A Study of Two-Tier Cellular Vehicular Communication Based on Deep Reinforcement Learning(2/2)

瀏覽統計 Email 通知 RSS Feed

簡歷

基本資料

Project title

Code/計畫編號

MOST109-2221-E019-051-MY2

Translated Name/計畫中文名

基於深度強化學習之雙層蜂巢式車聯網傳輸研究(2/2)

Project Coordinator/計畫主持人

Yao-Liang Chung

Department/Unit

Department of Communications, Navigation and Control Engineering

Website

https://www.grb.gov.tw/search/planDetail?id=13531479

Year

2020

Start date/計畫起

01-08-2020

Expected Completion/計畫迄

31-07-2021

Bugetid/研究經費

1040千元

ResearchField/研究領域

資訊科學--軟體

Over the last decade, the field of vehicular communications has attracted much academic and industry interest. 5G roadmaps are expected to open up more possibilities and improvements for connected vehicles. Therefore, the development of next-generation heterogeneous cellular vehicular transmission technologies that offer highly-efficient vehicular communications for various applications is a challenging and forward-looking task. Even though this issue has received a lot of attention in recent years, the high mobility, complexity, and variability of the communication environment has resulted in high computing requirements in transmission technologies. Consequently, it is very difficult to achieve real-time implementation and manage resources. Deep reinforcement learning (DRL) is a novel tool that has been developed in recent years as an effective means of solving problems and challenges pertaining to resource management in communication networks. However, as of now, there is limited literature related to the effective utilization of DRL tools in next-generation heterogeneous cellular vehicular transmission technologies. In particular, there is a dearth of studies that not only cover the transmission modes of connected devices and various base stations, but also take into account the allocation of resources and the configurations of possible transmission modes. This lack of comprehensive and in-depth discussions indicates that there is much room for further and proactive investigations, as well as in-depth explorations. In this proposal, an in-depth study that focuses on this direction of research will be implemented for two years. The current objectives are to overcome the challenges in the resource allocation of next-generation two-tier cellular vehicle-to-vehicle communication environments, as well as to improve the system’s overall effectiveness. On the systematic side, a cross-layer analysis was adopted to develop novel transmission insights, and DRL was applied to adjust the system’s parameters and achieve the dynamic optimization of system performance. This research, which considers both technical depth and interdisciplinary integration, is expected to strike a perfect balance between meeting engineering and socio-economic needs. 過去十年迄今，車輛通信吸引了學術界和業界的極大興趣。同時，5G發展藍圖將為聯網車輛帶來更多的可能性與優越性。發展下一代異質蜂巢式車聯網傳輸技術提供高效的車輛通信來滿足不同的要求是個極具挑戰性與前瞻性的目標，是近年來相當被重視的議題。然而，由於高移動性和複雜多變性的通信環境，因此傳輸技術在計算上的高度需求使得即時實施極具困難也使得資源變得難以管理。另一方面，深度強化學習最近一直在開發成為被用作有效解決通訊網路資源管理問題和挑戰的新興工具。然而，至目前，如何有效應用深度強化學習利器於下一代雙層蜂巢式車聯網傳輸之文獻卻非常有限(特別是聯合裝置對裝置傳輸模式與多類型基站傳輸模式並考慮其可能傳輸方式的組態設定之資源分配議題著墨甚少)，缺乏全方位且深入的探討，顯然有進一步積極探討與深入探索之學術研究空間。本計畫擬定兩年期著重此方向進行徹底之研究，目標將專注解決在此下一代雙層蜂巢式車聯網通訊環境未解的資源分配難題與整體系統效能之提升，從系統層面跨層解析提出新的傳輸見解藉由應用深度強化學習來適應調整系統參數達系統效能動態最佳化。期許此兼顧技術深度與跨領域整合之研究達成工程技術與社會經濟完美契合之願景。

Keyword(s)

下一代蜂巢式網絡
資源管理
效能分析
Next-Generation Cellular Networks
Resource Management
Performance Analysis

DSpace CRIS

A Study of Two-Tier Cellular Vehicular Communication Based on Deep Reinforcement Learning(2/2)

基本資料

Description