Answers are grounded in this paper and cite their sources.
Digital-twin (DT) systems have emerged as critical enablers of intelligent cyber–physical interaction, yet their effectiveness is limited by static modelling and non-adaptive optimisation mechanisms. This article presents a co-evolutionary framework that simultaneously evolves the DT model and its optimisation strategy, achieving self-correcting performance in dynamic environments. The framework integrates a Multi-Task Genetic Algorithm (MTGA) that unifies three concurrent evolutionary processes model calibration, decision optimisation, and uncertainty-guided exploration within a closed feedback loop. A formal theoretical foundation is developed to describe the coupled fitness landscapes and feedback–response operators that govern reciprocal adaptation between model and optimiser. The architecture incorporates an uncertainty-aware selection mechanism, where predictive variance from the surrogate model dynamically influences evolutionary pressure, thereby improving both model fidelity and solution robustness. Experimental validation on a Grid ALU Processor configuration problem demonstrates that the proposed MTGA achieves up to 60% reduction in model prediction error, 22% improvement in Pareto-front hypervolume, and enhanced stability under nonstationary conditions compared with NSGA-II, SPEA2, and CNSGA-II. These findings confirm that co-evolutionary coupling yields superior convergence and sustained reliability, positioning this framework as a foundation for next-generation self-learning, adaptive DTs in complex engineering domains.
No takes yet. Share an insight, caveat, or question.
Jalali-Varnamkhasti et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: