J Sign Process Syst DOI 10.1007/s11265-015-1061-x

Towards 5G: Context Aware Resource Allocation for Energy Saving Muhammad Alam1 · Du Yang1 · Kazi Huq1 · Firooz Saghezchi1 · Shahid Mumtaz1 · Jonathan Rodriguez1

Received: 14 April 2015 / Revised: 18 September 2015 / Accepted: 7 October 2015 © Springer Science+Business Media New York 2015

Abstract With the objective of providing high quality of service (QoS), 5G system will need to be context-aware that uses context information in a real-time mode depends on network, devices, applications, and the environment of users’. In order to continue enjoying the benefits provided by future technologies such as 5G, we need to find solutions for reducing energy consumption. One promising solution is taking advantage of the context information available in today’s networks. In this paper, we take a step towards 5G by utilizing context information in the scheduling process as conventional packet scheduling algorithms are mainly designed for increasing throughput but not for the energy saving. We investigate a Context Aware Scheduling (CAS) algorithm which considers the context information of users along with conventional metrics for scheduling. An information model of context awareness along with a context aware framework for resource management is also pre-

 Muhammad Alam

[email protected] Du Yang [email protected] Kazi Huq [email protected] Firooz Saghezchi [email protected] Shahid Mumtaz [email protected] Jonathan Rodriguez [email protected] 1

Instituto de Telecomunicac¸o˜ es, University of Averio, Aveiro, Portugal

sented in this paper. CAS is simulated applying a system level simulator and the results obtained show that considerable amount of energy is saved by utilizing the context information compare to conventional scheduling algorithms. Keywords 5G · Context information · Scheduling · Energy efficiency

1 Introduction In 5G Era, we need novel methods of abstraction to efficiently generate context-aware information, as well as new ways to share context information among applications, networks, and devices. In this sense, wireless systems play a key role as context aware enablers, as well as high capacity backhaul systems. This is provided by the unrelenting motivation in the capacity increase and latency decrease, especially in the case of future 5G systems. Alternative views applied to “context” lead to different definitions and different levels of applicability. In case of wireless networks the context is categorized into two basic categories, UE and network related context [1]. Contrary to the conventional scheduling mechanisms, there are a number of context information available related to user equipment (UE) that can be utilized in resource management based on the motivation and scenarios e.g. transmit power, battery level, mobility, traffic type etc. In order to focus on energy saving and QoS of UE, battery level along with traffic type and channel condition of the each UE are considered in our proposed scheduling algorithm. Energy efficiency (EE) and low carbon strategies have attracted a lot of concern in the recent years. Driven by the rapidly increasing demand of high data-rate, the throughput of today’s wireless system has dramatically improved over

J Sign Process Syst

the last few decades. In the recent years, there is a need to improve both the spectral efficiency as well as the Energy EE, since the energy consumption has become an important issue from both economic and environmental aspects; and future devices will be more power hungry than connectivity. The EE can be tackled through the exploitation of techniques and mechanisms from application to physical layer. We utilize the context information in scheduling process for energy saving in Long Term Evolution Advanced (LTEA). Round Robin (RR), Maximum Carrier to Interference ratio (Max C/I) and Proportional Fair (PF) scheduling are the three conventional and most popular scheduling methods. RR allocates equal resources to all users, regardless of their current channel condition. On the other hand, Max C/I scheduling aims at maximizing the total cell throughput by considering CQI values fed back to evolved NodeB eNB from the UEs. This leads to unfairness, as users that are further away from eNB (or have bad channel conditions) will not be allocated a fair share of the radio resources. The PF algorithm, [2], tries to provide fairness by increasing the priority of a mobile user who has a relatively low value of the C/I ratio. LTE and LTE-A aim to provide customers a new mobile experience providing higher data rates that makes user to use more bandwidth demanding application anywhere anytime e.g. video streaming, interactive gaming etc. However, these applications are highly energy demanding and drain out the limited battery of mobile devices which is a major challenge for modern telecommunication systems. On one hand, this challenge makes users more reluctant to use the high bandwidth demanding applications but on the other hand, it encourages the development of new architectures and mechanisms that are more power-aware or powerefficient and contribute to the energy savings of modern mobile devices. Keeping in view the power saving, LTE uses the idea of Discontinuous Reception (DRX) and Discontinuous Transmission (DTX) which makes the mobile devices aware of unnecessary and continuously monitoring the control channels and turns the radio to an extended sleep time and activate on defined time intervals. But these mechanisms only extend the sleep time to contribute to energy savings and do not considers the actual battery level of mobile devices and other important UE related context information in resource management for energy savings. There are situations when the UE’s battery level is low but gets no priority in the scheduling process. Therefore, in this paper, we have tackle this problem by introducing new metric in the resource management which contribute to the battery savings of UE while providing the same QoS to UE. The major contributions of this paper are as follows: •

A detailed context architecture and framework for context based scheduling algorithms. The framework can

• •

•

exploit any context information related to UE and eNB to achieve the desired goals based on the proposals for 5G. A detailed design and working of context information based signaling in LTE-A. A context base battery priority scheduling algorithm for the low battery devices in congested scenarios where UE has limited access to recharging and utilizing high data rate demanding applications. Implementation of context information based module in system level simulator for LTE-A.

Most of the traditional schedulers consider only the throughput but not the energy related information in the scheduling; therefore, we go beyond the state-of-the-art and develop a new scheduler which exploits the context information of UE for energy saving and QoS. The rest of the paper is organized as follows: in Section 2, we present the related work, Section 3 covers the scenario, the detail description of problem formulation is provided in Section 4 while the proposed scheduling algorithm along with the pseudocode and flowchart is presented in Section 5. Section 6 gives details about the context based framework for scheduling, Section 7 details the representation of the context information followed by the the acquisition of context information in Section 8. The details of the system level simulator along with simulations results are presented in Section 9 and finally, we concluded the paper in the last section.

2 Related Work When the mobile devices are powered on unnecessarily for an extended period of time they consume useful battery, which is considered one of the main reasons for energy consumption in both infrastructure and ad hoc networks. This problem is tackled by the introduction of proper sleep or idle modes of the mobile devices which is reported in [3, 4]. Therefore, to take advantage of the sleep or idle modes for energy saving 3GPP [5] has standardized discontinuous Transmission (DTX) and Discontinuous Reception (DRX). Similarly, in [6] the study evaluates several different parameter settings for LTE’s DRX, and attempts to discover a reasonable trade-off between VoIP performance and user terminal battery life. But these mechanisms contribute to energy saving only by extending the UE sleep time while ignoring the consideration of the context information in the scheduling process. For instance, to guarantee the QoS for real-time flows and also to minimize energy consumption of mobile devices a work is presented in [7]. The scheduling problem is formulated as an integer linear program to minimize the total number of active frames to save

J Sign Process Syst

energy consumption. The aftermentioned work adopts the scheduling process to contribute only to the mobile terminal’s sleep time thus ignoring the context information e.g. amount of battery power remaining for application in use etc. A detailed survey is presented on opportunistic scheduling in [8]. Opportunistic scheduling is considered to take advantage of the available information such as the channel quality and other QoS parameters (i.e., throughput, delay, and jitter) that consents the scheduler to the proper transmission resources for user [8]. An opportunistic scheduling mechanism for OFDM systems to minimize the overall transmission power is presented in [9]. Energy efficient and low complexity scheduling mechanisms for uplink cognitive cellular networks is presented in [10] along with a comparison of RR and opportunistic scheduling for EE. It is proven that RR is more energy efficient than opportunistic scheduling while providing the same QoS. Furthermore, existing schedulers mostly rely on either single parameter e.g. channel quality, throughput etc.[11–13], or a combination of more than one parameters, e.g. traffic type, channel quality or QoS metrics (jitter, delay) etc. But still these works are limited and do not go beyond the existing state-of-the-art work for considering context information n the scheduling process. On the other hand, some recent works consider the utilization of the context information in the scheduling process. To improve the QoS of, a context-aware resource allocation (CARA) scheduling scheme, for cellular wireless networks, is presented in [14]. This scheduling mechanism

is transaction-based and considers the running application’s foreground/background state as context information. Each transaction flow is provided a finish time, QoS requirements, and the context information attached. However, CARA considers the context information which is limited to the application in use, and contributes only the QoS improvement.

3 Scenario The proposed scenario for the context based scheduling is depicted in Fig. 1. The scenario shows an LTE-A cell having an eNB and several UEs randomly deployed inside the cell. Each UE gathers its required context information into a Context list (CX-list) and send to the eNB on the LTE-A uplink feedback channel. A context based database is maintained at each eNB which is updated each time when the UE triggers or when there is a change in the UE’s dynamic context information e.g. battery level, channel quality etc. Each UE once associated with a eNB, this particular eNB will create a temporary profile for this UE which contains UE’s temporary identity as well as key context information associated with the corresponding bearers including prioritized bit-rate and etc. The context information will be utilized for eNB downlink and uplink scheduling to guarantee QoS. The created profile for each UE will be removed once the UE moves into another cell, switches into idle mode, or switches off.

%

Context based UE profiles

Core Network MME

UE

Context Information

%

eNB

%

UE

CX-list

UE % % UE

%

UE

% % UE

UE %

Figure 1 Scenario for the investigated context based scheduling.

UE

UE

J Sign Process Syst

4 Problem Formulation

hkj

In this section, we formulate our problem. We define the following notations:

δjk

k K j J E Ptx N0 F

User index Total number of active users Resource block index Total number of resource blocks Energy (Joules) Transmit power per resource block at the eNB (Watts) AWGN noise variance File size (Bits)

Figure 2 Flow chart of the proposed algorithm.

PcRx

Instantaneous channel impulse response for the k-th user at j-th RB, including path-loss and shadowing (Complex value) If j -th RB assigned to k-th user, then δjk = 1; otherwise δjk = 0 (Binary value) Power consumption at the UE for receiving (Watts)

We target a scenario, where: 1) there are K active users in a single cell, all connected to one eNB; 2) file download application is considered for all users, the required file size is F k (bits) for the k-th user; 3) the remaining battery level of the k-th user is E k (Joule).

J Sign Process Syst

5 Proposed Algorithm

We made the following assumptions: 1) The up-to-date remaining battery level of each user E k (Joule) is known at the eNB through an error-free delayfree feedback channel. 2) The required download file size F k is known at eNB. 3) The active number of users K is smaller than the number of RBs in every transmission time interval (TTI). 4) Scheduling is performed every TTI (1 ms) in a RR fashion. 5) Equal power allocation over frequency/time is employed at the eNB. For every resource block, the transmit power is denoted as Ptx . The problem is formulated as follows ⎛

We investigate a context aware scheduling (CAS) algorithm which reduces the energy consumption by considering the context information based on the work presented in [15]. Most of the conventional schedulers, make decisions based on the throughput/QoS and instantaneous channel condition as part of a cross-layer scheduling approach. However, new factors that should be considered to enhance the system performance are the cost of energy per bit and the required energy (in battery level). In this context, the scheduling metric of a packet scheduler considers the ratio of the transmit energy to the number of transmitted bits [15] and multiply with the remaining battery level energy. For this reason, in a system with limited transmit energy, it is more efficient to allocate physical resource blocks (PRBs) to the users that require the least ratio of the transmit energy to the number of transmission bits and have low remaining energy. Thus, in the proposed packet scheduling scheme, the scheduling metric selects the UEs to be allocated in an order from lower to higher of the ratio of the transmit energy, Eum , to the number of transmission bits Bum , of the PRB m and BLu , battery level of the UE u as follows:

⎞

⎜ ⎟ K ⎜ ⎟  F k PcRx ⎜ k ⎟ max ⎜E − ⎟   2  k k ⎜ hj  δj Ptx ⎟ J k ⎝ ⎠ j log 2 1 + N0

(1)

Subject to the following constraints J 

δjk = LCM(J, K)/K

j

ϒ(u, m) = arg min u,m

Ek > 0

Eum Pum T = arg min ∗ BLu u,m Bum Bum

(2)

where ϒ(u, m) is the scheduling metric which denotes the index of selected UE u and PRB m respectively; energy is the multiple of power and time (Pum .T ).

Fk > 0

Decision And implementaon Engine

Core Network

UE Context Architecture

Schedulling Process

UE Context based scheduler

eNB

Context Reasoner

S-GW

Context Based Priority calculaon Context

MME

Internet

Context Filter Context Manager

BS Context P-GW

Context Provider

UEs context Context Provider

Context Informaon

Figure 3 Context architecture and framework for proposed scheduling algorithm.

Configuraon Profile

Policy Set Context Manager

Decision Engine Context Info Control Info

J Sign Process Syst Table 1 LTE QoS classes [17]. QCI Upper bound packet error rate Upper bound delay budget (ms) 10−2 10−3 10−3 10−6 10−6 10−6 10−3 10−6 10−6

1 2 3 4 5 6 7 8 9

100 150 50 300 100 300 100 300 300

As presented in [16], Pum = metric as follows: arg min u,m

6 Context Based Scheduling Framework and Architecture

(Bum ) , hm u

we can redefine the

Pum T (Bum )T ∗ BL = arg min ∗ BLu u m u,m hm Bum u · Bu

(3)

Let Pˆum denote the maximum transmit power at the transmitter that can be assigned for the UE u and the PRB m. Equation 2 can be rewritten as – ⎛ ⎜ ϒ(u, m) = arg min ⎜ ⎝ u,m

⎞ T m u (Bum )

+

battery level as in Eq. 5. The flowchart of the algorithm is shown in Fig. 2.

 1 Pˆum

Bum

⎟ ⎟ ∗ BLu ⎠

(4)

Because, arg min(x) = arg max( x1 ). Finally, the scheduling metric can be expressed as 

m 1 u ∗ ϒ(u, m) = arg max  m m ¯ u,m BLu E(Bu )/Bu

An information model of context awareness is presented in Fig. 3. This model basically illustrates how Context Information (Cx Info) is extracted and processed by various functional blocks in the context aware architecture. The outcome of this information model is the implementation of energy saving strategies based on the given context settings. Following is the brief description of different modules of our context architecture. Context Provider is the source of Cx Info. This Cx Info is obtained directly from the radio environment (e.g. from terminal measurement or network) without any processing. For instance, the information can be battery levels of MTs or signal strength to determine distance between UEs and eNB. If the mapping to the scheduling framework is considered, the context provider resides in both terminal and network side. The policy set is the set of strategies that can be used by radio, in other words it imposes constraints on the radio functionalities. The context manager is responsible for Cx Info processing to provide refined Cx Info for the decision engine. It consists of two blocks: Context Reasoner and Context Filter. The Context Reasoner collects raw Cx Info and generates rules for context filtering based on the constraints from the policy set. The reasoner may need to process the Cx Info to generate rules; however it will not alter the information content. The Context Filter filters the Cx Info based on the rules generated by



SPR

(5) PCRF EPS Bearer

The proposed energy efficient scheduler allocates the PRB m to the UE with larger excess channel gain which is distant to the required received energy per bit and lower

Radio bearer

UE

S1 Bearer

eNB

S5/S8 Bearer

S-GW

P-GW(PCEF)

PDN

Temp UE profile

Example bearer

Table 2 Representation of Battery level. Index 0 1

Battery level ≥ 30 % < 30 %

Index 0 1 2 3

Battery level [75 %–100 %] [50 %–75 %] [25 %–50 %] [0 %–25 %]

Dataflow 1(streaming ) Packet flow (video)

Packet flow (audio)

Data flow2 (video conference) Figure 4 QoS information flow.

J Sign Process Syst Figure 5 Mapping of physical channels and radio resources.

1) Uplink mapping Slot 1

2) Downlink mapping

Slot 2

Slot 1

... RB N

RB N

Slot 2

...

PDSCH

PUCCH format 2 (UE1/UE2/...) UE2 UE1 P D C C H

PRACH PUSCH

SIB2

PUCCH format 2 (UE1/UE2/...) RB1

RB1

3) MAC frame combined context informaon and service data MAC Header Sub-H for Sub-H for Sub-H for Control 1 Control 1 SDU 1

the context reasoner and output the high level operational Cx Info for the decision engine. The decision engine is the core of the context awareness framework. It makes decision based on operational Cx Info from the context manager and the constraints from the policy set. The decision will be implemented or used as a knowledge build-up. Configuration profiles represent a knowledge database built based on previous decisions. They can be seen as results of learning process. For example in a learning process, the implementation of a decision will be evaluated. A good (energy efficient) decision will be given a higher score. Good decisions with high scores are likely to be repeated in the future if the context setting permits. The framework is shown in the Fig. 3. The context related to scheduling is gathered at the UE and signaled to the eNB. The eNB gathers the information from all the connected UEs and its own information in UE context provider module. The collected information

...

MAC control 1

MAC control 2

MAC SDU 1

...

is filtered in the context manager and passed to the scheduling process where the context based priority calculation algorithm calculates the priority of each UE, based on the context parameters; which are battery level, channel quality and traffic type. The scheduling is performed in presence of network policies provided by policy set module. The scheduled decisions are passed to decision and implementation module.

7 The Representation of Context Information Context information, such as the received SINR, usually is a value/vector in continuous domain, which contains infinite entropy and cannot be processed by today’s digital systems. The common method is to predefine a table, which divides the original infinite-number of continuous values/vectors

Table 3 LTE-A defined list of control formats. PUCCH format

Release

Application

No. of UCI bits

No of PUCCH bits

1 1a 1b 2 2a 2b 3

R8 R8 R8 R8 R8 R8 R10

SR 1 bit HARQ-ACK and optional SR 2 bit HARQ-ACK and optional SR CQI, PMI, RI CQI, PMI, RI and 1 bit HARQ-ACK CQI, PMI, RI and 2 bit HARQ-ACK 20 bits HARQ-ACK and Optional SR

1 1 or 2 2 or 3 ≤ 11 ≤ 12 ≤ 13 ≤ 21

1 1 or 2 2 or 3 20 21 22 48

J Sign Process Syst

System Level Simulator

Simulator Mode Dynamic/Combined snapshot

Mobility models& User deployment

Cell Deployment

Context Aware Module

Look Up table from the PHY layer of the system

Context Information: BS context information, UE context information Battery level, Channel quality indicator, Traffic information, application in use etc.

RRM Admission Control, Link Adaptation, Channel resources Maganemnet, Scheduling, Power control and Interference calculations Context Aware schedulling (CAS) (RR, MCI, PF), HAndover, HARQ.

Computation of system level metrics (Energy Efficiency, spectrul efficiency etc.)

Figure 6 Components of system level simulator.

into finite-number of discrete regions. All the values within one region are represented by the index of this region. For example, LTE-A standardize a QoS table having 9 elements shown in Table 1. The index of each element in the table is called Quality Classification Indicator, which could be represented using 4 bits. Similarly, the SINR values are mapped into 16 combinations of modulation and coding schemes.

Following the same strategy, we could pre-define a table to represent the battery level. Two examples are given below and represented in Table 2. The first one set 30 % of remaining battery as the alarm threshold, which requires 1 bit to represent these two elements. While the other one provides have four elements, and requires at least 2 bits for representation. Having a longer table can represent more

Table 4 Simulations parameters. Parameters

Values

Carrier frequency fc Bandwidth Duplex mode Noise density Fast fading model Log-normal shadowing variance σ (dB) Number of cells Number of users BS transmit power Received SINR threshold Average snapshot Time transmission interval Number of resource block Link adaptation Traffic model Radio resource management Turbo decoder HARQ AMC P ERt arget CQI delay

2 GHz 10 MHz FDD −174 dBm/Hz Rayleigh fading using Pedestrian B model (6 taps, SISO) Urban LOS σ = 4 (dB), NLOS σ = 8 (dB) Multiple cell 50 43 dBm −3 dBm 100 1 ms (sub-frame) 50 RB in each slot, 7 symbol, number of subcarriers per RB=12,total subcarrier=600 EESM( Exp Effective SINR Mapping) Data (File) CAS, RR Max Log Map (8 iterations) Chase combining, Number of process=6,Retransmission interval=6 ms,Max Nb of retransmission=3 10 % Each TTI, with 2 ms delay

J Sign Process Syst

information, while it also increases the complexity and signaling overhead. Hence, there is a trade-off between the information accuracy and complexity plus overhead.

The quasi-/stable context information is stored in the core network, and transmitted to eNB via backhaul links. Taking the QoS information as an example, the process is illustrated in Fig. 4. As an IP-connected network, all services (voice, data, etc.) in LTE-A are connected to the external Packet Data Network (PDN) through the Packet Data Network Gateway (P-GW). It is natural to configure the quality of service at the P-GW. Since the expected service quality is related to charging, an element called Policy and Charging Enforcement Function (PCEF) is embedded in P-GW. This PCEF is responsible to communicate with other two network elements in the core network. One is called Policy and Charging Function (PCRF), which provides policy and charging control rules. The other one is called Subscription Profile Repository (SPR), which contains users’

subscription information such as his/her subscribed price plan. Having the information of how much a UE is willing to pay, the P-GW configures QoS for difference services through a mechanism named Evolved Packet Service (EPS) bearer. A EPS bearer could be considered as a bi-directional data pipe as a logical connection between the UE and the P-GW. It consisted by three other logical bearers, S5/S8 bearer, S1 bearer and radio bearer as shown in Fig. 4. Once a UE is switched on and connected to the network, a default EPS bearer will be set-up, and be remained until this UE switched off. More than one EPS bearer can be set up for difference services. Each EPS bearer is associated with a specific QoS, which defines how the data will be transferred using parameters such as error-rate, delay as shown in Table 1 (but not limited to these). One example of ESP bearer is also demonstrated in Fig. 4. This bearer defined a QoS suitable for real-time video transmission, for example QoS having low delay. In this bearer, two service data flows—one for video streaming from the UE to the network, and one for video conference from the network to the UE are supported. Within each data flow, one or more packet flows (e.g. audio and video) are contained for supporting this service. LTE gives the same QoS to all the packet flows within a particular EPS bearer. The eNBs are not suitable candidate for storage these quasi-/static pre-defined UE context information because: 1) The operators want to reduce the cost of eNBs; 2) eNB completely lose connection with certain mobile nodes once it moves out from this eNB’s coverage area. However, once a UE is associated with an eNB, this eNB will create a temporary profile for this UE, which contains this UE’s temporary identity as well as key QoS parameters associated with the corresponding bearers including priority prioritized bit-rate and etc. These QoS information will be utilized for eNB downlink and uplink scheduling. This profile will be deleted when this UE moves into another cell, switches into idle mode, or switches off.

Figure 7 Total remaining battery (%) vs simulation time (TTIs).

Figure 8 Average battery consumption vs number of users.

8 The Acquisition of Context Information Two types of context information, eNB context and UEs’ context, are shown in Fig. 4. The eNB’s context is measured, update, and stored at each eNB, which is easy to obtain. UEs’ context information can be categorized into two types according to its lifetime: quasi-/stable pre-defined context information with less or no changes; and unstable measured context information with frequently changes. For example, UE’s price plan and its expected QoS for different applications are pre-defined with low change rate; while the channel quality and battery level need to be measured with high variation. In this section, we will discuss the acquisition of these two types of UE context information. 8.1 Quasi-/stable Pre-defined UEs’ Context Information Obtained from Core Networks

J Sign Process Syst

8.2 Unstable Measured UEs’ Context Information Obtained from Uplink Feedback Channel The unstable measured UEs’ context information, such as the channel quality and battery level, is obtained from the UE and signaled via uplink feedback channel. There are two feedback modes: periodic and aperiodic. Periodic mode is carried on a regular interval according to the average variable rate of the context information. For example, the channel quality indicator is influence by the multi-path fading, and has a faster variable rate than the Rank Indicator, which is more influenced by shadowing. Hence CQI feedback is carried out 32 times more frequent than the RI feedback. By contrast, aperiodic node usually carried out on-demand or for abnormal situation. Aperiodic mode usually has a higher priority than periodic mode. The uplink feedback channel could be the Physical Uplink Share Channel (PUSCH) or the Physical Uplink Control Channel (PUCCH) depends on the states of UE. More explicitly 1) If the mobile is in connection mode and has data waiting for transmission via the PUSCH, the context information will be multiplex with the data at MAC layer, and transmitted back to the eNB. 2) If the mobile is in connection mode but has not data for transmission, context information will be transmitted via the PUCCH. 3) If the mobile is in idle mode, this mobile needs to invoke a random access request via Physical Random Access Channel (PRACH), re-establish the connection with the eNodeB, and then feedback its context information via PUSCH or PUCCH as described previously. In addition, for energy saving purpose, it is more appropriate for an idle-mode mobile to report in aperiodic mode trigger by the change of context information status, for example, the battery level dropped below a certain threshold. A simplified mapping of several physical channels (related to this paper) and radio resources are shown in Fig. 5. The outermost parts of the uplink band are reserved for PUCCH. The rest of the uplink bandwidth is mainly used by the PUSCH. Some resource blocks are reserved for PRACH. For downlink radio resources, a few symbols (varies from frame to frame) at the beginning of each subframe are reserved for control information such as Physical Downlink Control Channel (PDCCH). The rest of the subframe is reserved to downlink data transmission as the Physical Downlink Shared Channel (PDSCH).

The PUSCH is allocated to individual mobile in units of resource blocks within each sub-frame. An uplink scheduler at the eNB will decide allocate which resource blocks to which UE, and sending the UE a scheduling grant on the PDCCH. This grants permission for the mobile to transmit and states all the transmission parameters it should follow, such as transport block size, the resource block allocation and the modulation scheme. If one UE has data to transmit, it will initial a scheduling request through PUCCH, and receive such a scheduling grant. If this UE also has context information to feedback, it will concatenate its Service Data Unit (SDU) with context information (denoted as control) as shown in Fig. 5. The type of the context information, their length, and their place in the combined packet are included in the MAC header. The PUCCH is also shared by all UEs. An individual mobile transmits the PUCCH using two resource blocks, which occupies 1 ms and at the opposite sides of the frequency band. To efficiently utilize the limited PUCCH bandwidth, these two resource blocks are further shared by several UEs by using different cyclic shift or orthogonal sequence index, which are assigned to this mobile by eNB. Moreover, LTE-A standard has pre-defined a list of control formats, which are shown in Table 3. The resource blocks in the PUCCH are reserved for different control format. The way of reserving which resource blocks for what type of control format is again decided by the eNB, and advertised in the System Information Block No.2 (SIB 2) via PUSCH. As illustrated in the Fig. 5, the two highlighted resource blocks are reserved for transmitting a combination of CQI/PMI/RI. UE1 and UE2 both want to transmit these three types of information. As a result, they will spread these control information with their own orthogonal sequence, and occupy these two resource blocks.

9 Simulation Setup and Results This section presents the simulation results and analyses of our devised algorithm. Figure 6 demonstrates the component of the simulator we use for our simulation purpose. The simulation parameters set for proposed scenario to simulate is shown in Table 4. From Fig. 7, it can be observed that the battery consumption of UEs is reduced by around 10 % compared to the benchmark algorithm RR. It is also observed that our algorithm effectively saves more energy as simulation time increases, and reaches optimal results within 100 TTIs due to the context aware information in the scheduling. For instance, if we consider battery level as a context entity inside the context aware module, it demon-

J Sign Process Syst Figure 9 CDF vs energy consumption.

strates that the lower the battery level the higher the priority of our scheduling, while also considering other parameters. Therefore, in our investigated algorithm, low battery level and minimum energy per bit is assigned higher scheduling priority which eventually leads to reduced battery consumption. For the scheduling process, CAS considers the remaining battery of each UE along with channel quality, traffic demand and adaptive coding which are mostly ignored in the previous algorithms. As result, CAS saves energy and the average remaining battery level of the active users are much higher compared to conventional RR. Figure 8 represents the average remaining battery level compared with number of users. This figure depicts that CAS, on average, saves more energy compared to conventional scheduling algorithms. Figure 9 shows the Cumulative Density Function (CDF) of the UEs’ energy consumption. With the proposed method, almost 50 % of the users consume energy which is a value 0.5 mJ. Using the conventional RR, only 20 % of UEs’ consume the same amount of energy; the gain

Figure 10 Overall comparison of CAS with conventional RR.

is 25–30 %. This gain is achieved due to the context aware information available at context module for each mobile user. The context aware module provides the context to the RRM module to consider the battery level of each user in the scheduling process and to adapted its power according to the traffic load in each cell. Thus, the proposed algorithm saves energy and increase the number of UEs to be scheduled. A 3-D plot is demonstrated in Fig. 10. In this figure, we show an overall comparison between the proposed algorithm and the conventional RR in terms of energy consumption, number of users and simulation frames. Here, we assume all the context entities (battery level, CQI and traffic) that are defined in the proposed algorithm section. It can be observed that the proposed approaches saves almost 0.2 mJ of energy in contrast to the benchmark RR algorithm.

10 Conclusion In this paper, we present a Context Aware Scheduling (CAS) algorithm for 5G based on LTE-A exploiting the context information. CAS goes beyond the state-of-theart and exploits the context information of UE for energy saving and guarantee the requested QoS. Furthermore, we present an information model for context awareness which illustrates how context information is extracted and processed by various functional blocks in the context aware architecture of UE. The presented architecture is not only used for radio resource management, but can further be utilized in various context based mechanisms. The paper also discuss the design of context information based signaling in LTE-A that can be used in the future technologies. A context aware module is implemented in a system

J Sign Process Syst

level simulator to test the efficiency of the proposed CAS and to provide a comparison with conventional scheduling. The simulation results show that CAS has the potential to save energy compared to conventional RR scheduling. In fact, the battery consumption of the UEs are reduced by 10–15 % by using CAS in contrast to conventional RR scheduling.

References 1. Alam, M., Albano, M., Radwan, A.R., & Rodriguez, J. (2010). Context parameter prediction to prolong mobile terminal battery life. In Proceedings of the international mobile multimedia communications conf.—mobimedia, Lisboa, Portugal (pp. 1–10). 2. Jalali, A., Padovani, R., & Pankaj, R. (2000). Data throughput of cdma-hdr a high efficiency-high data rate personal communication wireless system. In Vehicular technology conference proceedings, 2000. VTC 2000-Spring Tokyo. 2000 IEEE 51st (Vol. 3, pp. 1854– 1858). 3. Marsan, M., Chiaraviglio, L., Ciullo, D., & Meo, M. (2011). Switch-off transients in cellular access networks with sleep modes. In 2011 IEEE international conference on communications workshops (ICC) (pp. 1–6). 4. Marsan, M.A., Chiaraviglio, L., Ciullo, D., & Meo, M. (2010). A simple analytical model for the energy-efficient activation of access points in dense wlans. In Proceedings of the 1st international conference on energy-efficient computing and networking, ser. e-Energy ’10 (pp. 159–168). New York: ACM. [Online]. Available: doi:10.1145/1791314.1791340. 5. (2010). 3GPP TS 25.214, Physical Layer Procedures (FDD), release 8, v8.9.0. 6. (2009). Waqas ahmad khan. Impact of DRX on VoIP Performance and Battery Life in LTE, master thesis. 7. Wu, C.Y., Ho, H.J., & Lee, S.L. (2012). Minimizing energy consumption with qos constraints over ieee 802.16e networks. Computer Communications, 35(14), 1672–1683. [Online]. Available: doi:10.1016/j.comcom.2012.06.012. 8. Asadi, A., & Mancuso, V. (2013). A survey on opportunistic scheduling in wireless communications (pp. 1–18). 9. Wong, C.Y., Cheng, R., Lataief, K., & Murch, R. (1999). Multiuser ofdm with adaptive subcarrier, bit, and power allocation. IEEE Journal on Selected Areas in Communications, 17(10), 1747– 1758. 10. Ding, H., Ge, J., Benevides da Costa, D., & Jiang, Z. (2010). Energy-efficient and low-complexity schemes for uplink cognitive cellular networks. IEEE Communications Letters, 14(12), 1101– 1103. 11. Kelly, F. (1997). Charging and rate control for elastic traffic. European Transactions on Telecommunications, 8, 1. 12. Viswanath, P., Tse, D., & Laroia, R. (2002). Opportunistic beamforming using dumb antennas. IEEE Transactions on Information Theory, 48(6), 1277–1294. 13. Kushner, H., & Whiting, P. (2004). Convergence of proportionalfair sharing algorithms under general conditions. IEEE Transactions on Wireless Communications, 3(4), 1250–1259.

14. Proebster, M., Kaschub, M., & Valentin, S. (2011). Contextaware resource allocation to improve the quality of service of heterogeneous traffic. In 2011 IEEE International Conference on Communications (ICC) (pp. 1–6). 15. Song, J., Gil, G.-T., & Kim, D.-H. (2010). Packet-scheduling algorithm by the ratio of transmit power to the transmission bits in 3gpp lte downlink. EURASIP Journal on Wireless Communications and Networking, 2010, pp. 6:1–6:13. [Online]. Available: doi:10.1155/2010/251281. 16. Wong, C.Y., Cheng, R., Lataief, K., & Murch, R. (1999). Multiuser ofdm with adaptive subcarrier, bit, and power allocation. IEEE Journal on Selected Areas in Communications, 17(10), 1747– 1758. 17. 3GPP TS 23.203 Policy and charging control architecture. Retrieved 13 Feb 2015.

Muhammad Alam holds a PhD. degree in Computer science from university of Aveiro. In 2009, he became a Researcher at Instituto de Telecomunicac¸o˜ es – Aveiro (Portugal) and concluded his Ph.D. in MAP-i Doctoral program Portugal in 2014. He has been involved in several research projects such as C2POWER, PEACE, SmartVision and ICSI. Currently, he is a Post-doctoral researcher at the Instituto de Telecomunicac¸o˜ es—P´olo de Aveiro, Portugal, working in the EU funded ICSI project. He is the author of several Journal and conference research papers. His research interests include Wireless Communication, Vehicular Communication, ITS and Context aware systems.

Du Yang received her BEng. degree from the Beijing University of Posts and Telecommunications (China), in 2005; and her MSc. and Ph.D. degrees from University of Southampton (UK), in 2006 and 2010 respectively. She was a recipient of the Mobile VCE Scholarship. She worked as a Post-doctoral researcher at the Instituto de Telecomunicac¸o˜ es—P´olo de Aveiro, Portugal, working in the EU funded WHERE2 project. Currently, she is working as Core Network Engineer at Huawei Technologies, U.K. Her research interests include MIMO techniques, multi-hop relaying communication, position information assisted communication, joint PHY and MAC layer optimization in LTE standard.

J Sign Process Syst Kazi Huq received the B.Sc. degree in computer science and engineering from Ahsanullah University of Science and Technology, Dhaka, Bangladesh, in 2003; the M.Sc. degree in electrical engineering from Blekinge Institute of Technology, Blekinge, Sweden, in 2006; and the Ph.D. degree in electrical engineering from the University of Aveiro, Aveiro, Portugal, in 2014. Since April 2014, he has been a Senior Research Engineer with the Instituto de Telecomunicac¸o˜ es, P´olo de Aveiro, Portugal. He is the author of several publications, including conferences, journals, and a book chapter. His research activities include fifth-generation (5G), energy-efficient wireless communication, radio resource management for green cellular networks, and coordinated scheduling.

Firooz Saghezchi received the MSc degree in Electrical Engineering-Communication Systems from Shiraz University, Shiraz, Iran in 2003 and the BSc degree in Electrical EngineeringTelecommunications from University of Tabriz, Tabriz, Iran in 2000. He secured a lecturer position at Electrical Engineering Department of Islamic Azad University of Garmsar, Garmsar, Iran for six years. Then, he joined 4TELL Wireless Communication Research Group at Instituto de Telecomunicac¸o˜ es, Aveiro, Portugal in 2010, where he has been involved in several European research projects such as HURRICANE, C2POWER and E2SG. He is currently pursuing his PhD under the umbrella of MAP-tele Doctoral Programme in Telecommunications, a joint degree offered by University of Minho, University of Aveiro and University of Porto in Portugal. He has authored several scientific works including book chapters, journal and conference publications and served as an active reviewer and TPC member for several high-profile journals and conferences. His research interests include 5G, energy efficiency, cooperative communications, game theory, demand response and smart grid.

Shahid Mumtaz received his MSc. degree from the Blekinge Institute of Technology, Sweden and his Ph.D. degree from University of Aveiro, Portugal. He is now a senior research engineer at the Instituto de Telecomunicac¸o˜ es—P´olo de Aveiro, Portugal, working in EU funded projects. He has been involved in several EC R&D Projects (5GPPSpeed-5G, CoDIV, FUTON, C2POWER, GREENET, GREEN-T, ORCALE, ROMEO, FP6, and FP7) in the field of green communication and next generation wireless systems. In EC projects, he holds the position of technical manager, where he oversees the project from a scientific and technical side, managing all details of each work packages which gives the maximum impact of the project’s results for further development of commercial solutions. He has been also involved in two Portuguese funded projects (SmartVision & Mobilia) in the area of networking coding and development of system level simulator for 5G wireless system. His research interests include MIMO techniques, multi-hop relaying communication, cooperative techniques, cognitive radios, game theory, energy efficient framework for 4G, position information assisted communication, joint PHY and MAC layer optimization in LTE standard. He is author of several books, conference, journals and book chapter publications.

Jonathan Rodriguez received his Masters degree in Electronic and Electrical Engineering and Ph.D from the University of Surrey (UK), in 1998 and 2004 respectively. In 2005, he became a researcher at the Instituto de Telecomunicacoes (IT)-Portugal where he was a member of the Wireless Communications Scientific Area. In 2008, he became a Senior Researcher where he established the 4TELL Research Group (http://www. av.it.pt/4TELL/) targeting next generation mobile networks with key interests on green communications, cooperation, security, and electronic circuit design. Since its inception, the group has steadily grown and now Dr. Rodriguez is responsible for supervising 36 research staff, including a project portfolio of over 25 research grants. He has served as project coordinator for major international research projects, that includes Eureka LOOP and FP7 C2POWER, whilst serving as technical manager for FP7 COGEU and FP7 SALUS. Since 2009, he became an Invited Professor at the University of Aveiro (PT) and Honorary Visiting Researcher at the University of Bradford (UK). He is author of more than 300 scientific works, that includes 6 books. His professional affiliations include: Senior Member of the IEEE and Chartered Engineer (CEng) since 2013, and Fellow of the IET (2015).