Utafiti wa kitaaluma, lugha inayoeleweka

Verianla | Akademik Araştırmalardan Türkçe Ekonomi ve Bilim İçerikleri

27 Septemba 2026, Jumapili
VERİANLAUchapishaji huru wa sayansi
Fungua au funga menyu
...
Home / Sayansi Tumizi / MATLAB / Uigaji wa uamuzi mahiri wa handover unaotegemea Support Vector Machine katika mitandao isiyotumia waya kwa MATLAB
MATLAB

Uigaji wa uamuzi mahiri wa handover unaotegemea Support Vector Machine katika mitandao isiyotumia waya kwa MATLAB

Simulation ya MATLAB/Simulink imetumia eneo la 1000 × 1000 m, base stations tano na mtumiaji mmoja anayesogea kwa kasi ya 3–60 km/h kwa random waypoint mobility. Channel imejumuisha log-distance path loss, log-normal shadowing na Rayleigh fading.

26/08/2026  Veri Anla Imetazamwa mara 54
Uigaji wa uamuzi mahiri wa handover unaotegemea Support Vector Machine katika mitandao isiyotumia waya kwa MATLAB

Simulation ya MATLAB/Simulink imetumia eneo la 1000 × 1000 m, base stations tano na mtumiaji mmoja anayesogea kwa kasi ya 3–60 km/h kwa random waypoint mobility. Channel imejumuisha log-distance path loss, log-normal shadowing na Rayleigh fading.

Idadi ya handover ilipungua kutoka 76.89 hadi 34.10, sawa na 55.7%. Ping-pong events zilipungua kutoka 41.16 hadi 17.56, au 57.3%. Packet loss probability ilishuka kutoka 0.2648 hadi 0.2486, huku effective throughput ikiongezeka kutoka 0.8486 hadi 0.9022.

Vipengele vya SVM

  • RSS — nguvu ya signal inayopokelewa.
  • SINR — ubora wa signal dhidi ya interference na noise.
  • Kasi ya mtumiaji — muktadha wa mobility.
  • Base-station load — matumizi ya resource, yaliyowekwa katika kiwango cha 0–1.

Mafunzo ya modeli

RBF kernel imetumika. C na kernel parameter zimetunishwa kwa grid search na k-fold cross-validation. Features zimenormalishwa; 70% ya data imetumika kwa training na 30% kwa testing.

Chanzo hakitoi final C, kernel parameter, thamani ya k au jumla ya samples.

Simulation environment

KigezoThamani
Eneo1000 × 1000 m
Base stations5
Watumiaji hai1
Kasi3–60 km/h
Muda200 s
Time step1 s

Verianla Live: Threshold dhidi ya SVM

KipimoThresholdSVMMabadiliko (%)Source
Handover76.8934.10-55.7Table I
Ping-pong41.1617.56-57.3Table I
Packet loss0.26480.2486-6.1Table I
Effective throughput0.84860.90226.3Table I
 

Umuhimu unaowezekana Afrika Mashariki

Utafiti haukutumia data ya mtandao halisi wa Afrika Mashariki. Hata hivyo, maeneo yenye ukuaji wa 4G/5G, mobility ya mijini na barabara za masafa marefu yanaweza kunufaika na utafiti wa adaptive handover unaotumia channel quality, kasi na cell load kwa pamoja. Deployment halisi ingehitaji training na validation kwa radio measurements na traffic patterns za eneo husika.

Vikwazo vya utafiti

Simulation ina mtumiaji mmoja tu na base stations tano. Haipimi user-density effects, large-scale cellular interaction au computational scaling katika operator network kubwa. Baadhi ya channel parameters, SVM hyperparameters na idadi ya Monte Carlo runs pia hazijaripotiwa kwa namba.

Mbinu na Matokeo ya Utafiti

KipimoThresholdSVMRelative change
Handover76.8934.10−55.7%
Ping-pong41.1617.56−57.3%
Packet loss0.26480.2486−6.1%
Effective throughput0.84860.9022+6.3%

Matokeo yanaonyesha kwamba katika mazingira yaliyosimuliwa SVM imepunguza unnecessary handover na ping-pong kwa kiwango kikubwa. Waandishi wanapendekeza future work kwenye multi-user scenarios, online learning na deep reinforcement learning.

Maelezo ya Chanzo na Mbinu

Waandishi: Andicho Haryus Wirasapta; Tiara Deta Pamungkas

Jarida: Jurnal LITEK: Jurnal Listrik Telekomunikasi Elektronika

Volume/Issue: 23(1), 30–37

Publication: 1 Machi 2026

DOI: 10.30811/litek.v23i1.104

Funding / data availability / conflict of interest: hakuna taarifa tofauti katika toleo lililochunguzwa.

Hii ni MATLAB/Simulink simulation, si field trial ya operator network. Parameter values ambazo hazikuwekwa wazi katika chanzo hazijaongezwa.


Shiriki:

Maoni huchapishwa baada ya kukaguliwa.Maoni yako yatapitia mchakato wa idhini na yataonekana yakikubaliwa.

Acha maoni

Anwani yako ya barua pepe haitachapishwa. Sehemu za lazima zimewekewa alama ya *

Your experience on this site will be improved by allowing cookies Cookie Policy