Robust optimization algorithm of RF MEMS switches considering uncertainties

• Published in: Integration (Amsterdam) Vol. 104; p. 102470
• Main Authors: Yan, Hao, Jia, Yaning, Zeng, Chuangyuan, Liao, Xiaoping, Shi, Xiao
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.09.2025
• Subjects: Evolutionary algorithm; RF MEMS; Robust optimization; Surrogate model; Robust optimization; RF MEMS; Surrogate model; Evolutionary algorithm
• ISSN: 0167-9260
• DOI: 10.1016/j.vlsi.2025.102470

Efficient robust design of RF MEMS switches requires balancing stringent performance criteria with inherent uncertainties. This paper proposes a Comprehensive Robust MEMS Optimization (CRMO) framework that integrates a Surrogate-assisted Differential Evolution with Screening Constraints (SDESC) and a Surrogate-assisted Multi-Objective Worst-case (SMOW) analysis method using both global and local regression models with particle swarm optimization (PSO). The SDESC algorithm adaptively adjusts constraint evaluations based on the proportion of feasible solutions, significantly reducing computational overhead, while SMOW efficiently handles multi-objective worst-case scenarios. Experimental evaluations on a 35 GHz series switch and an 10 GHz shunt switch demonstrate substantial performance and efficiency improvements. Specifically, for the series switch, the worst-case insertion loss improved from −6.742 dB to −0.134 dB, and the driving voltage was reduced from 58.345 V to 37.933 V; for the shunt switch, isolation was enhanced from −9.586 dB to −18.853 dB. Furthermore, the proposed algorithm achieves speedup from 3.2 × to 45 × over traditional PSO methods, confirming its advantage in both robustness and computational efficiency. This paper presents an efficient robust optimization approach for RF MEMS switches considering uncertainties. The key contributions of our work include:•Efficient optimization via individual-based model strategy adaptively adjusting constraints.•Handles MEMS worst-case analysis w/ multiple variables & objectives using hybrid regression & PSO.•Combines constrained optimization & worst-case analysis for robust RF MEMS design solutions.•Auto-generates MEMS design scripts; open-sourced code for reproducibility.