A Dual-Population Optimization Algorithm Based on a Collaborative Sloth–Vine Mechanism

This paper presents the Sloth–Vine Optimization Algorithm (SVOA), a dual-population method combining elite-guided differential search, vine exploration, opposition trials, and conditional role exchange. The SVOA implementation is evaluated on 12 CEC2022 functions at dimensions 10 and 20, using 30 prescribed seeds and counted function–evaluation budgets. Its average mean fitness ranks are 1.333 and 2.083, respectively; at dimension 20, the overall rank is tied with ABC. Its strongest consistent relative performance occurs on the three hybrid functions, while some competing methods perform better elsewhere. Engineering experiments distinguish a path-planning application variant from the unchanged core used for heterogeneous unmanned surface vehicle (USV) allocation. On one fixed obstacle map, the path variant obtains 30/30 feasible simplified paths at 3000 evaluations per run, with a mean pre-simplification fitness value of 225.31. For 43 tasks and 11 USVs, the unchanged core obtains feasible allocations in 5/30 runs at 250,500 evaluations, with a best feasible makespan of 11.5 h. The other 25 failures remain in the statistics. Ten-seed paired penalty tests preserve path feasibility on this map but show limited USV success across all settings. These results support problem-dependent performance, without establishing universal superiority or reliable feasible USV allocation.