蜂巢式系統下的合併多播排程與功率配置機制

Abstract

通道時變、頻率選擇性衰減的特性，使系統可以透過動態的資源(時槽、頻寬、功率、位元)分配來增加系統吞吐量(throughput)，實現了多用戶分集(multi-user diversity)，多播(multicast)則是透過傳遞相同的內容給同一個群(group)內的多個使用者來增加頻譜效率，但同一個群內的使用者的通道狀況可能差異非常大，傳統的方法選擇同一個群內最小的通道增益作為子通道的通道增益，限制了系統的吞吐量，為了最大化頻譜效率，我們必須在多用戶分集增益與多播增益間取得適當的平衡。

除了系統吞吐量，公平性也是資源分配效能的重要指標，我們考慮正交分頻多重接取(Orthogonal Frequency Division Multiple Access)的無線多播系統，首先將不同的無線資源配置抽象化成三個較簡單的最佳化問題─單一群的鏈路調適(不同的使用者可以正確地收到那些資料片段?)、子通道指定(那些群的資料可以在那些通道上傳送?)、功率分配，並提出相應的演算法，接著將所有的資源分配抽象成一個最佳化問題，並結合前述的演算法達到使用者間的正比公平(proportional fairness)。

此外我們同時也考慮了更一般的狀況─單一使用者可以同時屬於兩個以上的群(i.e.同時接收兩份以上的資料流)。模擬結果顯示，我們提出的方法在系統的吞吐量與使用者間的公平性都有較佳的表現。

Due to the nature of time-varying and frequency selective fading, wireless communication system can increase system throughput via dynamic resource allocation (time slot, bandwidth, power and bit), exploiting multi-user diversity. Multicast deliver the same content to several users in the same group to enhance frequency efficiency, but the channel conditions may vary greatly between users of the same group. Conventional approach selects the worst channel gain as equivalent channel gain of group’s sub-channel, limiting the system throughput. Aim to maximum frequency efficiency, we have to make appropriate trade-off between multi-user diversity gain and multicast gain. In addition to system throughput, fairness is another performance index of resource allocation. Considering orthogonal frequency division multiple access (OFDMA)-based wireless multicast system, firstly, we separately formulate different resource allocation as three simpler optimization problem — Link adaption of single group (Which data fragments can be correctly received by different users?), sub-channel assignment (Which group’s data can be transmitted over different sub-channels?) and power allocation — and propose corresponding algorithms successively. Subsequently, we formulate all resource allocation as a joint optimization problem and combine it with aforementioned algorithms to approach proportional fairness among users.

Besides, we also consider more general case — One user may belong to more than two different groups (i.e., receives different data streams more than two simultaneously). Simulation results show that our proposed approaches have outer performance on system throughput as well as fairness among users.