Hsiang-Hung Liu, Huan-Yi, Ho, Cheng-Fu Chou CONTENT-AWARE SPECTRUM AND POWER ALLOCATION IN TWO-TIER FEMTOCELL NETWORK To mitigate the co-tier and cross-tier interference in two-tier femtocell network, our works focus on the actual effects of interference on users’ perceived quality, which has much to do with the content of users’ application. In this paper, we tradeoff between interference and spectral efficiency and a victim-based algorithm is proposed to reuse as much resources as possible without severely impacting on other users’ quality. In this way, the resources could be allocated efficiently and improved the quality of users’ experience. Hsiang-Hung Liu, Huan-Yi, Ho, Cheng-Fu Chou

Introduction Popularity of smart phones and tablets Higher demand in wireless communication Provide service comparable to Wi-Fi networks in cellular data network Over 70% data traffic originate indoor[1] Coverage dead zones Because of the ever increasing popularity of smart phones and tablets, mobile data traffic in wireless network is also expected to grow exponentially. Moreover, more and more subscribers not only use their smart phones to make phone calls, but also run applications with higher demand for data rates, like streaming or online games, this will make more congestion in conventional cellular network. Although the Wi-Fi network is capable of providing good network capacity, the level of mobility and coverage may not able to be supported as the same level as the cellular network. As a result, cellular data systems will still have the need to provide service comparable to the Wi-Fi network. [1] G. Mansfield, “Femtocells in the US Market –Business Drivers and Consumer Propositions,” FemtoCells Europe, ATT, London, U.K., June 2008

Challenge Additional interference M-BSs  F-UEs F-BSs  M-UEs 那在這種two-tier的網路中 最大的challenge就是會產生一些額外的干擾( femtocell user equipment (FUE))(Macro Base Station MBS) macro users (MUEs)  femtocell base station (FBS) Although the advantages of femtocells(室內) are no doubt, there are still many technical challenges before they are widely deployed in existing macrocells. One of the major concerns is the interference problem. Since femtocells provide coverage using typical cellular system air interface, the co-channel interference may significantly limit the capacity improvement in this two-tier network F-BSs  F-UEs

Challenge Distributed nature Deployed by end users Limited information between macrocells and femtocells 除此之外 因為這些額外干擾通常跟femtocell所在的位置有很大的關係 但是基本上這些femtocell是使用者自己決定要不要deployed 所以在一個macrocell底下 對於femtocells的數量跟他們的位置是很難被掌握的 所以這種所謂Distributed的nature 會讓這些interference很難有效的被網路的operator控制 Furthermore, because FAPs are installed by customers, the location and the number of femtocells are not coordinated by operator. Due to this distributed nature, self-configuration and self-organization techniques are necessary for each femtocell to adapt itself into the environment. In order to efficiently take advantage of the femtocells, resources should be allocated to avoid persistent collision with neighboring cells. In addition, the content of application and capacity demand for each user also has much to do with the effects of interference.

Related Works Most of the existing works… Focus only on the interference between cells Not consider the interference effect to cell users

Motivation Although… But… When design resources allocation scheme… Femtocell introduces additional interference But… Interference may or may not be sensitive to users When design resources allocation scheme… Inter-cell interference Actual impact on the victims 所以我ㄇ的motivation就會在於說 在這種two-tier的環境底下 雖然會造成一些額外的干擾 但是因為使用者在需求上的一些差異性 這些干擾可能對有些user是很sensitive但也有可能對某些user造成的影響不是很明顯甚至根本是不會被察覺到的 所以我們認為在做resource allocation的時候 不只是要考慮intercell的interference 而真正要考慮到的是這些干擾對於被干擾的user有多少程度實質上的影響 Although Femtocell introduces additional interference But Interference may or may not be sensitive to users When design resources allocation scheme, we have to not consider only Inter-cell interference, but more important is what atual impact on the victims

Goal In the environment of interference… Reuse as more resources as possible Without severely impacting users’ quality Victim-based allocation scheme 所以我們的goal就是 在…盡可能的去對資源做reuse的動作 也就是說 在對被干擾者沒有太嚴重的影響下 用比較積極的方式去做resources allocation 那我們會採用victim based的方式 也就是像前面提到 我們是以使用者的觀點來看待干擾的嚴重性 也就是說干擾本身不一定是 一件嚴重ㄉ是 嚴重的是那些需求低的user去影響到需求高的user In this paper, we proposed a victim-based method, which reuses resources as many as possible while minimizing the impact on users’ perceived service quality.

Assumption Assume every user is running streaming application Watching different types of video TV news, action movie, drama , etc. Different capacity demand The value of capacity is different 那我們會先假設在系統中… 都是在使用streaming相關的應用程式 因為我們認為在這種每個仁觀看的影片類型不同的情況底下 干擾對每個使用者所造成的影響有比較顯著ㄉ差異 像前面的例子 看新聞… Take streaming application, which is popular in mobile subscribers for example, the influences on perceived video quality due to the channel condition has much to do with the type of the videos. For users watching videos with more actions or more dynamic such as sports game or action movie, the channel variation will significantly impact the video quality perceived by the viewers. As a result, we should consider not only the inter-cell interference but also users’ network content and capacity demand when designing the resources allocation scheme.

Algorithm v3:k5 u2 v1:k1 u1 u3 v2:k3 那因為B包含那些沒被victim所使用的subchannel     v3:k5 u2 v1:k1   那因為B包含那些沒被victim所使用的subchannel 所以我們會先對B做allocate的動作 且會用比較aggressive的方法 In this work, the purpose is to figure out a simple mechanism to assign system resources (sub-channels and transmission power) efficiently and we put the emphasis on the effects of interference to users instead of the interference itself. First, we separate all the available sub-channels into two sets. One contains those sub-channels which are being used by victims , and another one contains remaining sub-channels that are not holding by any victims. u1   u3 v2:k3

Algorithm For set B estimate PSNR gain of each user after assigning sub-channel with max power level greedily choose the best sub-channel to assign iterate till B becomes empty For these two sets, we use two similar ways to assign the sub-channels. Because the sub-channels in set B have less chance to interfere other users, we can assign these resources first and with a more aggressive way. This algorithm is executed iteratively. In every iteration, it firstly estimates the PSNR gain for every sub-channel in B under the situation that if it is assigned to each user with maximum power. After that, for each sub-channel, there will be a best user that can earn the most PSNR gain. Then finally, the algorithm greedily chooses the sub-channel with maximum PSNR gain and assigns it to the user causing that gain and then the sub-channel is removed from B. These iterations continue till the set B becomes empty.

Algorithm For set A estimate PSNR gain of each user and PSNR loss of victims after assigning sub-channel with all power levels greedily choose the best sub-channel to assign iterate till A becomes empty 那在集合A的部分 因為這些subchannel是被那些victim所使用的 所以我們正allocate之前 除了要去估計對user造成的PSNR gain以外 還要去估計他可能會對victim所產生的影響 已就是他們可能的PSNR loss 那除此之外 我們不能像前面依樣只考慮最大傳輸功率的情況 我們要比較小心地去考慮使用每一個power level所會造成的影響 In the case of set A, the sub-channels can be allocated in similar way but with the concern of the effects on the victims. This algorithm is also executed iteratively as the previous one. The difference is that the femtocell needs to estimate the PSNR gain for not only each user but also for each power level. Besides, the loss of video quality of victims should be deduced from the PSNR gain to discover the actual effect of allocating the sub-channels.

Simulation and Results Environment 10~90 femtocells overlaid under a macrocell Three different video types Parameter 12 sub-channels (0.18MHz for each) Power of MBS: 43dBm Power of FBS: 0~23dBm Number of femto users: 4 Number of macro users: 12

Simulation and Results Methods Split-band Reuse-1 Victim-based Victim-based with rate Cell-based Split band: 那在這個方式底下 macrocell跟femtocell是使用互相orthogonal的頻帶 而在reuse1則是相反 所有的cell都會去共用所有的資源 那victim based則是我們所提出的方法 Victim-based with rate的部分也是使用victim based的概念 只是他去maximize的是每個user的傳輸速率 那最後一個cell based的部分是像前面提到以cell的觀點去計算干擾的程度 也就是可能當Cell跟cell比較近的時候他就會認為之間干擾的機會很大 而去使用比較orthogonal的方式做channel assignment Reuse-1: In this scheme, all the macrocells and femtocells share the total available sub-channels. Split-band: In this case, femtocells use orthogonal set of sub-channels to macrocells. Specifically, for the femtocell is located at the outer region of macrocell, it only uses the inner band, and vice versa. Cell-based: The sub-channel assignment is based on the received pilot signal power of neighboring cells. The femtocell will try to allocate orthogonal subchannels to the cells with larger received power of pilot signal. Victim-based with rate: This scheme runs the algorithm similar to the one we proposed but it focuses on maximizing the transmitting rate.

Simulation and Results All users are randomly watching one of the three video types all users 那接下來是看到系統中所有使用者的psnr值加總 那在這部分可以看到 在兩種最naïve的方式底下 因為較低的spectrum efficiency和比較嚴重的彼此間的干擾 造成雖然在femtocell deployed數量變多的情況下 對系統的效能到了60個femtocell之後 就沒辦法在有明顯的improve 那我們方法 隨者deploy的femtocell變多 也還能因為這些femtocell本身所帶來的好處 而繼續對系統的performance有正向的影響 同時跟其他比較保守的方法比起來 也能比較有效率的運用這些資源

Simulation and Results Difference of PSNR sum compared to split-band 所以針對我們的方法跟其他方法之間的差距 我們就用這張圖來看 那他的橫軸會是不同的環境 最左邊是所有使用者都觀看同種類型的影片 那中間的看三種影片類型中最動態跟最靜態的影片中的一種 最右邊則是隨機觀看三種類型中的一種 那縱軸是其他方法 針對系統中user psnr的總和跟split-band這種方法之間的差距 可以看到當使用者都看同種類型影片的時候呢 因為user彼此之間沒有太明顯的差異性 所以我們的方法跟其他方法比較起來比較沒有太大的差距 但是當使用者的diversity變高的時候 就可以看到差距有變多的趨勢 那在最極端的情況底下 也就是當使用者要馬是看最動態或者是看最靜態的影片的時候 會有最大的差距 那主要就是因為如果提供給那些需求低的使用者過多的保護 而讓需求高的人沒辦法運用這些資源的情況下 會造成一些資源上的浪費

Conclusion Consider the actual effects of interference on users Trade off between interference and spectral efficiency Victim-based scheme Reuse as much resources as possible without severely impacting on users quality To mitigate the co-tier and cross-tier interference in two-tier femtocell network, our works focus on the actual effects of interference on users’ perceived quality, which has much to do with the content of users’ application. In this paper, we tradeoff between interference and spectral efficiency and a victim-based algorithm is proposed to reuse as much resources as possible without severely impacting on other users’ quality. In this way, the resources could be allocated efficiently and improved the quality of users’ experience. The simulation results proved that proposed mechanism increased the users’ watching video quality than other more conservative schemes and the gap would be even larger as the users’ content became more diverse.