Resource Optimization Algorithm Based on Vertical Federated Learning in VLC/RF Hybrid Systems

Zhongtian Du; Wuwei Huang; Yang Yang

doi:10.3788/LOP232054

Laser & Optoelectronics Progress, Volume. 61, Issue 13, 1306001(2024)

Resource Optimization Algorithm Based on Vertical Federated Learning in VLC/RF Hybrid Systems

Zhongtian Du¹, Wuwei Huang², and Yang Yang^2、*

¹Science and Technology Innovation Department of China Telecom Digital Intelligence Technology Co., Ltd., Beijing 100035, China

²School of Information and Communication Engineering, Beijing University of Posts and Telecommunications, Beijing 100876, China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(10)

Fig. 1. Vertical federated learning architecture based on VLC/RF hybrid system

Download full size

Fig. 2. MLP Structure

Download full size

Fig. 3. Comparison of training loss in different algorithms

Download full size

Fig. 4. Comparison of model accuracy in different algorithms

Download full size

Fig. 5. BER comparison of different algorithms

Download full size

Fig. 6. Comparison of training loss using different channel estimation algorithms

Download full size

Fig. 7. Comparison of model accuracy using different channel estimation algorithms

Download full size

Table 1. Parameter of MLP
View table
Table 1. Parameter of MLP
Parameter Value
Number of inputs 2
Number of hidden layers 1
Number of neurons 512
Iteration rounds 200
Training algorithm Levenberg-Marquardt
Activation function TanSig

Table 2. VFL-VLC/RF algorithm

View table

Table 2. VFL-VLC/RF algorithm

Input	Packet error rate per user $n$ ， $q_{n} (r_{n}^{U}, P_{n})$ Calculate energy consumption per user $n$ ， $E_{n}^{P}$ Transmission energy consumption per user $n$ ， $E_{n}^{M}$
Output	Transmission power per user， $P_{n}^{*}$ Selected user collection， $S$
1	Optimizing PER during transmission using MLP based channel estimation algorithm， $q_{n}^{M L P} (r_{n}^{U}, P_{n})$ ← $q_{n} (r_{n}^{U}, P_{n})$
2	Obtain the optimal transmission power for each user in uplink RB based on Eq. （15）， $P_{n}^{*}$
3	Using the definition of weights in bipartite graph matching problems Eq. （18） and the Kuhn-Munkres algorithm to solve optimization problems Eq. （17）
4	Process of longitudinal federated learning using the optimal user selection set $S$ and the optimal transmission power vector $P^{*}$

Table 3. Simulation parameters

View table

Table 3. Simulation parameters

Parameter	Value
Transmission optical power of each VLC AP， $P_{v}$ /W	9
Modulation bandwidth of LED lights， $B$ /MHz	40
Physical area of PD， $A_{p}$ /cm²	1
Half intensity radiation angle， $θ_{1 / 2}$ /（°）	60
Optical filter gain， $T_{s} (θ)$	1.0
FoV of receiver， $Θ_{F}$ /（°）	90
Refractivity， $n$	1.5
Photoelectric conversion efficiency， $γ$ /（A/W）	0.53
Noise power spectral density， $N_{0}^{V L C}$ ， $N_{0}^{R F}$ /（A²/Hz）	10^-21
RF total bandwidth， $B_{R}$ /MHz	20
Transmission power of BS， $P_{B}$ /W	1
Total number of users， $N$	50
Delay requirement， $T_{r o u n d}$ /s	2.5
Energy consumption requirements， $γ_{n E}$ /J	2
Energy consumption coefficient， $α$	2×10^-28
User update size， $s$ /Mb	1

Tools

Get Citation

Copy Citation Text

Zhongtian Du, Wuwei Huang, Yang Yang. Resource Optimization Algorithm Based on Vertical Federated Learning in VLC/RF Hybrid Systems[J]. Laser & Optoelectronics Progress, 2024, 61(13): 1306001

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites