一個基於馬可夫決策過程之異質網路接取策略

異質網路提供行動通訊裝置多種無線接取技術。在這種具有多種無線接取技術可採用的環境裡，接取網路的選擇變的很重要。為此我們提出一個新型的網路選擇接取策略，此策略稱為多種使用者種類的最大化獎勵策略(MU-MRP)。此策略目的是為了在合理的價格中提升無線通訊服務的品質, 降低換手次數與換手失敗次數。MU-MRP 為基於半馬可夫決策過程的策略，此數學模型所的到的策略確保此策略可以達成最大的整體獎勵。我們使用Q 學習演算法找出最優策略，許多模擬結果顯示MU-MRP 比其他策略達到更大的獎勵，此外，MU-MRP 可以大幅減低換手次數與換手失敗次數。

關鍵字

網路選擇策略；半馬可夫決策過程；多種使用者種類； Q-學習

並列摘要

Heterogeneous Networks provide user equipments (UEs) diverse radio access technologies (RATs). Under this environment with RATs, network selection goes important. We propose a novel network selection policy (NSP) so-called Multi-User-Typed Maximum Reward Policy (MU-MRP). The incentive is to enhance the quality of service (QoS), reduce the number of handoffs and dropping events in consideration of a reasonable monetary cost. MU-MRP is based on semi Markov decision process (SMDP), which offers an optimal policy to maximize the overall rewards. The Q-learning algorithm is used to determine the optimal policy. Numerical results show that MU-MRP earns more total system rewards than other policies. Also, MU-MRP reduces the number of handoffs and dropping events obviously.

並列關鍵字

Network Selection Policy ； SMDP ； Multi-User Type ； Q-learning

參考文獻

[1] Cisco, “Cisco visual networking index: Forecast and trends, 2017–2022,” White Paper, Nov 2018.

Google Scholar

[2] 3GPP TS 24.234 v. 12.2.0, “3GPP System to Wireless Local Area Network (WLAN) interworking; WLAN User Equipment (WLAN UE) to network protocols; Stage 3,” Mar 2015.