Cost-efficient Deployment of Relays for LTE-Advanced Cellular Networks Yijie Wang, Gang Feng and Yide Zhang 2011 IEEE International Conference on Communications(2011 ICC) Advisor：Han-Chieh Chao Student：Weijen Yu Date ：2011/8/9

Outline Introduction System Model Cost-Efficient Deployment of Type I Relay Numerical Results and Discussions Conclusion

Introduction Relay is used in LTE-Advanced cellular networks to assist eNB (evolved Node Base-station) coverage extension and throughput enhancement. Compared to eNB, relay is advantageous on unit cost. Aim at analyzing the tradeoff deployment cost VS. the cell performance gain overall cell performance VS. satisfaction level for individual users. Two types of relay node (RN) have been defined Type 1:Extend signal and coverage Type 11: Improve QoS and raise system capacity 在佈建成本及效能增益取得平衡 在整個cell效能及user的滿意程度取得平衡

System Model A. System Infrastructure A network of 7 cells is modeled Omni antennas for direct link (eNB to UE link) and access link (RN to UE link), Directional antennas for backhaul link (eNB to RN link). Frequency reuse applies among multiple RNs and eNB Distance among RNs is large enough to avoid interference. 2*2 MIMO and advanced transmission schemes (SDMA) eNB can transmit simultaneously to the multiple RNs 在LTE-A裡，有一些因素會影響throughput 是用2*2 MIMO的天線加上的進階傳輸方法

B. Propagation Models Path loss models significantly affect system performance The detailed quantitative path loss values for three types of links can be found in Table II. 使用最新的propagation模型in [5]

C. Backhaul Subframe Allocation Use FDD mode Type I relay sends its own broadcast channel and synchronous channel in subframe 0,4,5,9. Due to RN’s half duplex characteristic subframe 0,4,5,9 cannot be assigned as backhaul downlink subframe for Type I relay to receive information from eNB. Backhaul Subframe Allocation也會影響到throughput 用4個Backhaul Subframe 消除backhaul 瓶頸影響

D. Interference Models Pi is the transmission power of the sender SINR at the i -th link in each receiving antenna of UE is calculated Pi is the transmission power of the sender k is the rank of MIMO channel matrix vi,k ( k =1, 2 …) is the singular values of i-th link’s channel matrix N denotes the noise power 因為RN之間可能會有干擾 User的接收天線在第 i個 link的SINR P是sender的transmission power K是矩陣的rank V是第 i個 link的channel matrix的奇異值 N是噪音

The capacity of these UEs is given by Calculate the achievable rate for i-th link by Shannon capacity formula The capacity of these UEs is given by where Cbackhaul and Caccess are the achievable rates for backhaul and access links respectively 用Shannon 來算 achievable rate

Cost-Efficient Deployment of Type I Relay An example of total cost per unit for different RAP(Radio Access Point) types is given in Table I Type 1Relay 也是一種小型的BS,管理自己擁有的UE

λb and λc are the eNB and RN deployment densities respectively. A. Iso-performance Curves for LTE-A Relay–enhanced Systems The linear cost model for total system cost C is where cb and cr are the total cost of an eNB and an RN respectively. λb and λc are the eNB and RN deployment densities respectively. 在LTE-A network裡，total cost包含eNB和RN的佈建

Examine the necessity of deploying RNs where λb,0 is the eNB density for pure eNB systems 在一樣的throughput 根據Iso-performance relay 密度增加 則BS密度減少 Iso-performance是一個理想的情況 若是frequency reuse 時 佈建relay 不一定會增加performance gain

B. Cost Efficiency with Type I Relays Optimal static deployment through investigating tradeoff between the deployment cost and the cell performance gain. The cost efficiency η defined where n is number of relays in a cell B is the cell system bandwidth C(n) is cell achievable rate with n relay deployed

C. Tradeoff Analysis of Progressive Deployment for Type I Relays Based on the Customer Satisfaction Index (CSI) used in economical theory Define relative user satisfaction level Sn,k of user k when n relays are deployed as where Cn,k is the throughput of user k when n relays in a cell. 接著考慮 Progressive Deployment , 當有預算限制時 network operator較喜歡逐漸地佈建RN 這是符合消費者需求方法，更多的relay可以幫助cell capacity和服務更多user 但是user可能會遭受到嚴重的干擾 因此降低performance 換句話說，需要在 system performance 和 single user performance level做取捨 依照經濟理論的Customer Satisfaction Index 我們定義了Average Customer Satisfaction Index為了達到cellular system capacity and average user satisfaction level. C是user k的throughput 在一個cell有n個relay時 C n-1,k =0 表示 user k是在佈建第 n個 才被服務到

Card() is the number of elements of a set. Use Kn to denote the number of users served in phase n Kn users can be divided into two set φn for users also being served in phase n-1 φ n’ for users initially admitted into LTE-A systems in phase n. Defined the ACSI of phase n to evaluate the effectiveness of progressive deployment ACSI where α = Card(φn)/Kn Card() is the number of elements of a set. ACSI的值是正的表示average user satisfaction level較好 隨著RN逐漸增加，ACSI的值會因為干擾而下降 因此在RN的佈建時要檢查ACSI是不是正數

Numerical Results and Discussions A. Simulation setting and parameters Simulation model in MATLAB The transmission power of eNB and RN is 46dBm and 37dBm respectively. 1.從Iso-performance curve 來觀察 RN密度 2.研究靜態佈建RN dependent on backhaul link distance 和 unit cost 3.對於逐漸佈建RN，會去看overall cell capacity and average user satisfaction level

B. Iso-performance Curves on Deployment of Type I Relay Relay 個數從1~12 因為當超過12會造成performance 下降 看到2條線相切的地方不會是在X軸 表示使用RN比較好，因為使用frequency reuse 但是用太多的RN不一定會得到較好的performance 固定X0 RN由Y1到Y0 這曲線也可以拿來預測 當希望capacity達到3*10平方MBPS時 eNB密度是0.5 則RN的個數上界為4個 多蓋就浪費了

C. Cost Efficiency η VS. Backhaul Link Distance d where R is the radius of the cell and Dbackhaul is backhaul link distance. 這張是看說eNB跟RN之間的距離 當D愈大則d越小 一開始Cost Efficiency 會隨著relay個數增加 到了高峰後 就隨著relay個數減少 不過太多的RN會讓成本持續增加但capacity並沒有增加 太多的RN也會因干擾而造成performance下降

D. Cost Efficiency η vs. the Total Cost Ratio r r=Cr/Cb 一個RN的cost大小，也會影響Cost Efficiency 分別在0.17跟0.33來看 在relay的個數等於7的時候Cost Efficiency 下降 r越小表示RN越便宜 r=0.1是Cost Efficiency 較大的

E. Tradeoff Gain for Progressive Deployment of Type I Relay 這事說明在用逐步佈建RN的時候user滿意度與relay個數的關系 1~4的時候ASCI都是正的，表示使用者對於目前佈建的滿意度大於下一個佈建的滿意度， 但performance gain是減少的 當ASCI是負的時候 意思是說滿意度變差 即使cell capacity增加 作者舉列 若一個網路營運商 每年都佈建一個RN來滿足使用者需求的話 最大的滿意度是 RN= 4 不過若是佈建到5.6.7的話 雖然會稍微降低滿意度 但能夠有較好的cell capacity 和有更多的使用者被服務

Conclusion Investigated the optimal deployment of RNs in LTE-A networks. The iso-performance curve for tradeoff between eNB density and RN density is not an ideal case in LTE-A networks. The optimal number of RNs is between 7 and 11 for static deployment. For the progressive deployment, the optimal number is 4 to achieve tradeoff gain between cell capacity and relative average user satisfaction level.