| Presentation | 2018-05-25 Traffic Control with Deep Reinforcement Learning Using an RGB-D Camera for Millimeter-wave Communications Tomoya Mikuma, Takayuki Nishio, Masahiro Morikura, Yusuke Asai, Ryo Miyatake, |
|---|---|
| PDF Download Page |  PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In millimeter-wave (mmWave) communications, throughput is decreased seriously when line-of-sight paths are blocked by human bodies. In order to solve the throughput degradation problem, a proactive traffic control system based on RGB-D camera images has been proposed. The system predicts the human blockages and controls data transmission to each STA in order to increase the system throughput. The previous works employ the heuristic traffic control strategy which requires manual tuning suitable for a specific wireless communication environment. In order to solve the problem, this paper proposes a traffic control system with deep reinforcement learning using depth images obtained from an RGB-D camera and states of traffic buffers in a proxy server. The proposed system explores the optimal traffic control strategy appropriate to a wireless communication environment autonomously through trial and error, and it is expected to obtain a traffic control strategy improving the system throughput. The performance of the proposed traffic control system is evaluated by experiments. The experimental results show that the proposed system increases the system throughput as the learning progresses. In addition, the proposed system achieves higher performance compared with a simple heuristic algorithm. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | mmWave communication / IEEE 802.11ad / human blockage / RGB-D camera / deep reinforcement learning |
| Paper # | MoNA2018-1 |
| Date of Issue | 2018-05-18 (MoNA) |
| Conference Information | |
| Committee | MoNA / IPSJ-DPS / IPSJ-MBL / IPSJ-ITS |
|---|---|
| Conference Date | 2018/5/24(2days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | Eef Information Plaza |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | 5G, Mobile Applications, Ubiquitous Services, Mobile Distributed Cloud, ITS, etc. |
| Chair | Ryoichi Shinkuma(Kyoto Univ.) / 田上 敦士(株式会社KDDI総合研究所) / 河口 信夫(名古屋大学) / 重野 寛(慶應義塾大学) |
| Vice Chair | Shigeaki Tagashira(Kansai Univ.) / Gen Kitagata(Tohoku Univ.) |
| Secretary | Shigeaki Tagashira(Kyushu Univ.) / Gen Kitagata(NTT) / (NEC) / (明治大学) / (（株）富士通研究所) |
| Assistant | Takayuki Nishio(Kyoto Univ.) / Takato Saito(NTT) |
| Paper Information | |
| Registration To | Technical Committee on Mobile Network and Applications / Special Interest Group on Distributed Processing System / Special Interest Group on Mobile Computing and Ubiquitous Communications / Special Interest Group on Intelligent Transport Systems and Smart Community |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Traffic Control with Deep Reinforcement Learning Using an RGB-D Camera for Millimeter-wave Communications |
| Sub Title (in English) | |
| Keyword(1) | mmWave communication |
| Keyword(2) | IEEE 802.11ad |
| Keyword(3) | human blockage |
| Keyword(4) | RGB-D camera |
| Keyword(5) | deep reinforcement learning |
| 1st Author's Name | Tomoya Mikuma |
| 1st Author's Affiliation | Kyoto University(Kyoto Univ.) |
| 2nd Author's Name | Takayuki Nishio |
| 2nd Author's Affiliation | Kyoto University(Kyoto Univ.) |
| 3rd Author's Name | Masahiro Morikura |
| 3rd Author's Affiliation | Kyoto University(Kyoto Univ.) |
| 4th Author's Name | Yusuke Asai |
| 4th Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
| 5th Author's Name | Ryo Miyatake |
| 5th Author's Affiliation | Nippon Telegraph and Telephone Corporation(NTT) |
| Date | 2018-05-25 |
| Paper # | MoNA2018-1 |
| Volume (vol) | vol.118 |
| Number (no) | MoNA-56 |
| Page | pp.pp.159-164(MoNA), |
| #Pages | 6 |
| Date of Issue | 2018-05-18 (MoNA) |