A memory-adaptive Karma–Dharma optimizer for global optimization and VLSI design

While many recent metaheuristics rely on elaborate biological or physical metaphors, their underlying mathematical mechanics often remain obscured. This paper introduces the Karma–Dharma Optimizer (KDO), a transparent, memory-adaptive metaheuristic designed for continuous and constrained engineering optimization. KDO couples two complementary phases: a memory-guided exploration phase that utilizes a historical success archive (the Dharma operator) and a structured exploitation phase (the Karma operator). An adaptive stagnation-recovery mechanism, termed Nirvana Reset, is further incorporated to help the search escape locally stagnant regions. The global search capability and dimensional scalability of KDO are extensively evaluated on the CEC-2017 and CEC-2022 benchmark suites across dimensions up to $$D=50$$ . Furthermore, KDO is applied to four constrained analog, RF, and VLSI circuit design problems: a two-stage Miller operational amplifier, an RF CMOS cascode LNA, a low phase-noise LC-VCO, and a 6T FinFET SRAM cell. These circuits are evaluated under both closed-form analytical surrogates and an automated, physics-driven MATLAB–ngspice-in-the-loop simulation framework accounting for process–voltage–temperature (PVT) variations and Monte Carlo yield constraints. Non-parametric statistical analyses alongside component-wise ablation studies, reveal a problem-dependent performance profile for KDO. It achieves the best overall rank on the constrained VLSI design problems under both analytical and physics-based evaluations, while remaining competitive against strong mathematical optimizers (DE, GBO) on the standardized CEC benchmarks.