International Journal of Wireless Communications and Mobile Computing

2018; 6(2): 37-42

http://www.sciencepublishinggroup.com/j/wcmc
doi: 10.11648/j.wcmc.20180602.11
ISSN: 2330-1007 (Print); ISSN: 2330-1015 (Online)

Beamforming Technique Assisted by Machine Learning

Algorithm for Next Location Prediction
Hussein Safwat Hasan Hasan, Humor Hwang*
Department of Information and Communications Engineering, Myongji University, Yongin City, Republic of Korea

Email address:
*
Corresponding author

To cite this article:

Hussein Safwat Hasan Hasan, Humor Hwang. Beamforming Technique Assisted by Machine Learning Algorithm for Next Location
Prediction. International Journal of Wireless Communications and Mobile Computing. Vol. 6, No. 2, 2018, pp. 37-42.
doi: 10.11648/j.wcmc.20180602.11

Received: December 5, 2018; Accepted: January 2, 2019; Published: February 14, 2019

Abstract: Technological improvement towards the development of location prediction advancement had attracted a great
attention due to its broad application. Herein, intercalation of two widely scrutinized techniques were fused to form a
synchronized location forecasting system. Using the underlying concept of beamforming (BF), an array of retro directive
beams towards the phase sectioned field were emitted to determine the specific location of an entity or receiver. The receiver
collects and sends back the data of beam emissions with respect to time and phase, machine learning (ML) technique were
used to analyze the transcribed data to determine the phase with optimum beam reading that corresponds to the location of the
receiver. Series of historical context will be analyzed by ML to predict the next location of the entity, emitting an array of
signals pointing at the predicted location. Automatic location forecasting synchronization due to intricate systematic design
were demonstrated. It should be noted that BF-ML technique collaboration for location prediction had never been reported
before and driven by its advantages in wireless networking (such as elimination of interference and privacy issues) field of
utilization can still be expanded.
Keywords: Machine Learning (ML), Beamforming (BF), Scans, Phase Delay, Tracking Algorithm

network (Wi-Fi), Global System for Mobile Communications

1. Introduction (GSM) and Global Positioning System (GPS). And
depending on the application, ML algorithms can be tuned
Forecasting the next location of the receiver through its and constructed to satisfy its purpose [8, 9].
historical motion is useful for creating smart applications. In this publication, we introduce a new approach on next
The advancement on the techniques on approximation and location prediction through collaboration of ML
anticipation of the future position of the receiver enable the methodology and beamforming technique (BF). Phase array
progression of location predicting application and services. trajectory guiding and receiving using BM have attracted
Generally, wireless system monitors the location predicting great attention in the past decade due to its robust
system by gathering and providing them the necessary application. The array of beams can be synchronized on the
information prior application [1, 2]. In order to communicate desired direction by using the BF phase array technique. And
with the receiver effectively, the sender should be intelligent to take advantage of the versatility of ML technique we
enough to determine the position of the receiver at any point propose a new location prediction system with automatic
in time [3]. The idea of anticipating the location of the synchronization ability which have never been reported on
receiver by machine learning (ML) algorithms has already any BF application. Experimental design use to collect
been established [4]. Status data of the receiver at specified context of information through scans from lower to higher
time is recorded and analyzed accordingly and transcribe as time phases delay until the optimum phase with fix gap
context [5, 6]. Numerous machine learning techniques had among the phases. The context readings were projected as the
been recognized towards the technological advancement of predicted value of the transported power to the entity or
location forecasting [7], such as wireless communication receiver on a specific amount of time had been received.
38 Hussein Safwat Hasan Hasan and Humor Hwang: Beamforming Technique Assisted by
Machine Learning Algorithm for Next Location Prediction

Historical motion of the entity was gathered from several the signal arrives at the antenna “n” leads in phase with n d
transitions, analysis of the labeled context features was sin α. This gives the description of the Array Propagation
supervised by ML classifications [10, 11]. The collected data Vector that it comprises.
about sequential movement were analyzed and interpreted to
generate a suitable location prediction. Once forecasted
position has been determine, the BF and ML system will
synchronize the array of beam targeting and directing
towards the proposed next location of the receiver, enabling
the proposed system to have an auto-synchronization
capability onto a repetitive cycle.
Location forecasting has garnered increased attention by
deploying in location control expecting to boost the standard
of the services. Therefore, the utilization of BF approach
effectively manipulates the high-performance beam
emissions channeled towards the receiver with reduced
interference and elevated intensity. The proposed system
addresses the issues governing privacy breach and reception
dilemma of the next location prediction, the utilization of
beamforming technique to forecast the succeeding position of
the entity using the context of its current and previous
location without invading the receiver’s privacy and
eliminating the reception interference. New technique for the Figure 1. Beam forming diagram.
resolution towards the challenge in determination of the next
probable location of an entity with ML-BF system had been The signal can be manipulated in specific direction while
scrutinized with advantageous properties such as auto- on the other hand use of amplitude weights permits to
synchronization, avoidance of privacy violation and steer the sidelobe intensity and neglect the control for radio
reception interference elimination. intrusion rejection. At the beamforming system, antenna N
can control the signal on the chosen direction and nullify
(rejects beam manipulation) at optimum N-1 intrusive signals
2. Beamforming approaching from N-1 opposite directions.
Anticipating the possible location of an entity has been a
∑ (2)
challenge in the networking industry. Location forecasting
has received a great attention due to its robust application in Wherein the radiation response S (α) at the array, while
various discipline. Hence, the reason of applying ξ=2π/λ and λ corresponds to the wavelength with the same
beamforming technique observing the elevate performance of weights all over the array, ( =1) in the diagram. The
beam navigation to collect a lower level of interference with details of the direction of influx of signals can be discern as:
higher intensity [12]. Beamforming is defined as the spatial
filtering wherein the suitable signal receptor that corresponds 1 … … (3)
to group of antennas that can be used to maneuver the
In this manner the Array Factor can be described as:
transmitted signals in a fix direction [13]. Various methods
$
and technologies have been generated to focusing beam ! ", (4)
emissions towards the receiver direction to achieve an
effective communication and tracking ability [14]. Time Hence, the main benefits of applying the beamforming is
delay shift (time delay) and phase shift were the most that numerous access is achieved through the space
common strategy to direct the beams projection whether in partitions, since a beamformer can manipulate its direction
analog or digital mode. The time delay program can form and towards a certain signal, the beamformer focuses in a specific
manipulate the beam by altering and adjusting the time delay direction, the antenna’s capability to sense can be improved
pace that are independent from the steered frequency and for a better signal to noise ratio, specifically when getting
bandwidth. The increase in normalized depends on the weak signals. And these properties are all attained
bandwidth B and the delay ∆t (variation in time of influx of electronically with no physical activity of the receiving
the front wave at the receiving antenna due to the dimension antennas [15, 16].
of array) as described in the following expression.
∆ 3. The Proposal
(1)
∆
Herein, we demonstrated a method of finding the next
Shown in Figure 1 a basic beamforming diagram, destination of a receiver with reference on its present
considering the signal reception of the antenna 1 at phase = 0, location. To address the issues governing the estimation of
 International Journal of Wireless Communications and Mobile Computing 2018; 6(2): 37-42 39

the next location prediction, we historically log the places time having fixed amplitude, the receiver on the other hand
that the receiver visited with variations on the antennas will store the coming signals and transcribe the value of
beamforming phase at a given moment of time. The power and time under each phase Figure 3. The stored data of
beamforming system is composed of array of antennas each scan is fed to the tracking algorithm to determine phase
equipped with phase delay shifters as projected in Figure 2. with the highest power received, the results were supplied to
Assisted by machine learning, shifters automatically navigate the machine learning algorithm to predict the next location of
signals towards the receiver relying onto historical context. the receiver. The phase delay automatically directs the
The forecasting is initiated by several scans with beam signals to the receiver based on the predicted phase value of
shifted on different phase delay settings in a certain amount the machine learning.

Figure 2. Beamforming System Structure.

Figure 3. The Tracking algorithm.

Providing a sequence of various places that comprises the specified amount of time as defined in the chart projected
necessary data needed for training. Considering % below.
&' , … , ' ( to be the set of values of location visited, the )
&* , … , * ( will be the set of time records, and + ./01.2' ∈ % 456% 78 '|) *- , + ,*2 *- , %
&, , … , , ( to will be the set of context data,
':, *- ;< (5)
where , signifies as a set of featured pair values that are
available at *- . The context details in our experiment are
It should be noted that this model has been applied on
gathered by administrating a full scan starting from the
numerous studies that deals with location forecasting that is
lowest phase shift up to the highest phase with a firm gap in
cited to be a classification problem that uses the historical
between each phase. The scanning outcomes represented as
the assessed value of the delivered power to the receiver at a context to determine the next point location using the current
40 Hussein Safwat Hasan Hasan and Humor Hwang: Beamforming Technique Assisted by
Machine Learning Algorithm for Next Location Prediction

location as a starting point.

4. Experimental Analysis
The performance of the proposed technology was analyzed
and evaluate by implementing Python algorithm. In the
evaluation, the pygame library were applied to create and
simulate the virtualization of the proposed phenomenon. An
array of antennas was set and will serve as the beamformer
transmitter (with number of antennas from 1 to N), once the
receiver transmitted a signal as shown in Figure 4.
The power will be transmitted and capture through the
antennas. To calculate the performance of the system, free-
space path loss (FSPL) were expressed through the value of
power losses and corresponds to the reciprocal of the power
gain.
= ?
! 5% 4 < (6)
>

Where λ is stands for the speed of light while d will be

determine by applying the distance between two circles
formula.

@ A 22 C 21 ? D E2 C E1 ? (7)

The power transmittance will on either watt or dBm media,

to maximize the power reception, the power fed to the
program which is in watt will be transformed to dBm prior
sending it to the receiver. Then the receiver will interpret the Figure 4. Transmittance of signal at (a) Zero phase delay between the
power signals and save it in watt unit and feed it back antennas, (b) Maximum phase delay between the antennas and (c) Minimum
towards the antennas, the power will be calculated with the phase delay between the antennas.
aid of the equation below.
The scan will start from the lowest phase (-100) to the
LM
5 F 10 ∗ log 41000 ∗ < (8) highest phase (0) in a gap of 20 degrees between each phase
N
in a time of *- from 0 to 50 for each phase, decreasing the gap
5N or adding more phases will give us more accurate readings.
5 10 log D 30
1 One scan phase can be interpreted through the projected
Table 1 and Figure 5.
P / : 1 30 @R1
S , TU @ 5:V / W X@ / Y:V / 1 30 @R1 C
=
20 log 4 < (9)
>

Table 1. Power readings with respect to time.

Phase Power For Each Amount Of Time

[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2.583463619340189, 2.5864906755903156,
-100 2.5896283049689104, 0, 2.5928442962395457, 2.5960807761738103, 0, 2.5992427224954247, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
0, 0, 0, 0, 0]
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 7.759582599899415, 5.188925072413356,
-80
2.5992427224954247, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 5.195323498669235, 5.182472601208456,
-60
2.5864906755903156, 2.583463619340189, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2.5992427224954247,
-40 2.5960807761738103, 0, 2.5928442962395457, 2.5896283049689104, 0, 2.5864906755903156, 0, 2.583463619340189, 0, 0, 0,
0, 0, 0, 0]
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2.5992427224954247, 0,
-20 2.5960807761738103, 0, 0, 2.5928442962395457, 0, 2.5896283049689104, 0, 0, 2.5864906755903156, 0, 0,
2.583463619340189, 0]
[0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 2.5992427224954247, 0, 0,
0
2.5960807761738103, 0, 0, 0, 2.5928442962395457, 0, 0, 2.5896283049689104, 0, 0, 0]
 International Journal of Wireless Communications and Mobile Computing 2018; 6(2): 37-42 41

Figure 5. Power readings through time as shown in Table 1.

The collected data of power in a certain amount of time on were also addressed by utilizing beamforming since the
every phase will documented, each phase data sets were platform and mode applied enhance the connection with
analyzed to determine which phase provided or projects the receiver to avoid the noise and fluctuation caused by the
highest intensity power, three loops scan were applied to environment compared to the one applied in the mobiles, this
filter out and determine the highest power intensity. In our makes it more effective and efficient in predicting locations.
proposed method, efficient creation and interpretation of the The proposed scheme can be applied on single and multiple
recordings of time with respect to the location is users depending on the applications requirement to cater the
demonstrated. The corresponding values of L (location) at a markets demand.
given T (Time) on every C (phase) were demonstrated on
Table 2.

Table 2. Corresponding values.

References
[1] Gomes, João Bártolo, Clifton Phua, and Shonali
Location Z Time [ \] Context C (phase)
Krishnaswamy. "Where will you go? mobile data mining for
' ,.X 1 32 -80
next place prediction." International Conference on Data
'? ,.X 2 32 -60
Warehousing and Knowledge Discovery. Springer, Berlin,
'^ ,.X 3 29 -60
Heidelberg, 2013.
'= ,.X 4 26 -60
'` ,.X 5 30 -20 [2] Xia, Linyuan, Qiumei Huang, and Dongjin Wu. "Decision
'b ,.X 6 29 -20 Tree-Based Contextual Location Prediction from Mobile
' ,.X X X X Device Logs." Mobile Information Systems 2018 (2018).

[3] Barwise, Patrick, and Colin Strong. "Permission‐based

5. Conclusion mobile advertising." Journal of interactive Marketing 16.1
(2002): 14-24.
In this paper, a new approach on predicting the next
location were proposed and scrutinized by combining BF [4] Anagnostopoulos, Theodoros, Christos Anagnostopoulos, and
with ML technique to generate an automatic location Stathes Hadjiefthymiades. "Mobility prediction based on
machine learning." Mobile Data Management (MDM), 2011
prediction synchronization system. Algorithm composition 12th IEEE International Conference on. Vol. 2. IEEE, 2011.
and simulation of the proposed proposition were run and
tested with the aid of python. Systematic and visual [5] Jaiswal, Ayush, et al. "Location Prediction with Sparse GPS
interpretation were explored and demonstrated to fully gasp Data." Proceedings of the 8th International Conference on
Geographic Information Science. 2014.
the propounded system design. Supporting theoretical and
mathematical calculation were also presented to discerned [6] Anagnostopoulos, Theodoros, Christos Anagnostopoulos, and
the algorithm component makeup and validate the system Stathes Hadjiefthymiades. "An adaptive machine learning
legitimacy. With sufficient theoretical results it was found algorithm for location prediction." International Journal of
Wireless Information Networks 18.2 (2011): 88-99.
that BF supported with ML algorithm scheme can be a
platform for location prediction application with [7] Wu, Ruizhi, et al. "Location prediction on trajectory data: A
advantageous characteristics. Such as the elimination of review." Big Data Mining and Analytics 1.2 (2018): 108-127.
privacy exploitation that is common in next location [8] Lu, Zhongqi, et al. "Next place prediction by learning with
prediction in the mobile application. Interference reduction multiple models." Mobile Data Challenge Workshop. 2012.
42 Hussein Safwat Hasan Hasan and Humor Hwang: Beamforming Technique Assisted by
Machine Learning Algorithm for Next Location Prediction

[9] Laurila, Juha K., et al. "The mobile data challenge: Big data for Efficient Digital Beamforming." Red River Engineering.
mobile computing research." Pervasive Computing. No. EPFL-
CONF-192489. 2012. [14] Tall, Abdoulaye, Zwi Altman, and Eitan Altman. "Multilevel
beamforming for high data rate communication in 5G
[10] Yavaş, Gökhan, et al. "A data mining approach for location networks." arXiv preprint arXiv: 1504.00280 (2015).
prediction in mobile environments." Data & Knowledge
Engineering 54.2 (2005): 121-146. [15] Darwish, Mohammad, and Cecil Lau. "A Software Radio
Architecture for Smart Antennas." Spectrum Signal
[11] Palma, Andrey Tietbohl, et al. "A clustering-based approach for Processing, Inc., Vancouver, Canada (1998).
discovering interesting places in trajectories." Proceedings of
the 2008 ACM symposium on Applied computing. ACM, 2008. [16] Shenghua, Zheng, Xu Dazhuan, and Jin Xueming. "A new
receiver architecture for smart antenna with digital
[12] Kalousis, Alexandros. "Predicting the Location of Mobile beamforming." Microwave, Antenna, Propagation and EMC
Users: A Machine Learning Approach." (2009). Technologies for Wireless Communications, 2005. MAPE
2005. IEEE International Symposium on. Vol. 1. IEEE, 2005.
[13] Schreiner, Clint. "Utilizing Digital Down Converter for