UDC 004.021
ADAPTIVE SELF-TUNING OF GLOBAL PARAMETERS IN TF-QIMOA MULTIOBJECTIVE REAL-VALUED OPTIMIZATION ALGORITHM BASED ON SUCCESS HISTORY OF NONDOMINATED SOLUTIONS
L. A. Demidova, Dr. in technical sciences, full professor, professor of the Department of ERP Systems, RTU MIREA, Moscow, Russia;
orcid.org/0000-0003-4516-3746, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
V. V. Maslennikov, Senior lecturer at the Department of ERP Systems, RTU MIREA, Moscow, Russia; orcid.org/0000-0003-3201-2228, e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
The study presents an adaptive version of quantum-inspired multiobjective real-valued optimization al gorithm TF-QIMOA, enhanced with a self-tuning mechanism for global parameters based on success history of nondominated solutions. In contrast to the original TF-QIMOA algorithm which employs fixed global pa rameter values, a new algorithm, TF-QIMOA-SHA, dynamically adjusts its key control parameters by quan titatively assessing the contribution of each nondominated solution to the evolution of the Pareto front. This assessment integrates three components: hypervolume improvement, temporal stability of solutions, and the degree of exploration of sparsely populated segments of objective space. This approach enables real-time adaptation of search process without any prior assumptions about the structure of objective functions. Ex perimental validation confirms the superiority of TF-QIMOA-SHA over state-of-the-art multiobjective opti mization algorithms, namely TF-QIMOA, QI-NSGA-III, MOWOATS, and MOEA/D-DE-SHA, in terms of solution distribution uniformity along the Pareto front, convergence speed, and the proportion of nondomi nated solutions relative to previous generation. These results demonstrate high potential of the proposed algorithm for solving complex engineering problems characterized by a high degree of conflict among objec tive functions.
Key words: : multiobjective optimization, quantum-inspired algorithm, thermonuclear fusion, adaptive self-tuning, success history, hypervolume, Pareto front, solution evaluation.