Many-objective optimization problems (MaOPs) face the difficulties of convergence decay and diversity maintenance caused by the dimension catastrophe. The traditional two-archive strategy makes it difficult to balance the performance of the solution set due to the local convergence preference of the \(I_{\varepsilon +}\) index. This paper proposes the Two-Arch2-DAC algorithm and a two-stage collaborative optimization architecture: (1) The DAV-dominated stage uses a dynamic angle vector-based dominance relation to reconstruct convergence archives to retain the dominant solutions in the global objective space; (2) the \(I_{\varepsilon +}\) assisted stage refines the solutions within the dominance layer to enhance convergence capability. The innovation lies in the synergistic mechanism between DAV and \(I_{\varepsilon +}\) , which effectively balances the convergence and diversity of the population when solving MaOPs using the dual-archiving algorithm. Across 16 benchmark functions in the DTLZ and WFG series, the Two-Arch2-DAC algorithm demonstrates significant advantages over six mainstream many-objective evolutionary algorithms in terms of population convergence and diversity, regardless of the number of objectives. Project page: https://github.com/snow2121/Two-Arch2-DAC.
