Energy and spectral efficiency trade-off in OCDMA-PON assisted by non-linear programming methods

• Published in: Computer networks (Amsterdam, Netherlands : 1999) Vol. 189; p. 107920
• Main Authors: Pendeza Martinez, Cristiane A., Abrão, Taufik, Martinez, André Luís Machado
• Format: Journal Article
• Language: English
• Published: Amsterdam Elsevier B.V 22.04.2021; Elsevier Sequoia S.A
• Subjects: Bi-objective optimization (BOO); Code Division Multiple Access; Convexity; Efficiency; Energy conversion efficiency; Exhaustive search (ES); Linear programming; Majoration–minimization (maMi); Networks; Nonlinear programming; OCDMA networks; Optimization; Performance evaluation; Quadratic programming; Quality of service architectures; Resource efficiency (RE); Sequential quadratic programming (SQP); Spectral efficiency (SE); Tradeoffs; Weighted sum method (WS); Weighted sum method (WS); Sequential quadratic programming (SQP); Spectral efficiency (SE); Resource efficiency (RE); OCDMA networks; Majoration–minimization (maMi); Exhaustive search (ES); Bi-objective optimization (BOO)
• ISSN: 1389-1286; 1872-7069
• DOI: 10.1016/j.comnet.2021.107920

Two important metrics for performance evaluation in optical code division multiple access networks (OCDMA) are energy efficiency (EE) and spectral efficiency (SE). Both metrics are investigated in this work through the bi-objective optimization approach by analyzing the trade-off between SE and EE. We first formulate the SE–EE optimization as a bi-objective optimization (BOO) problem with minimum rate requirement constraints. Then, considering the non-convexity of the BOO problem, it is converted into a single-objective optimization (SOO) problem by utilizing weighted sum method. Nonlinear programming methods as sequential quadratic programming method (SQP) and majoration–minimization (MaMi) are applied to solve the BOO problem, with QoS guarantees for the OCDMA networks. Resource efficiency (RE) is able to explore the trade-off between EE and SE in OCDMA networks. The optimization methods were applied and their performance-complexity trade-offs are compared with the exhaustive search (ES) method. Numerical results were performed considering practical and realistic scenarios, including a wide range of nodes, while the solutions obtained by the methods are represented in the Pareto front.