BROWSE
Detail
Journal Article
DNN-based algorithm for joint SIC ordering and power allocation in downlink NOMA-enabled heterogeneous networks
- Authors
- Donghyeon Kim, Jung-Bin Kim, Haejoon Jung, In-Ho Lee
- Issue Date
- 2024-12
- Citation
- ICT EXPRESS, v.10, no.6, pp.1301-1307
- ISSN
- 2405-9595
- Publisher
- ELSEVIER
- Language
- English
- Type
- Journal Article
- Abstract
- In the heterogeneous network (HetNet) employing downlink non-orthogonal multiple access (NOMA), we focus on the non-convex optimization problem to optimize the spectral efficiency (SE) while the users satisfy the quality-of-service (QoS) requirement. In the previous work, the optimal joint successive interference cancellation and power allocation (JSPA) algorithm for maximizing SE is proposed to solve the mixed-integer non-linear programming (MINLP) problem in NOMA-enabled HetNet. However, the optimal solution requires exponential complexity by the number of base stations (BSs). Therefore, we present a deep neural network (DNN)-based algorithm for JSPA to reduce the complexity. In particular, to deal with the MINLP-based JSPA problem, we reformulate it into an equivalently simple problem that optimizes only the power consumption of BSs. Then, we introduce the unsupervised DNN-based method for JSPA to handle the simplified problem. The presented scheme yields improved SE and outage performance compared with traditional DNN-based methods. Additionally, we propose a user selection scheme with low complexity to enhance the SE of the proposed DNN-based power allocation. Through simulations, we illustrate that the suggested DNN-based scheme can attain SE performance similar to that of the optimal scheme.
- KSP Keywords
- Deep neural network(DNN), Low complexity, Mixed-integer non-linear programming(MINLP), Non-convex optimization problems, Nonorthogonal multiple access(NOMA), Outage performance, Power Consumption, Spectral efficiency(SE), base station, heterogeneous network, linear programming(LP)
This work is distributed under the term of Creative Commons License (CCL)
(CC BY NC ND)
(CC BY NC ND)
