Authors:
Saman M. Almufti, Renas Rajab Asaad, Ajibise Abdul-Lateef Ajibola, Ereeny Aziz, Abdulrahman Sherif, Abdelrahman Hassan

Addresses:
Department of Information Technology, Technical College of Informatics-Akre, Akre University for Applied Sciences, Akre, Nineveh Governorate, Iraq. Department of Electrical and Electronics Engineering, Olabisi Onabanjo University, Ago-Iwoye, Ogun State, Nigeria. Department of Computer Science, Zewail City of Science and Technology, 6th of October City, Giza Governorate, Egypt.

Abstract:

This paper proposes a continuous multi-objective U-Turning Ant Colony Optimization (MO-U-TACO) method and evaluates its parameter configuration on 21 standard benchmarks from the ZDT, DTLZ, and WFG families. The original U-TACO concept assigns a preparatory role to partial U-ant routes before complete ants construct final tours. The present formulation transfers this distinction to continuous vectors: a U-ant samples a subset of coordinates, the remaining coordinates are completed, the full objective vector is evaluated, and pheromone credit is returned only to the sampled coordinates. A bounded external non-dominated archive, crowding-distance truncation, archive-guided interval sampling, and sparse local refinement replace the single incumbent used in single-objective applications. Researchers evaluated eight-parameter scenarios with 50 matched seeds per benchmark, producing 8,400 independent problem runs. Researchers retained the first 10 and 30 seeds as nested samples for run-count analysis. Researchers assessed performance using IGD+, normalized hypervolume, additive epsilon, archive size, function evaluations, and validation runtime. Under the 50-run analysis, S7 obtained the best mean composite rank (1.381) and the best rank on 12 of 21 problems. S8 ranked second (2.825) and produced the smallest arithmetic mean IGD+ (0.473454), but S7 achieved a larger mean hypervolume ratio (0.579963 versus 0.527794). The Friedman test detected a scenario effect (χ_F^2=284.226667, p=1.408e-57, Kendall’s W=0.812076). Holm-adjusted paired tests favoured S7 over every alternative, including S8 (p_Holm=2.931e-14). The ordering was stable across 10, 30, and 50 runs. These results identify the large-colony configuration as the most consistent setting under the adopted low-budget protocol; they do not compare it with established multi-objective algorithms.

Keywords: Multi-Objective Optimization; U-Turning Ant Colony Optimization; Pareto Archive; Hypervolume and Parameter Sensitivity; Run-Count Analysis; Validation Runtime.

Received on: 20/10/2025, Revised on: 13/12/2025, Accepted on: 04/02/2026, Published on: 03/06/2026

DOI: 10.69888/FTSCIS.2026.000766

FMDB Transactions on Sustainable Critical Infrastructures, 2026 Vol. 1 No. 2, Pages: 108-125

• Views : 36
• Downloads : 5

Download PDF