應用於無線通訊網路中以網路容量為依據的媒體接取控制

Abstract

隨著行動裝置及智慧型手機的普及，目前的數據流量正以指數的速率成長。因此，我們正面臨提高數據傳輸速率的挑戰。在下一個世代(5G)的行動無線通訊裡，小型細胞、多封包接收技術以及裝置對裝置通訊等，都被認為是提高網路吞吐量及頻譜效率的有效方法。由於小型細胞的低成本以及快速佈署的特性，小型細胞被大量地部署在行動細胞網路中，用來從減輕負載超載的行動細胞網路的數據資料流量。和傳統的碰撞通道模型比較，具有多封包接收能力的存取點能夠同時成功解出來自多個傳送裝置同時傳送的封包。為了提升小型細胞的容量，設計具有多封包接收能力的媒體存取控制協定是必須的。在細胞網路中的裝置對裝置通訊，由於裝置對裝置用戶與細胞用戶共享相同的資源，細胞網路的頻譜效率可以進一步的提高。然而裝置對裝置通訊會對細胞用戶產生干擾。為了降低裝置對裝置用戶對細胞用戶的干擾，我們必須謹慎的分配資源給裝置對裝置用戶。除此之外，裝置對裝置用戶之間的干擾也不可被忽視。下一代無線網路由許多本質不同的網路所組成，在這樣的整合網路裡，涉及網路選擇的議題變得相當重要。如果行動用戶存取一個不適當的網路，將會導致用戶頻繁的換手或是頻譜的使用效益低落 小型細胞在減輕細胞網路負荷扮演著一個關鍵的腳色。為了提高小型細胞的容量，我們針對無線網路，為具有多封包接收能力的分散式媒體存取控制提出學習演算法。根據此演算法，節點能夠藉由回饋資訊以及其他節點先前使用的策略，調整它的數據傳輸速率。在對稱式的網路中，我們提出了對稱式學習演算法來最大化網路吞吐量。在非對稱的網路中，為了在吞吐量與公平性之間取得一個良好平衡，我們提出了非對稱學習演算法。除此之外，由於在真實的網路環境中可用的數據傳輸速率的數量是有限的，節點無法使用任意的數據傳輸速率傳送數據，因此我們提出計算有限個最佳數據傳輸速率的方法。相較於傳統的碰撞通道模式，所提出演算法的吞吐量可隨著節點數量的上升而增加。 在細胞網路下的裝置對裝置通訊，彼此距離較遠的來源裝置應該使用相同的資源來改善頻譜效益。而彼此距離較近的來源裝置應該使用正交的資源來降低彼此之間的干擾。基於這個原則，我們利用圖形理論來計算裝置對裝置通訊所需要的資源數量。根據需要的資源數量，我們將來源裝置分成若干個群組。為了降低對每個群組吞吐量最佳化的運算複雜度，我們提出一個快速的迭代演算法來計算群組裡來源裝置的數據傳輸速率。而透過此演算法，與裝置對裝置用戶分享資源的細胞用戶的數據傳輸速率可以獲得保證。另外，我們應用消息理論至裝置對裝置通訊，使得目的裝置具有多封包接收的能力。模擬結果顯示，利用所提出演算法得到的吞吐量與最佳的網路吞吐量有相當好的近似。 在異質網路中，我們為通話請求提出一個新穎的以議價式賽局為基礎的存取網路選擇機制，用以降低換手機率與通話阻塞機率。當有一個新的通話請求或是一個換手通話請求出現時，存取網路選擇機制就會被觸發。為了避免經常性的換手，除了考慮通話請求在一個網路中的停留時間，我們也考慮了通話請求的殘餘通話時間。首先，通話請求的候選存取網路集合會根據三個限制條件被決定出來。之後，與通話請求關聯的行動裝置分別與每個候選存取網路進行議價。在一個議價賽局中，雙方玩家基於他們的偏好來對候選存取網路的剩餘資源議價。從通話請求的觀點來看，為了有較好的服務品質，通話請求偏好獲取較多的資源。從候選存取網路的觀點來看，為了未來能夠服務更多的通話，候選存取網路偏好保留較多的資源。最後，通話請求的最終存取網路會根據議價的結果而被選擇出來。藉由所提出的基於賽局的存取網路選擇機制，通話請求能存取適當的網路，同時，進行中的通話也能得到保護。

As mobile devices and smart phones are popularized, the demand for data traffic is growing exponentially. Therefore, we are facing the challenge of increasing data transmission rates. In the next generation (5G) mobile wireless communications, small cells, multipacket reception (MPR), Device-to-Device (D2D) communications, etc. are considered as promising approaches to enhancing network throughput and spectral efficiency. Due to the low cost and rapid deployment of small cells, small cells are massively deployed in cellular networks to offload mobile data from overloaded cellular networks. Compared to the conventional collision model, an access point with MPR is capable of simultaneously decoding more than one packet from multiple concurrent transmissions. To increase the capacity of small cells, there is a need to design a medium access control (MAC) protocol with MPR for small cells. In D2D communications underlaying cellular networks, D2D users share the same resource with cellular users so that the spectral efficiency of cellular networks can be further improved. However, D2D communications would cause interference to cellular users. To reduce the interference from D2D users to cellular users, D2D users have to be allocated resource cautiously. Moreover, the interference among D2D users cannot be ignored. The next generation wireless networks are composed of essentially different types of networks. In such an integrated network, the topic related to network selection becomes significant. It would lead to frequent handoff or inefficient spectral usage if mobile users incorrectly access to an inappropriate network.

Small cells play a key role of relieving the loading of cellular networks. To improve capacity of small cells, we propose learning algorithms for a distributed medium access control with MPR in wireless networks. Based on the proposed algorithms, a node can dynamically tune its data transmission rates according to feedback information and the previous strategies used by other nodes. For symmetric networks, the proposed symmetric learning algorithms are designed to maximize network throughput. For asymmetric network, the proposed asymmetric learning algorithms are designed to seek a good balance between throughput and fairness. In the real network environment, a node cannot transmit data with an arbitrary data transmission rate since the number of available data transmission rates are finite. Hence, we propose methods to determine a finite number of the optimal data transmission rates. Compared to conventional collision model, the throughput of the proposed algorithms increases as the number of nodes increases.

In D2D communications underlaying cellular networks, source devices that are far away from each other should use identical resource to improve spectral efficiency, while source devices in proximity should use orthogonal resources to reduce interference from each other. On the basis of this principle, we utilize graph theory to calculate the amount of resources required for D2D communications. According to the amount of required resources, source devices are partitioned into subgroups. To reduce the complexity of optimizing throughput for a subgroup, we propose a fast approximation algorithm to calculate data transmission rates of source devices in a subgroup. Based on the proposed algorithm, the data transmission rates of the cellular devices that share the same resources with D2D devices are guaranteed. We adopt network information theory for D2D communications so that destination devices are capable of decoding multiple packets from multiple concurrent transmissions. Simulation results show that the proposed algorithm can reach a good approximation of the globe optimal throughput.

For heterogeneous networks, we propose a novel bargaining-game-based access network selection scheme for call requests in heterogeneous networks to decrease handoff probability and call blocking probability. The access network selection scheme is triggered when a new call request or a handoff call request arrives. To avoid frequent handoff, not only the sojourn time of the call request in a cell but also the residual life time of the call request is taken into consideration. First, the set of candidate access networks is determined for the call request according to three constraints. Then, the mobile device associated with the call request play bargaining games with candidate access networks one by one. In a bargaining game, two players bargain over the residual resource based on their preferences. On the perspective of the call request, the call request prefers to attain larger resource for better quality of service (QoS). On the perspective of the candidate access network, the candidate access network prefers to remain more resource in order to accept more calls in the near future. The final access network is selected based on the bargaining results. Via the proposed game-based access network selection scheme, call requests can access to appropriate networks, and meanwhile the QoS of ongoing calls be can be protected.