| 講演場所(会議室等) |
|
Meeting 10
|
| 講演番号 |
|
B-5-19
|
| タイトル |
|
Markov Decision Process (MDP) for Ensemble Learning Method-based Slice Admission Control in Adaptive RAN
|
| 著者名 |
|
○Seung Il MOON, Haruhisa HIRAYAMA, Yu TSUKAMOTO, Shinobu NANBA, Hiroyuki SHINBO,
|
| キーワード |
|
5G, Adaptive RAN, RAN Slicing, Slice Admission Control, Ensemble Learning, Reinforcement Learning
|
| 抄録 |
|
The advanced 5G system is expected to support further diverse services with the performance targets of low-latency, high-reliability or high-capacity. In advanced 5G system, we have proposed an adaptive Radio Access Network (RAN) system [1]. In the Adaptive RAN, an efficient slice admission control (SAC) scheme for RAN slices with machine learning (ML) is required. Because the RAN resources are limited and need to manage efficiently by the SAC scheme in order to satisfy various requirements from users. We introduced the EML-based SAC in adaptive RAN to solve the above problem in [2], where we formulated the SAC problem in Adaptive RAN to maximize the utility of RB. In this paper, we model the MDP to solve the formulated SAC problem using the ML approach, to find the optimal policy that can be described as a probability distribution. The policy vector θ in the SAC problem is to make a decision for an action a_t at a state s_t to maximize the reward r_t in the environment.
|
| 本文pdf |
|
PDF download
|
