You are on page 1of 6



   



 
 


        
   ! 
Abdulraqeb Alhammadi1, Mardeni Roslee1*, Mohamad Yusoff Alias1, Ibraheem Shayea2, Saddam Alraih3
1
Center for Wireless and Technology, Multimedia University, Cyberjaya, Selangor, Malaysia
2
Istanbul Technical University, Electronics and Communication Engineering Department, Istanbul, Turkey
3
National University of Malaysia, Bangi, Selangor, Malaysia
mardeni.roslee@mmu.edu.my

Abstract— Deploying ultra-dense small-cell base stations in the management is a major function in mobile wireless communi-
next-generation mobile networks is one of the most expected ap- cation; it allows mobile devices to work in global systems for
proaches to overcome the uncertainty of millimeter wave (mm- mobile communications (GSM) and universal mobile telecom-
wave) communications. Mobility management is a critical issue munications system (UMTS). Mobility management primarily
that requires much concern to achieve seamless and highly reliable
traces and tracks mobile location and delivers services, such as
connection through user mobility. Handover in heterogeneous net-
calls and SMS. Given the movements and changes in the loca-
works (HetNets) has gained attention, especially when 5G ultra-
dense small cells coexist with the current 4G networks. Handover tion of a mobile device while a call or connection is in progress,
failure is one of the main issues in mobility, and it can be avoided this action must be performed smoothly without disconnecting
by adjusting handover control parameters (HCPs): time-to-trig- the connection or dropping the call. A mechanism called the
ger (TTT) and handover margin (HOM). In this paper, we pro- handover enables mobile devices to stay connected to the core
posed dynamic HCPs in HetNets (LTE-A and mm-wave networks) network through transferring from one channel to another.
with dense small cells. The proposed algorithm was compared with Although remarkable improvements have been provided by
different settings of HCPs for different user mobile speed scenar- the 4G system, several issues related to mobility management
ios. The simulation results show that the proposed algorithm sig- became more complicated in heterogeneous networks (Het-
nificantly reduces the probability of ping pong handovers and ra- Nets). These issues must be solved efficiently to provide seam-
dio link failure, thus improving network performance.
less and high reliable connection through user mobility. Hand-
over control parameters (HCPs) are one of the major challenges
Index Terms—Mobility management, small cells, HetNets. facing the implementation of 4G and future systems; they must
be adjusted according to dynamic events.
In this paper, a dynamic-HCP (D-HCP) algorithm is pro-
I. INTRODUCTION
posed to address handover parameter optimization (HPO)
Several decades have passed since the first generation of based on handover types (too early handover, too late handover
mobile wireless communication (1G) was established. Nowa- and wrong cell handover). The D-HCP aims to reduce the radio
days, mobile wireless communication is one of the most widely link failure (RLF) and handover ping-pong probability (HPPP).
used technology. The evolution of mobile wireless communi- The rest of the paper starts by giving background knowledge
cation generation series gradually changes. Only voice and text and related works in Section 2. Section 3 provides details on
data were supported in 1G, but video and internet streaming are the HPO. Section 4 presents the proposed algorithm for Het-
supported in the current generation of mobile networks (4G). Nets. Section 5 presents the evaluation of handover perfor-
Despite the high speed and numerous features provided by 4G, mance metrics that must be used for system evaluation. Section
the next network generation (5G) will signify an important rev- 6 evaluates and discusses the simulation results in different sce-
olution in the technology field, especially in mobile wireless narios. Finally, the last section presents the conclusions and
communication. suggests several points for potential future works.
The 5G mobile wireless communication will provide a num-
ber of significant features, such as a high data rate (up to 10 II. BACKGROUND AND RELATED WORKS
Gbps), a remarkably low latency (less than 1 ms round-trip
Mobility management is one of the most important areas in
time) and a high bandwidth in the unit area [1]. Researchers and
HetNets, which are still under development and enhancement
developers are working on developing 5G in all its areas to
to provide a seamless communication. Two different mobility
achieve and meet the 5G requirements and prepared for its re-
managements exist in cellular networks: user equipment (UE)
lease in the prospected year, which is 2020. Mobility manage-
mobility with respect to spectrum and UE mobility with respect
ment is an important area in the 5G research field [2]. Mobility
to base station (BS). Spectrum mobility is more related to spec-
trum utilization in cognitive radio technology [3, 4], whereas

  !"#$%%% 39
mobility with respect to BS maintains the connection between in non-heterogeneous wireless networks, the mobile UE main-
UE and BS, and the UE moves from one BS to another. tains its radio connection when it moves within the cells by per-
Different techniques and algorithms have been applied to forming a handover process from the serving eNB to another
handover, and each one exhibits a different performance and eNB that provides better signal quality. However, in HetNets
accuracy. Researchers proposed a highly efficient approach for with different cell types, handover maintains the UE connection
automatic optimization handover; it provides self-organizing with eNB while moving through different cells. Thus, the hand-
handover for LTE networks [5]. This technique aims to reduce over will become more complex and require efficient algorithm
unnecessary amounts of ping-pong handovers. Although this as each cell features different specifications, such as cell radius,
approach achieves almost zero drop call ratio for the complete transmission power and path loss model. Fig 1 demonstrates an
simulation time, it features a slow response time (after 60 s). example of a UE trajectory between macro and small cells. Ef-
Other studies have proposed a methodology for certain radio ficient handover can support service continuity and enhance
frequency conditions and different scenarios in which handover UE throughput, ideally, without any service interruption. LTE-
parameters are flexibly customized depending on radio fre- A system introduces features to enhance system performance,
quency conditions [6]. The BS in this technique is built-in with such as improved handover mechanisms that provide short
a learning machine and data mining techniques to determine handover interruption time. In addition, the LTE-A system sup-
and identify the received signal strength and characteristic pat- ports mobility speeds of up to 500 km/h [10].
terns. The spectral efficiency at the cell-edge is improved, and
the downlink data rate in a certain loading scenario can reach
up to 150% better than the traditional approach of selecting a
set of handover parameters.
Other scholars proposed a new green procedure using self-
optimization self-organizing network (SON) function which
minimizes energy consumption in LTE-Advanced/5G [7]. In
general, this technique reduces energy consumption in all sce-
narios, but shows better results when the mobile device moves
at low speed. A new handover detection self-organizing-hand-
over parameter (HD-SOHP) based on reinforcement learning
(RL) concept was introduced by [8]. Call drop decreases due to
effective session handovers. The performance of using HD-
SOHP improved up to twice as that when using SON-based
handover techniques in terms of call drop handover failure and Fig. 1. Handover concept in HetNets
ping-pong and support mobility speed of up to 120 km/h.
Wei Zheng et al. [9] developed a handover parameter adjust- The handover algorithm could be divided into three catego-
ing-based mobility robustness optimization scheme based on ries, namely, reference signal-received power (RSRP), RSRP
gradient algorithm; this scheme aims to reduce unnecessary with threshold and handover margin (HOM) and time-to-trig-
handovers (UHOs) and RLFs caused by too late handover, too ger (TTT) interval. In the first category, handover decision al-
early handover or wrong handover in open access femtocells. gorithm is used based on RSRP only, which decides to initiate
They also introduced a cost function incorporating UHOs and the handover procedure once the target RSRP becomes greater
wrong handovers for femtocell networks. The simulation re- than that of the serving RSRP. In the second category, the hand-
sults show that the proposed scheme can achieve reduction of over decision algorithm is constructed based on the RSRP with
RLF ratio and UHO ratio with minimal handover signaling a threshold level. That is, the handover decision procedure is
modifications. initiated upon fulfilling two conditions: the serving RSRP is
less than a predefined threshold, and a stronger target RSRP is
III. HANDOVER PARAMETERS OPTIMIZATION compared with the serving RSRP. The third category, on the
other hand, adds the HOM to the RSRP level to initiate the
Mobility management becomes essential in wireless mobile handover decision algorithm. Thus, the handover decision is
communications, requiring seamless communication at differ- considered when the target RSRP is greater than the serving
ent mobility levels. Radio resource management maintains the RSRP plus the HOM. The third category is the most practical
radio link connection between UEs and BSs within the cover- algorithm used for the handover decision algorithm [11, 12].
age area by handing over the UE from one cell to another. The However, the HOM level between the serving and target RSRP
handover refers the process of establishing a new radio link and the TTT interval contribute in reducing the UHO procedure
connection from the source to the target BSs [7, 8]. Therefore, (preventing the ping-pong effect). The ping-pong effect is a fre-
quent UHO scenario between two neighbor stations and results

40
from rapid fluctuations in the signal strength from both stations. values: TTT = [40; 80; 160; 480] (ms); HOM = [-1; 0; 1; 2; 3]
Thus, HCP values are considered highly sensitive for consider- (dB). We considered three types of RLF that disrupt the link
ing a robust handover decision, which in turn contributes to the between the UE and eNB, namely, too early handover, too late
enhancement of throughput and outage probability. Thus, the handover and wrong cell handover. For example, UEs in high
handover decision algorithm based on RSRP with a marginal speed state will pass through many eNBs which require a low
level is the adopted, and conventional handover decision algo- TTT value to avoid too late handover. On the contrary, when
rithm used in in LTE and LTE-A systems [13, 14]. This hand- UEs move with low speed, they will have experienced short
over decision is given mathematically as follows: distance and good quality signal, requiring high TTT value to
avoid too early handover. Algorithm 1 shows the proposed D-
்ܴܴܵܲ௔௥௚௘௧ ൐ ܴܴܵܲௌ௘௥௩௜௡௚ ൅ ‫ܯܱܪ‬ሺͳሻ
HCPs algorithm based on the type of handover detected.
where ்ܴܴܵܲ௔௥௚௘௧ and ܴܴܵܲௌ௘௥௩௜௡௚ represent the target and
serving RSRP, respectively. Algorithm 1: D-HPO
1. Initialize system’s parameters (LTE-A and mmwave)
2. If t=1
Handover Decision false
3. else
Target eNB
4. Update HPPP and HOF
Serving eNB
5. If the early Too handover detected
Adaptive Margin Value 6. Updating TTT and HOM
Max Handover margin value 7. ࡹሶ ࢚ ൌ ࡹ࢚ି૚ ൅ ࢻǡ ࢀሶ ࢚ ൌ  ࢀ࢚ି૚ ൅ ࢼ
8. Handover Decision  True
RSRP

Update ࡹሶ ࢚ and ࢀሶ࢚


Average Handover margin value
9.
Min Handover margin Value
10. elseif Too late handover detected
Qrslevmin 11. Updating TTT and HOM
12. ࡹሶ ࢚ ൌ ࡹ࢚ି૚ െ ࢻǡ ࢀሶ ࢚ ൌ  ࢀ࢚ି૚ െ ࢼ
T1, represent the beginning of TTT 13. Handover Decision  True
Tn, represent the end of TTT 14. Update ࡹሶ ࢚ and ࢀሶ࢚
Tn
15. elseif Wrong cell handover detected
ࡹሶ ࢚ ൌ ࡹ࢚ି૚ െ ࢻǡ ࢀሶ ࢚ ൌ  ࢀ࢚ି૚ ൅ ࢼ
T1
Time [millisecond]
16.
Time-To-Trigger (TTT)
17. Handover Decision  Ture
Fig. 2. Handover decision algorithm in LTE-A system 18. ࡹሶ ࢚ and ࢀሶ࢚
19. else
Thus, when the target RSRP becomes greater than the serv- 20. Handover Decision  false
ing RSRP by HOM level continuously during a TTT interval, a 21. end if
true handover decision is considered. Fig 2 illustrates the con- 22. end if
cept of using handover decision based on the RSRP and HOM
levels with TTT interval. HCP values can be fixed values or where‫ܯ‬ሶ ௧ and ܶ௧ሶ are the adaptive HOM and TTT, respec-
adaptively adjusted. Fixed HCP values mean that the HCPs are tively.ߙ represent the step levels to adjust HOM which carry a
set at constant values through all transmission time intervals value of 0.5, and ߚ represents the transition level in given TTT
(TTIs). Conversely, adaptive adjustment refers to automatic ad- values. These steps are implemented when a handover condi-
justment of HCPs based on different network factors. These ad- tion is encountered. The TTT and HOM are updated for each
justments could be optimized by identifying the types of hand- UE in cell to avoid RLF in each simulation time.
over that occur during UE mobility in HetNet.
V. EVALUATION METRICS
IV. PROPOSED ALGORITHM In wireless networks, various performance metrics are
In this section, we describe the proposed novel algorithm usually defined to characterize network service quality. Thus,
called D-HCP to address HPO based on mobility scenarios dis- such metrics are essential in this work to investigate the
cussed in the previous section. This novel algorithm reduces efficiency of the proposed algorithms compared with the
RLF and HPPP and maximizes the spectral efficiency of the previous algorithms. These performance metrics include the
system. The framework of HetNets (LTE-Advanced and 5G) following:
are used for evaluating the proposed algorithm. The HCP val- HOP is the probability of UE handover from the serving
ues (HOM and TTT) are selected according to the 3GPP [15] eNB to target eNB and also determines the quality of service.
for HetNets, in which HOM and TTT possess the following In other words, HOP is the probability of switching the radio

41
link connection when the signal quality of the serving eNB VI. SIMULATION AND PERFORMANCE ANALYSIS
worsens. HOP can be calculated by the following expression:
A. Simulation Environment
‫ܱܲܪ‬ǡ௎ா ሺ்߲ ǡ ߲௦ ሻ ൌ ܲ௥ ሾ்߲ െ  ߲௦ ൒ ‫ܯ‬ሿሺʹሻ
The system model and simulation parameters of our work is
where ்߲ and ߲௦ denote the signal strength of target and serving based on 3GPP-specified evaluation methodology using
cells at time ‫ݐ‬, respectively, and ‫ ܯ‬denotes the HOM level. The MATLAB. Two-tier model comprises seven macro tri-sector
average of HOP in each simulation time over all UEs in the hexagonals (MeNB) where 5G SeNBs are placed at the middle
network can be given by the following equation: of each sector of a MeNB cell. The inter-site distances for
σே ೆಶ
௝ୀଵ ܲ௝ ሺ‫ܱܪ‬ሻ
MeNB and SeNB are 500 and 100 m, respectively. Table 1
തതതതതത ൌ
‫ܱܲܪ‬ ǡሺ͵ሻ provides the main simulation parameters. A random mobility
ܰ௎ா model is considered for all UEs throughout the network. That
HPPP is an important metric that calculates the frequency is, in every simulation cycle ‫ݐ‬, the UEs can move in any of the
of UHOs between two adjacent cells. The HPPP counts when four directions in the range of [0", 360"] selected randomly.
the UE disconnects its communication link from the serving Four different UE level speeds are considered: slow, medium
eNB and establishes a new connection with the target eNB, and and fast, ranging from 40 km/h to 160 km/h in steps of 40 km/h
then bounces back to the serving eNB under a time period that to investigate the effect of UE speed on the system perfor-
is less than the critical ping-pong interval ܶ௖ . ܶ௖ is defined as a mance. These considered speeds, which were assumed for the-
short time period (which is assumed to be 2 s) required to cal- oretical investigation, represent typical vehicle speeds in urban
culate UHO between neighboring cells. The HPPP should be and suburban areas.
considered if the condition of the following expression is satis-
fied: Table 1. Simulation parameters
Parameter Value
‫ ܲܲܲܪ‬ൌ  ܲ௥ ሾܶ௜ ൑  ܶ௖ ሿሺͶሻ Carrier frequency (GHz) 2.1 (MeNB), 28 (SeNB)
System bandwidth (MHz) 10 (MeNB), 500 (SeNB)
Number of eNB 7 (MeNB), 21 (SeNB)
ܶ௜ ൌ  ܶ௅ െ  ܶ௛௕ ሺͷሻ Inter-site distance (m) 500 (MeNB), 100 (SeNB)
Macrocell path loss model (dB) 128.1 + 37.6 log10 (R)
where ܶ௜ǡ denotes the time passed by the UEs starting from ini-
Small cell path loss model (dB) 79.2 + 26 log10 (R)
tiating handover in eNB and its return to the same eNB. ܶ௅ de- Shadowing standard deviation (dB) 8 (MeNB), 10 (SeNB)
notes the time to initiate handover from the serving eNB, Shadowing correlation 0.5 (1) between cells
whereas ܶ௛௕ denotes the time required by UE to return to the Transmit power (dBm) 46 (MeNB), 30 (SeNB)
same eNB. When the UE is handed back to the previous serving UE speeds (km/h) 40, 80, 120, 160
UE noise figure (dB) 9
eNB, then the ܶ୧ is less thanܶ௖ ሺܶ୧ ൏ ܶ௖ ሻ, and the handover is
Thermal noise density (dBm/Hz) –174
recorded as ping-pong handover for each UE in the network.
The average HPPP in each simulation time over all UEs in the
network can be given by the following equation: B. Performance analysis
The proposed algorithm was compared with three differen-
ܰு௉௉
തതതതതതതത
‫ ܲܲܲܪ‬ൌ  ሺ͸ሻ tiated static HOM and TTT settings for all UE mobile scenarios
ܰோுை shown in Table 2. The performances of these different settings
of HCPs were analyzed using three key performance indicators,
ܰோுை ൌ  ܰௌ ൅  ܰி ሺ͹ሻ namely, HOP, HPPP and average handover link failure proba-
where ܰு௉௉ denotes the number of ping pong handovers over bility.
the entire simulation. ܰௌ ƒ†ܰி  are the successful handover
and handover failure, respectively. Table 2. HCPs settings
Set HCPs
RLF is recorded when the UE loses connectivity over eNBs HOM (dB) TTT (ms)
in the network. Intuitively, the main sources of RLF include S1-HPO 5 480
S2-HPO 2 512
cases of handover procedure failure or the disruption experi- S3-HPO 0 40
enced by the radio link. To present RLF quantitatively, the av-
erage RLF probability (RLFP), that is, the average RLF rec-
We specifically studied the effect of user mobility on the
orded over all UEs ܰ௎ா , is calculated as follows:
system performance. Four different mobile speed scenarios
σே ೆಶ
௝ୀଵ ܴ‫ܲܨܮ‬௝
were investigated in this simulation, namely, 40, 80, 120 and
തതതതതതത
ܴ‫ ܲܨܮ‬ൌ ሺͺሻ 160 km/h. Fig 5 shows the average HOP versus different
ܰ௎ா mobile speeds for S1-HPO, S2-HPO, S3-HPO and the

42
proposed D-HPO algorithm. The results show that the proposed each mobile speed scenario. The results demonstrate that the
D-HPO algorithm causes a significant reduction in the average proposed D-HPO algorithm achieves a significant reduction in
HOP compared with other algorithms for all investigated RLF compared with other HPO algorithms for all mobile
mobile speeds. speeds. Although S3-HPO achieves a lower rate of RLF
compared with S1-HPO and S2-HPO, it yields higher rates of
HPPP due to low values of HOM and TTT (Fig 6).

Fig. 5. HOP with different mobile speeds and HPO algorithms

Fig 6 shows the average HPPP over the entire simulation


Fig. 7. RLFP with different mobile speeds and HPO algorithms
time for each mobile speed. The HPPP achievable with the D-
HPO algorithm is slightly lower than that with the S1-HPO, S2-
VII. CONCLUSION
HPO, S3-HPO and the proposed D-HPO algorithms for all mo-
bile speeds. This result is attributed to the higher HPPP ob- In this paper, a handover algorithm based on HCPs is pro-
tained when the HPO algorithm with static HCPs (S1-HPO, posed to minimize RLF and HPPP for HetNets. The proposed
S2-HPO and S3-HPO) uses non-optimized HCPs, leading to algorithm optimizes TTT and HOM values according to mobil-
non-optimal HCP values. Thus, the number of UHOs increases, ity scenarios identified for mobility robustness optimization.
resulting in high HPPP, especially in high mobility speeds. The performance of D-HPO algorithm was analyzed using two-
Moreover, a high HOP leads to high HPPP, wherein a signifi- tier model simulation, which consists of 4G and 5G networks,
cant reduction in HOP will lead to reductions in HPPP. where it was compared with three scenarios of static HCP val-
ues. Simulations have been performed for different mobile
speeds. The results have shown that the D-HPO improves the
system performance compared with the static values for most
mobile speeds. In future works, several parameters can be con-
sidered for handover decisions, such as UE mobile speed and
traffic load of each cell, in the optimization process.

ACKNOWLEDGMENT
This project has been sponsored by Telekom Malaysia, TM
R&D with project ID: MMUE/170014.

REFERENCES
[1] M. Agiwal, A. Roy, , & N. Saxena, ‘ Next generation 5G wireless
networks: A comprehensive survey’. IEEE Communications Surveys &
Tutorials, 18(3), 2016, 1617-1655.
Fig. 6. HPPP with different mobile speeds and HPO algorithms
[2] T. Minh. S. Nguyen, Kwon and H. Kim, "Mobility Robustness
Optimization for Handover Failure Reduction in LTE Small-Cell
Fig 7 shows the average RLFP for S1-HPO, S2-HPO, S3- Networks," in IEEE Transactions on Vehicular Technology, vol. 67, no.
HPO and proposed D-HPO algorithm with different mobile 5, pp. 4672-4676, May 2018.
speeds. The average RLF was independently calculated over all [3] A. Alhammadi, M. Roslee and M. Y. Alias, "Analysis of spectrum
handoff schemes in cognitive radio network using particle swarm
the UEs in the system network and over all simulation times for optimization," 2016 IEEE 3rd International Symposium on

43
Telecommunication Technologies (ISTT), Kuala Lumpur, 2016, pp. 103-
107.
[4] A. Alhammadi, M. Roslee and M. Y. Alias, "Fuzzy logic based
negotiation approach for spectrum handoff in cognitive radio
network," 2016 IEEE 3rd International Symposium on
Telecommunication Technologies (ISTT), Kuala Lumpur, 2016, pp. 120-
124.
[5] T. Jansen, I. Balan, J. Turk, I. Moerman, & T. Kurner, ‘ Handover
parameter optimization in LTE self-organizing networks’. In Vehicular
Technology Conference Fall, 2010 IEEE 72nd (pp. 1-5).
[6] D. Castro-Hernandez, , & , R. Paranjape. Optimization of handover
parameters for LTE/LTE-A in-building systems. IEEE Transactions on
Vehicular Technology, 2017
[7] M. Boujelben, S. B. Rejeb, & S. Tabbane, SON Handover Algorithm for
Green LTE-A/5G HetNets. Wireless Personal Communications, 95(4),
4561-4577, 2017.
[8] S. Chaudhuri, I. Baig, & D. Das, Self-organizing method for handover
performance optimization in LTE-advanced network. Computer
Communications, 110, 151-163, 2017.
[9] W. Zheng, H. Zhang, X. Chu, and X. Wen, "Mobility robustness
optimization in self-organizing LTE femtocell networks," EURASIP
Journal on Wireless Communications and Networking, vol. 2013, p. 27,
February 12 2013.
[10] 3GPP 2017. TR 25.931 V14.0.0., UTRAN functions, examples on
signalling procedures (Release 14). Valbonne - FRANCE,
http://www.3gpp.org/DynaReport/25931.htm.
[11] 3GPP 2017. TR 21.905 V14.4.0., Vocabulary for 3GPP Specifications
(Release 14). http://www.3gpp.org/DynaReport/21905.htm.
[12] Ericsson. 2015. R4-154516, Modified RRH Arrangement for HST SFN.,
http://www.3gpp.org/ftp/tsg_ran/WG4_Radio/TSGR4_76/Docs/.
[13] Po-Chiang Lin, Lionel F. Gonzalez Casanova, and K. S. Bakary Fatty,
“Data-Driven Handover Optimization in Next Generation Mobile
Communication Networks,” ‘„‹Ž‡ ˆ‘”ƒ–‹‘ ›•–‡•, vol. 2016,
Article ID 2368427, 11 pages, 2016.
[14] S. Barbera, et al., "Synchronized RACH-less handover solution for LTE
heterogeneous networks," 2015 International Symposium on Wireless
Communication Systems (ISWCS), Brussels, 2015, pp. 755-759.
[15] 3GPP 2017. TS 36.839 V11.1.0., Evolved Universal Terrestrial Radio
Access (E-UTRA); Mobility enhancements in heterogeneous networks.
France, http://www.3gpp.org/dynareport/36839.htm.

44

You might also like