CQI Classification Based Method To Improve Throughput in Wireless LTE Systems
CQI Classification Based Method To Improve Throughput in Wireless LTE Systems
CQI Classification Based Method To Improve Throughput in Wireless LTE Systems
Abstract
Advanced wireless networks such as 4G LTE have effective ways of reporting channel
quality through CQI values of the time and frequency varying channel whether it is fast
fading or slow fading. The main objective of using CQI values is to schedule the resource
blocks and maintain proper SNR by adjusting signal power or modulation. In doing so,
one must precisely be able to map CQI values to SNR values so that the objectives can be
fulfilled properly where an imperfect mapping can hamper the performance of wireless
system as packet scheduling can become difficult and unmanageable. In this work, a
detailed study of variation of channel SNR is presented and a classification based method
for scheduling packets is proposed for varying SNR. The proposed method is shown to
improve performance at link level and packet level throughout. Our future work is to
create a fine grain system considering various channel models and incorporate same in
the proposed model.
Keywords: LTE, Imperfect CQI, packet scheduling, fair queuing, proportional fair,
channel allocation, resource blocks
1. Introduction
Future Wireless Networks such as LTE (Long Term Evolution) and LTE-A (LTE-
Advanced) have novel ways to utilize frequency spectrum efficiently may it be using
OFDM (Orthogonal Frequency Division Multiplexing) or using optimal channel and
source coding. LTE also utilizes prediction resource allocation schemes including but not
limited to Adaptive Modulation and Coding (AMC), Channel Quality Index (CQI) and
Hybrid Automatic Retransmission Request (HARQ) as detailed in [1]. LTE system could
make utilized for uplink (towards the resource allocator from UE) alternately as a
downlink (from the resource allocator to the UE). In time division Multiplexing (TDM),
the accessible bandwidth is time divided among end users same time in frequency
division Multiplexing (FDM) same frequency groups are used to differentiate separate
end users. Different end users might utilize the FDM at the same time utilizing FDMA
(Frequency division different Access). In LTE and LTE-A, Orthogonal Frequency
Division Multiple Access (OFDMA) is used to attain higher data information. For
scheduling the resource, of LTE, RBs (Resource Blocks) are designed where user perform
hop beyond certain RB (time or frequency block) after the subsequent Resource Block.
Scheduling this hoping users among exceptional RBs a challenge so specific users have
special QoS (Quality of Service), delay, throughput and bandwidth requirements. On one
hand, LTE Release 8 is an outcome of research of many years in telecom industry which
is based on OFDM techniques. The deployment of LTE is mostly considered as a prime
contender for macro and micro cells providing improvised system capacity, coverage,
data rate and latency as compared to its previous generations. LTE-A (advanced) on the
Received (April 20, 2017), Review Result (August 30, 2017), Accepted (September 8, 2017)
other hand is an important improvement of LTE release 8 which enhances the user
experience in high mobility situations [8]. The community has added up to Release 14
into the research domain for which is all collectively known as LTE-A after release 8.
Wireless throughput is a term used to measure the performance of a wireless channel.
Techniques to improve performance of a wireless channel not only include the coding,
modulation and scheduling, but also require a deep understanding about the channel
features such as multipath fading, slow and fast fading, and channel gain. Upon
successfully analyzing these features, one can optimally design a scheme to schedule
resource blocks (RBs) to provide better throughput. There are many ways of improving
throughput of an LTE channel. Rate and delay feedback, channel gain equalization and
diversity techniques are well known in literature. Moreover, OFDM provides a neat way
to use single subcarrier in both time and frequency domain to transmit signal in a resource
block. While it is known that techniques such as FHSS can be used, packet scheduling in
a RB is still a major challenge as it requires a multi-criteria optimization methodology.
Various packet scheduling schemes are proposed in the literature including proportional
fair, round robin, fair queuing, weighted fair queuing, inverse weighted queuing and
inverse weighted fair queuing [2-5]. In order to create a better scheduling, focus is
provided on different aspects of a channel such as number of subcarriers, number of
resource blocks, quality of channel, modulation scheme used e.g. Most of the work in
literature focus on specifics of parameters as mentioned above. While this approach is
useful in many cases, the results can be improved if we can provide a holistic approach
which can focus on group based usage of scheduling [9-10].The advantage of
classification and grouping are as follows: (a) One can be characterized as a part of group,
(b) a group level guaranteed QoS can be maintained, (c) There is a lower computation
complexity at the downlink transmitter and computing resources and (d) Estimates of
channel can be used directly to improvise the system as a whole. In this work, a CQI
classification based method is proposed to improve throughput of a 4G LTE wireless
system. The paper is organized as follows: Section II details various techniques of
resource scheduling in LTE. Section III describes the importance of CQI and its usage in
modern networks for better channel utilization. Section IV explains our proposed
classification based method for packet scheduling. Results and analysis is presented in
Section V while Section VI concludes the paper with briefs on our future work.
On the off chance that eNB needs to allot just 20 RB's to UE, the data inside DCI will
empower UE to figure correct RBs distributed to it. In case eNB needs to send 100 bits,
the system will build control data information. To decrease the extent of control data,
LTE presents diverse resource distribution methodologies.
Network congestion and instability problems are reduced by different resource
scheduling methods. A plenty of methodologies are proposed for the segregation of
resources among users which can be elaborated as follows: (1) channel unaware, (2)
channel aware/QoS unaware, (3) Channel aware/QoS aware, (4) semi-persistent for VoIP
support and (5) energy aware [1].
Initially, “Channel Unaware” systems are basically produced for wired
correspondences wherein the transmission media is invariant regardless of the possibility
that there is any adjustment in time. Likewise, the likelihood of error occurrence in a
wired media is insignificant. In demonstrate hatred for that wired communications'
suspicions don't function admirably in wireless situation, they shape a base for some
practical and powerful scheduling strategies which can be explained as takes after:
(a) First In First Out: This is most straightforward calculation to distribute resources for
UE's. The main demand will be served or will get the assets first. This is an inefficient
calculation because of the way that meriting users don't get the resources.
(b) Round Robin: Here every single user will get resources in their designated time
term, yet even this is not an effective technique as it prompts deferral of execution.
(c) Blind Equal Throughput (BET): This is a genuinely better strategy as it takes the
past normal throughput of every user and based on the data received for every one of the
users, it will plan the users with slightest normal throughput by sharing a few resources
from the most elevated normal throughput user.
(d) Resource Preemption: Here it makes utilization of the priority techniques to allot
the resources to the UE. The user with most astounding need experiences resource
acquisition.
(e). Weighted Fair Queuing: To maintain a strategic distance from the starvation
condition happened due to the need strategies, weighted reasonable lining is utilized. Here
resources will be designated in view of the weights given.
(f). Guaranteed Delay: In this strategy an opportunity to live for every packet is
assigned. The packet must touch base inside that specific due time or is disposed of.
Secondly, in “Channel aware/QoS unaware methodology”, the channel condition
assumes a vital part. The most extreme channel limit is administered by Shannon's
principle. Given the transmission capacity and noise in the channel, the maximum
information rate in the channel can be resolved. The algorithms can be recorded as takes
after:
(a) Maximum Throughput (MT): In the event that the user has the best channel
condition, at that point that specific user will be allocated with most extreme throughput.
(b) Proportional fair (PF): Channel unaware BET can be utilized along with MT to
dispense resources proportionately which enhances the range productivity and
reasonableness in resource conveyance.
(c) Joint Time and Frequency space Schedulers: For this situation an arrangement of
dynamic UEs are chosen by Time Domain Packet Schedulers (TDPS) and the portion of
resources to those UEs is finished by the Frequency Domain Packet Scheduler (FDPS).
(d) Inverse weighted Fair Queuing (IWFQ): The network sections must be swapped so
as to represent BET and thus the weights related with every user will be straightforwardly
corresponding to throughput. In this way, in this calculation the resources are shared
proportionately among users by taking inverse of their weights joined to it [2].
𝐼𝑊𝐹𝑄 𝑅𝑅
𝑦𝑖,𝑙 = (1/𝑤𝑖 ). 𝑦𝑖,𝑙 (1)
(e) Proportional Inverse Weighted Fair Queuing (PIWFQ): PF, BET, IWFQ all these
three methods combined together, gives PIWFQ to achieve fairness in the scheduling
strategies [2].
𝑃𝐼𝑊𝐹𝑄 𝑀𝑇 𝐵𝐸𝑇 𝐼𝑊𝐹𝑄 𝑙 𝑔𝑖 (𝑇)𝑅𝑅 1
𝐺𝑖,𝑙 = 𝐺𝑖,𝑙 . 𝐺𝑖,𝑙 . 𝑦𝑖,𝑙 = 𝑓𝑖(𝑇−1) . 𝑤 . 𝑦𝑖,𝑙 (2)
𝑖
Figure 2 shows a Rayleigh faded signal with two different Doppler frequencies. A
simple channel model can be explained using equation below where the received signal is
a linear combination of: (a) x: transmitted signal, (b) h: channel gain and (c) n: Gaussian
noise.
u
y =h* x + n (3)
While the signal is highly noisy and time varying, one can also find the
correlation between two instances of a same signal and signal correlatio n with itself
as shown in Figure 3. We can clearly observe that when the signal correlates itself at
fd*deltaT = 0.4 that is after around every λ/2 cycle. Therefore, one can take
advantage of this property and receive correlated signal at the receiver every half
cycle of a signal. This property is highly useful in determining the SNR (signal to
noise ratio) of the signal at the transmitter and the receiver.
The analysis has many advantages. For example, one can tune the receiver to
receive only highly correlated signal and therefore to avoid highly noisy signal.
Also the channel inversion can be used at the receiver to overcome the phase shift of
the signal. These properties of a channel quality are highly important to understand
in order to optimize LTE channel scheduling and therefore throughput as CQI
(Channel Quality Indicator) values are provided in LTE to understand the quality of
channel. Channel utilization and CQI are detailed in the next section.
fluctuations and therefore one must verify its value by suitable pilot methods. The channel
utilization can be written as a function of accuracy of CQI reporting as it directly impacts
the modulation scheme and transmitted signal power.
Channel Utilization (CU) = function (CQI)(1-p(e)) (4)
Where p(e) is the probability of error in reporting CQI. As p(e) is dependent upon
fluctuations in the channel (Rayleigh channel is this case), we may argue that p(e) is a
Rayleigh distributed random variable with whose probability density function (pdf) can be
written as:
y = x2 exp(- x /2b )
2 2
b (5)
In case of a slow faded flat channel, we can assume that the CQI value do not change
much frequently. The problem arises in case of a fast fading channel wherein the SNR
values change much frequently and therefore there might also be error in reporting CQI to
the receiver. Figure 4 illustrates the channel utilization of a Rayleigh faded channel using
the appropriate CQI value. We can clearly see that the channel is utilized very poorly in at
many a times. In order to overcome this fluctuation, a method is proposed in the next
section of this paper.
4. Proposed Method
In order to overcome the situation of imperfect CQI we can use Forward Error
Correction (FEC) which requires previous levels of CQI [9]. In FEC approach, we
can use levels of CQI using a Group Based Moving Average (GBMA) technique as
shown in Figure 5. We arrange users in view of their evaluated CQI levels and pick
the proper packet scheduling algorithm for each gathering. Once the SNR levels are
calculated, we can choose the calculation use for various gathering of users. The
decision is sensible in light of the fact that when the clients who are as of now at
most noteworthy SNR levels can simply be given resources proportionately and they
would have the capacity to get great throughput owing their better channel
conditions when contrasted with the ones who are battling with resources and in
addition with the channel conditions. We utilize the GBMA technique for allotting
diverse booking calculation according to the value of user gathering. The scheduling
strategies utilized are Maximum Throughput (MT), Proportional Fair (PF), Inverse
Weighted Fair Queuing (IWFQ) and Proportional Inverse Weighted Fair Queuing
(PIWFQ).
Group 1
Group 2
Group N
On the off chance that, we consider the need of premium users, we can rapidly improve
the technique by utilizing a pre-settled measure of drifting transfer speed for need users
which can be allotted to others when not being used. In this way, the planning won't be
influenced for others. There might likewise be situation when the user bunches are unique
for quite a while and for this situation the general normal throughput will be low for the
system. We have similarly separated the sizes of SNR to allot the required planning
calculation. The same can likewise be proportionately partitioned and is a zone left for
investigation in future. This framework intensely depends upon the blemished CQI levels
which are a typical situation in wireless channels.
the group based method, the fairness is improved on a group level and therefore is
appropriate to say that a better group may provide better throughput.
Figure7 compares the average throughput for different number of users taking
advantage of the proposed scheme. It can be observed that even though the individual user
might be hampered from the scheme (Figure 6), the overall throughput is higher as
expected for a group as we treated the group as a whole to provide throughput. Table I
details various parameters used in the simulations. In cases where comparisons are made,
a higher or lower value of the parameter is used to compare when deemed appropriate. It
can be noted that appropriate values of frequencies are used in Section II to compare the
Rayleigh fading. Also, the simulations are run for a minimum of 50 times and average is
taken in the all the cases to avoid the impact of any randomness occurred due to software
itself.
References
[1] F. Capozzi, “Downlink packet scheduling in LTE cellular networks: Key design issues and a survey”,
IEEE Communications Surveys & Tutorials, vol. 15, no. 2, (2013), pp. 678-700
[2] S. Suganya, M. Sumit, S. L. Yaram and C. Ramesh, “Resource Scheduling Algorithms for LTE using
Weights”, in 2nd International conference on Applied and Theoretical Computing and Communication
Technology, SJBIT, Bangalore, (2013).
[3] Phunchongharn, Phond, E. Hossain and D.I. Kim, “Resource allocation for device-to-device
communications underlaying LTE-advanced networks”, IEEE Wireless Communications, vol. 20, no. 4,
(2013), pp. 91l.-100.
[4] M. Sumit, “A joint parametric prediction model for wireless internet traffic using Hidden Markov
Model”, Wireless networks, vol. 19, no. 6, (2013), pp. 1171-1185.
[5] R. Kwan, C. Leung and J. Zhang, “Proportional fair multiuser scheduling in LTE”, IEEE Signal
Processing Letters, vol. 16, no. 6, (2009), pp. 461-464.
[6] G. Piro, “Two-level downlink scheduling for real-time multimedia services in LTE networks”, IEEE
Transactions on Multimedia, vol. 13, no. 5, (2011), pp. 1052-1065.
[7] 3GPP TS 36.101, “User Equipment (UE) radio transmission and reception,
[8] A. Ghosh, “LTE-advanced: next-generation wireless broadband technology [Invited Paper]”, IEEE
Wireless Communications, vol. 17, no. 3, (2010).
[9] S. Suganya, and C. Ramesh, “SNR based Predictive Packet Scheduling in LTE with Varied CQI”,
Indian Journal of Science and Technology, vol. 9, no. 47, (2016).
[10] Y. X. Wang, “Packet Scheduling in LTE with Imperfect CQI.”, International Journal of Advanced
Research in Computer Science and Software Engineering, vol. 3, no. 6, (2013), pp. 6-13.
[11] G. Wang, “Power scheduling of Kalman filtering in wireless sensor networks with data packet
drops”, arXiv preprint arXiv:1312.3269, (2013).
[12] W. K. Lai and C.-L. Tang, “QoS-aware downlink packet scheduling for LTE networks”, Computer
Networks, vol. 57, no. 7, (2013), pp. 1689-1698
[13] M. Kalil, A. Shami and A. Al-Dweik, “QoS-aware power-efficient scheduler for LTE uplink”, IEEE
Transactions on Mobile Computing, vol. 14, no. 8, (2015), pp. 1672-1685.
[14] R. E. Kalman, “A new approach to linear filtering and prediction problems”, Journal of basic
Engineering, vol. 82, no. 1, (1960), pp. 35-45.
[15] K. You, N. Xiao and L. Xie, “Kalman filtering with scheduled measurements”, Analysis and Design of
Networked Control Systems. Springer London, (2015), pp. 269-291.