Rock Street, San Francisco

\author{ Mahmoud~Mandour, Fayez~Gebali, Ashraf~D.~Elbayoumy, Gamal~M.~Abdel~Hamid \IEEEcompsocitemizethanks{ \IEEEcompsocthanksitem M.\ Mandour is a PhD Student in the Department of Electrical and Computer Engineering, University of Victoria, Victoria BC, V8W 3P6, Canada. E-mail: [email protected] \hfil\break \protect \IEEEcompsocthanksitem F.\ Gebali is with the Department of Electrical and Computer Engineering, University of Victoria, Victoria BC, V8W 3P6, Canada. E-mail: see http://www.ece.uvic.ca/$\sim$fayez/info/contact.html \hfil\break \protect \IEEEcompsocthanksitem A.\ Elbayoumy is with the Department of Communications, Military Technical Collage, Kobry El-Kobbah, Cairo, Egypt. E-mail: [email protected] \hfil\break \protect \IEEEcompsocthanksitem G.\ Abdel~Hamid is with the Department of Communications, Military Technical Collage, Kobry El-Kobbah, Cairo, Egypt. E-mail: [email protected] }} \markboth{ Enhanced Handover Algorithm in LTE Femtocells Network } %IEEE Transactions on Parallel and Distributed Systems}% {Shell \MakeLowercase{\textit{et al.}}: Bare Demo of IEEEtran.cls for Computer Society Journals} \IEEEtitleabstractindextext{% \parbox{\linewidth}{\emph{ \begin{abstract} Data rate and Coverage are the main secret factors for customer satisfaction. Today’s, small cell technology as femtocell can achieve the upcoming demand for higher data rate for cellular networks as well as can extend the coverage area. Femtocells are known as low cost, low power and efficient way to boost mobile signal in indoor or in areas with little mobile coverage. Deploying femtocells networks results in more frequent initiation of a handover (HO) procedure. The reduction of unnecessary handovers challenges is getting more considerable while more femtocells are used in the network. We propose a novel HO algorithm to eliminate the redundant handovers in Long Term Evolution (LTE) femtocells network and to nominate the most proper target femtocell for HO among many candidates. This algorithm based on the transition probability matrix to predict the user movement using Markov Chain equations. Toward this goal, we are considering Reference Symbol Received Power (RSRP) and Reference Symbol Received Quality (RSRQ), User Equipment (UE) moving direction and in which zone the UE inside the femtocell as decision criteria. \end{abstract} }} \begin{IEEEkeywords} Femtocell, Markov Chain, Handover, RSRP, RSRQ. \end{IEEEkeywords} } \maketitle \IEEEdisplaynontitleabstractindextext \IEEEpeerreviewmaketitle %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \section{Introduction} \label{sec.intro} Recently, a huge progression done in the deployment of cellular networks. LTE and LTE Advanced (LTE-A) technologies are the most tremendous technologies for future deployment of cellular networks and wireless communication system. The 3rd Generation Partnership Project (3GPP) is a communion between groups of telecommunications standards associations, have been inovated for of (2G, 3G, 4G, and 5G) cellular networks, to meet the aims of high-speed data communications networks \cite{ahmed_j17}. In the mobile network, whenever a UE moves from one cell to another, the call needs to be handed over to another base station, and network resources must be reallocated. HO technique can be defined as allowing the connected UE’s with Evolved Node-B (eNB) to be handed-off to the next eNB without any disconnection \cite{miyim_j14}. HO is one of the remarkable strategies in cellular networks. Therefore, seamless HO is responsible for mobility, user reliability, and high quality-of-service (QoS). Fast seamless HO should occur, especially when the UE starts moving from the serving cell to the target cell \cite{hasbollah_j16}. During the HO procedure, the reason for handover latency is the time taken for resource allocation. In order to fulfill a fast and seamless HO process, the handover latency needs to be minimized. Predicting the next location of the User Equipment (UE) is one of the most important techniques that used to reduce this latency and the delay in resource allocation. Mobility prediction detects the identity of the target cell for resource reservation before starting the actual handover procedure \cite{duong_j12}. On the other hand, LTE/LTE-A supports different kinds of deployment cells in the cellular network (i.e. Macro, Micro, Pico or Femtocell), where used to circulate the coverage area depending on the area. Basically, (Macrocell) is the largest coverage area and used for outdoor subscribers or UEs, it is known as (eNB) in the LTE networks. So, whenever the coverage of macrocell starts decrease imply that (microcells to femtocells) founded. Actually, those are used to increase the capacity and coverage for the indoor subscribers \cite{salman_c14}. Recently renowned that the indoor mobile users are the origin of most networks traffic. However, in modern buildings, service providers hardly trying to provide high-quality service on-the-go for indoor subscribers, but many users facing problems with poor indoor connectivity due to the signal attenuation \cite{ariffin_j13}. There are many driving factors behind deploying LTE and LTE-A networks. One of the most important methods of boosting celullar network capacity is by shrinking the cell size. So, the need for a technology such as femtocells is mandatory to achieve high capacity, high data rates, signal quality, and low latency for multimedia services indoor or in shadowed areas. Also, to increase throughput in areas with a high density of users, and to offload a large amount of capacity from the network \cite{ulvan_j11}. The LTE Femtocells represented as Femtocell Access Point (FAP) and also known as Home Evolved Node-B (HeNB). Figure~\ref{fig.Femtocells} shows an example of today’s and future’s deployments everywhere. FAP’s, are supposed to be deployed especially (e.g. in households, in offices, schools, universities and in shopping malls .. etc). The FAPs is connected to the service provider’s backbone via a wired line such as Digital Subscriber Line (DSL) or optical fiber. \begin{figure}htbp! \begin{center} \includegraphicswidth=8.5cm{../figures/Femtocells} \end{center} \caption{ \label {fig.Femtocells} Femtocell deployments networks. } \end{figure} %%%%%%%%%%% %%%%%%%%%%% Overall, the paper is arranged as follows. Section II describes the related work on HO management in mobile communication networks, while Section III illustrates the background of LTE femtocell network. Section IV contains the details of modeling the system and all the assumptions. We present and discusses the result of the study of network performance evaluation in Section V and finally, Section VI summarizes the conclusion and further work. %%%%%%%%%%%% %%%%%%%%%%%% \section{Related Works} \label{sec.review} Recent studies aimed to improve the HO performance by alleviating the unnecessary handovers and the probability of dropped calls. Also by reducing the HO delay and reserve in advance radio resources by predicting the next handed-over target cell (i.e. either eNB or HeNB). There are many surveys shows different solutions for this issue; one of them by using prediction and creating a profile based on user mobility history, the other solutions based on Markov chain or based on user’s location and RSRP. In \cite{chowdhury_c11} a tentative study was done to create a list with a minimum number of neighbor femtocells to start the femtocell-to-femtocell HO in a dense homogeneous network. The proposed method count on the RSRP and the operating frequency of each femtocell are then defined. So, any femtocell that uses a frequency as the serving FAP is rejected. As presented in \cite{si_c10}, the authors proposed a Hidden Markov Model (HMM) mobility predictor way to enhance the performance of macro-cellular networks. The network deployment as nodes represent macrocells and at the edges representing neighboring relationships. Based on the dynamic data of UE’s and HMM, the next macrocell is identified. The authors in \cite{ge_c09}, have proposed a prediction scheme based on the history of UE mobility. The network will recognize the UE whom visit the cell frequently, and then track and records the movement data. Using this information and the position of the user, the network can find a route of the user and the RSRP is considered. This technique minimizes the number of HO’s in LTE systems. The work in \cite{ulvan_c09}, propose a prediction technique based on the Markov Chain. This method using the path prediction to predict the target cell for the users. Position and velocity of the users are the main factors helps to expect the heading next cell for the user. In this technique, the author assumed that UE can send its location to the serving cell. Simultaneously; the serving cell is able to keep a database of the network coverage. Based on Markov Process, the UE movement can be predicted. In \cite{zhang_c10}, the authors use the UE mobility history that provides frequent position and time spend on particular places. Then, using the time as an input to the transition probability matrix. After that, this matrix used to predict the UE movement by Markov Chain equations. In \cite{kim_c07}, a position-based path prediction scheme are used. Using a Markov Chain, the user movement has been predicted before the proper handover strategies are done. \section{Background of LTE Femtocell Network} \label{sec.femtocells} Femtocells can be defined as small, plug-and-play, and low-power base stations. They are designed for home/enterprise usage and provide short-range coverage over the licensed frequency band. Femtocells offer a money-wise approach for improving the spectrum efficiency without the need to upgrade the infrastructure of the carrier network. It works as a major part of the indoor cellular network infrastructure to get higher QoS. Generally, LTE is designed to provide connectivity between the UE and the Packet Data Network (PDN). LTE was known as an IP-based and flat core network architecture. Figure~\ref{fig.LTE_architecture} shows a simple LTE network architecture. There are two areas of the network at the highest level that we can consider for LTE. Those would be the Evolved Universal Terrestrial Radio Access Network (E-UTRAN) and the core (EPC) which is the Evolved Packet Core, combined together to form the LTE network. \begin{figure}htbp! \begin{center} \includegraphicswidth=8cm{../figures/LTE_architecture} \end{center} \caption{ \label {fig.LTE_architecture} LTE network architecture } \end{figure} In E-UTRAN we encounter the eNB (macro base station) and HeNB (femto base station), which responsible to manage the air interface and look-after the scheduling, and handover. The architecture may deploy a HeNB Gateway (HeNB GW), to organize the large number of HeNBs. The eNBs/HeNBs are inter-connected by the X2 interface, and the S1 interface is between the E-UTRAN and the EPC. As we move closer to the network core, in EPC we encounter the first packet core device the Serving Gateway (S-GW). The SGW is responsible for user plane (U-plane) during handovers process, and also organize the data packets for the user. (i.e. it utilize standard Internet security protocols to authorize, authenticate and encrypt the femtocells traffic between the HeNBs and the network operator’s core network). Moving further to the core we encounter the Packet Data Network Gateway (PDN-GW), and this gives UE access to external packet data networks (PDN) and assigns IP addresses to the UE and QoS fulfillment. And the controlling element of the LTE network is the Mobility Management Entity (MME), This device looks after mobility management and session management. (i.e. it is responsible for control plane functions (C-plane). Supporting the MME with Home Subscriber Server (HSS), acting as a central database that contains user-related and subscription-related information. The HeNB shall acts as the eNB in terms of function supported and the procedures run between the HeNB and the EPC \cite{3GPP_ts16}. \subsection{LTE Femtocell Handover Scenarios} %%%%%%%%%% There are three possible handover scenarios in the femtocell network: \be \item Hand-in: represents the UE connection is handed over from the eNB to the HeNB. \item Hand-Out: represents the handover performed from the HeNB to the eNB. \item Inter-HeNB: represents the handover scenario done between the HeNBs. This handover is similar to Hand-in scenario. \ee \subsection{The Handover Types} As shown in fig.~\ref{fig.HO} LTE/LTE-A networks are supporting two different types of HO schemes: \be \item Horizontal HO (homogeneous): used in the HO between the same type of network cells. \item Vertical HO (heterogeneous): used in the HO between different types of network cells \cite{barja_j10}. \ee \begin{figure}htbp! \begin{center} \includegraphicswidth=9cm {../figures/HO} \end{center} \caption{ \label {fig.HO} Handover types (Vertical and Horizontal). } \end{figure} HO mechanism should accept mutation between the homogeneous and heterogeneous of cellular networks, and also should be incorporated with different radio access networks (RAN), such as (LTE, WiMAX, WiFi, UMTS, WLAN, ..~etc). Moreover, the preceding types of HO mechanisms are done after passing by three essential phases: measurement reports, taking the decision, and HO execution \cite{zhou_j14}. In the measurement reports phase, all information RSRP/RSRQ related to UE and network cells are gathered for the HO triggering mechanism. Then, this data is used to evaluate the next hop (target) cells in the second phase (taking a decision). After choosing it, a fresh link is established in the last phase (HO execution) where all resources are transferred from the served cell to the target cell \cite{yaz?c?_m14}. \subsection{Handover Decision} Handover decision is one of the most challenging parts in the handover call flow. The handover decision based on some common system metrics included in the measurement report provided by the UE. The main parameters are: \be \item RSRQ … Reference Signal Received Quality. \item RSRP … Reference Signal Received Power. \ee Both are used to choose the best target cell for the UE. The RSRQ provides more information about the interference level at the location. Furthermore, RSRQ is more favorable than RSRP for the dense HeNB’s network, where the interference level is high. In LTE network, the UE every 200 ms should update two parameters on the reference signal RSRP/RSRQ . Basically, we should discuss how the conventional (classical) HO occurred. The classical HO scheme for wireless and mobile networks show up two styles of threshold area in the coverage of the cell. The handover threshold and the exit threshold can be imagined as two different hexagons with an irregular radius around the cell core. The inner hexagon with small radius refers to exit threshold hexagon, and the outer hexagon with larger radius refers to the handover threshold hexagon \cite{kirsal_j16}. Nevertheless; the outer threshold hexagon is the beginning of the HO process. So, for complete a successful mobility, any UE has to end the handover process before finishing the handover threshold hexagon area. If the UE is not handed-over efficiently before this threshold area, the UE call will be blocked (i.e. dropped) \cite{ever_j15}. \section{System Model for Mobility and Handover} \label{sec.model} We consider our proposed system shown in figure~\ref{fig.HEXA}, as a homogeneous cellular network consists of many merged femtocell stations and the UE is moving randomly inside the network. Where every UE inside the network has a unique ID and neighbors depending on the UE serving cell and its type (corner/edge). \begin{figure}htbp! \begin{center} \includegraphicswidth=6cm{../figures/HEXA_Numbers} \end{center} \caption{ \label {fig.HEXA} Geometry of hexagonal grid for cellular-femtocell network. } \end{figure} Figure~\ref{fig.HEXA} shows the architecture of our cellular-femtocell network, which has rings around the central hexagon. The darker tone represents a corner hexagon, and lighter tone represents an edge hexagon. In network modeling, we employ the following schema: \be \item We assume 60 femtocells, the numbering sequence follows a “spiral” path such that the numbers of hexagons per ring are multiple of 6, e.g. first ring 1-6, second ring 7-18, third ring 19-36, and the last ring 37-60. \item Two femtocells can be adjacent by being neighbors along the spiral-ling path indicated above, in which case their IDs are also consecutive. In this spiral-ling counting pattern, we call the femtocells such as 7, 19, 37, … (gate cells) where the path jumps to an outer larger circular ring. \item In figure~\ref{fig.HEXA}, all hexagons were labeled using a single non-negative integer $ID$, $0$~$\leq$~${ID}$~$\in$~${Z}$; or, equivalently, using the ring number $R$, $0$~$\leq$~${R}$~$\in$~${Z}$, and the user site within the ring $S$, $0$~$\leq$~${S}$~$\in$~${Z}$, where \begin{eqnarray} ID=\begin{cases} 0, ; \text{$R = 0$} \\ 3R(R – 1)+1+S, ; \text{$R \ge 1$} \end{cases} \end{eqnarray} or inversely, \begin{eqnarray} R=\begin{cases} 0, ; \text{$ID = 0$} \\ \lfloor ((3 + \sqrt(12*ID-3))/6) \rfloor, ; \text{$ID \ge 1$} \end{cases} \end{eqnarray} where, $\lfloor \rfloor$ denote rounding towards zero, and \beqa S =ID – 3R(R-1)-1 \label{eq:20} \eeqa \item A special case at ring $R=0$, there is only one hexagon (the white hexagon in center). It has six hexagons neighbors are located in ring $R=1$. \item At ring $R=1$ is kind of special because it consists of only corner hexagons, but it turns out the same rules below handle this ring as well as all outer rings just fine. \item All rings except ring $R=0$ have $(R-1$) edge hexagons between consecutive corner hexagons. Particularly, each outer ring has one more edge hexagon between corner hexagons: which allows to compensate the $P$ indexing along the ring and in any other rings. \item UE located inside any ring, where R = $(0)$ to $(R-1)$ have 6-neighbors, except the (last) outer ring has special conditions. \item As shown in figure~\ref{fig.HEXA}, if the user located in the outer ring, first we have to check the type of cell either (corner/edge). For the user being served by outer edge cell, it has 4-neighbors and only 3-neighbors for the one being served by an outer corner cell. \item We assumed that we have 6-directions rounded counterclockwise with the first ring (R=1) which labeled as direct 1 for cell ID 1 to the direct 6 for cell ID 6. \item The main direction is known as the Top Corner direction which has the IDs of (1, 8, 21, 40, .. etc). \item Whenever UE moving to the boundaries of any direction inside the network (i.e. Not Valid destination cell) which located in the outer ring, the moving direction will be laterally inverted, such as mirror-like reflection. The outer ring acts Specular Reflection. \item All characteristics with general forms of our proposed hexagonal femtocells network are listed in the table (\ref{table.hex}). \ee \begin{algorithm}htbp! \caption{Pseudo code for assigning station types.} \label{alg.types_of_stations} \begin{algorithmic}1 \small \REQUIRE $ID$, $R$, $S$, $Total\ Femtocells$, $Corner$, $Edge$ \ENSURE $Station\ Type$ \STATE $Total\ Femtocells \leftarrow N$ \% Initialization step \IF {$ID = 0$} % the center hexagon \STATE $R \leftarrow 0$; \STATE $ID.Corner \leftarrow 0$; \STATE $ID.Edge \leftarrow 0$; \ELSIF {$ID = 1:6$} % the 1st ring (R=1) \STATE $R \leftarrow 1$; \STATE $ID.Corner \leftarrow 1$; \STATE $ID.Edge \leftarrow 0$; \ELSIF {$ID = 7:Total\ Femtocells$} % R;1 rings \STATE $R \leftarrow floor((3 + sqrt(12*ID-3))/6)$; \STATE $S \leftarrow ID-3*R*(R-1) -1$; \STATE $modulo \leftarrow mod(S,R)$; \ENDIF \WHILE {$modulo$} \IF {$modulo = R-1$} \STATE $ID.Corner \leftarrow 1$; \STATE $ID.Edge \leftarrow 0$; \ELSE \STATE $ID.Corner \leftarrow 0$; \STATE $ID.Edge \leftarrow 1$; \ENDIF \ENDWHILE \RETURN $Station\ Type$ \end{algorithmic} \end{algorithm} Algorithm~\ref{alg.types_of_stations} shows how to assign station type in our network. We define the input parameters as (ID, R, S, N, Corner, and Edge), an output parameter is the Station Type. %new call blocking probability and handover call blocking probability. \begin{table*}htbp! \caption{ \label{table.hex} Characteristics of hexagonal femtocells. } \begin{center} \begin{tabular}{*{6}{p{20mm}}} \hline \textbf{Ring } ; \textbf{\#edge cells per sector} ; \textbf{\#cells on ring}; \textbf{Start ID} ; \textbf{End ID} ; \textbf{Top Corner ID} \\ \hline 0 ; 0 ; 1 ; 0 ; 0 ; 0 \\ 1 ; 0 ; 6 ; 1 ; 6 ; 1 \\ 2 ; 1 ; 12 ; 7 ; 18 ; 8 \\ 3 ; 2 ; 18 ; 19 ; 36 ; 21 \\ 4 ; 3 ; 24 ; 37 ; 60 ; 40 \\ $\vdots$;$\vdots$;$\vdots$;$\vdots$;$\vdots$;$\vdots$ \\ $R$ ; $R-1$ ; $6R$ ;$3R(R-1) + 1$;$3R(R+1)$ ;$R(3R-2)$\\ \hline \end{tabular} \end{center} \end{table*} \begin{figure}htbp! \begin{center} \includegraphicswidth=8cm{../figures/network} \end{center} \caption{ \label {fig.network} Proposed LTE femtocells cellular network. (a) Network structure. (b) Detailed coverage areas of two neighbor femtocells } \end{figure} Figure~\ref{fig.network} shows the femtocells network being studied in details, where all femtocells are modeled as a hexagon shape, overlapped coverage and with a HeNB located at the center. Figure~\ref{fig.network}(a), shows the general geometry of the femtocells under study (central blue hexagon) and one of the six neighboring femtocells (top blue hexagon). Figure~\ref{fig.network}(b), shows the two adjacent femtocells that will be explained in details. We employ the following assumptions: \be \item Assuming that our indoor cellular network has 60 femtocells overlapped arranged and labeled in a specific sequence (will be discussed in the consecutive sections). \item We assume that every two neighboring femtocells are divided into three zones, we use subscripts (0, 1, and 2) to indicate the femtocell zones in both (outgoing/incoming) directions as: \\ Zone 0 represents the core of the source femtocell. \\ Zone 1 represents the edges between the source femtocell and neighboring target femtocell at the top. (i.e. HO region in grey color). \\ Zone 2 represents the core of the target femtocell. \item Each femtocell with its neighbors have six identical queues located in HO region, one queue per each hexagonal sector. \item We consider that there are many users with active calls or not. They are allocated inside the core of HeNB and the moving UE considered to be handed-over from the served HeNB to the next target HeNB. (i.e passing form (zone 0 to zone 2) direction or vice versa). \ee As depicted in figure~\ref{fig.network}(b), we investigate the queue exist in the HO region (grey hexagons area) between the two neighbor femtocells. Queuing analysis is one of the most important methods for studying wireless communication systems. It helps to get the truth of many inquiries about the system performance. %%%%%%%% %%%%%%%%% The basis of the approach is to propose a novel HO scheme that enhances the performance of the LTE femtocells network and optimizing the HO process. Since the number of users in the queue depends on its immediate past state of the queue only, our system can be modeled using Markov chain Models. % $M^m/M^m/1/B$ queue type (Multiple-Birth Multiple-Death) is used. So, at any time step $t$, most multiple calls could arrive and at most multiple calls could leave. (i.e the queue size, at each time step $t$ can increase by more than one, and also decrease by more than one) \cite{gebali_b15}. %%%%%%%% %%%%%%%%%% %%%%%%%%% %%%%%%%%%%% \begin{figure}htbp! \begin{center} \includegraphicswidth=6cm{../figures/queue} \end{center} \caption{ \label {fig.queue} The queue in handover region between two neighbor femtocells } \end{figure} Figure~\ref{fig.queue} shows the two neighbor femtocells with a queue in the gray handover region. Our proposed algorithm considering two types of calls: the incoming calls have two types are $(a, b)$, and the outgoing calls have two types are $(c, d)$. We have a number of users per sector considered as $n$. The queue size $B$ can define as the waiting area for the UE as long as they have not been served yet, also the UE that is being served. In figure~\ref{fig.queue}, we employ the following simplifying assumptions: \be %\item The Markov Chain states represent the number of users in the handover queue. \item Each femtocell is further divided into two parts, the core, and the edge areas. The core area is in yellow color, and the edge area where merged or overlapped with the target neighbor HeNB is in gray color. \item We use a column stochastic matrix that operates sequentially and it represents the transitioning from one state to another. \item The incoming call types defined as: \\ ${a}$ = represents the probability of a new call initiated inside the handover region, which is equivalent to a birth event or an increase in the queue population. \\ ${b}$ = represents the probability of incoming call comes from the core of serving HeNB to be handed-over to the target HeNB. \\ Also, we consider the probability of no incoming call enters the queue is $(1-b)$. \item And, we have two different types of outgoing calls leaving the queue in the handover region. These two types of outgoing calls are defined as: \\ ${c}$ = represents the probability of that new call initiated (${a}$), is terminated. So, it is known as an outgoing terminated call. \\ ${d}$ = represents the probability of that incoming call (${b}$) comes from the core of serving HeNB, is successfully handed over to the target HeNB. So, it is known as an outgoing handed-over call. \\ Also, we consider the probability of no outgoing HO call leaves the queue is $(1-d)$. \item We assume, ${n}$~$\leq$~${B}$ to warrant that the number of incoming users would not exceed the queue size ${B}$, where ${n}$ is the number of user’s per sector. \item As a basic condition, if the number of users ${n}$ is greater than the queue size ${B}$ would imply the next user will lose the call (Blocked Calls). \item Also, we assume: ${N}$ = represents the total number of user’s per HeNB. and, ${k}$ = represents the number of active calls. \item The probabilities of ${k}$ calls started, are given by: \beqa p_a (n,k) = \binom{n}{k} a^k (1-a)^{n-k} \label{eq:1} \eeqa And, the probabilities of ${k}$ calls ended, are given by: \beqa p_c (n,k) = \binom{n}{k} c^k (1-c)^{n-k} \label{eq:2} \eeqa We assume that, whenever $n$ \$