Dual-neighborhood iterated local search for routing and wavelength assignment

The minimum routing and wavelength assignment (min-RWA) problem is a classic and challenging NP-hard combinatorial optimization problem which aims to reduce the required wavelengths in wavelength-division multiplexing optical networks. In order to tackle this problem, we present a dual-neighborhood iterated local search (DN-ILS) by alternately evaluating a shift-shaking and a swap-shaking neighborhood to improve the current solution. Both neighborhoods are based on an ejection chain-based shaking (ECS) procedure, while the high-level procedures are shift moves and swap moves, respectively. These two move operators change the wavelengths of certain lightpaths, while the ECS refines routing for the related lightpaths. Experimental results on commonly used benchmark instances demonstrate that our proposed DN-ILS algorithm outperforms the best methods in the literature by improving the best known results for 35 out of 113 instances while matching the best known results for the remaining ones. The success of the shift-shaking and swap-shaking neighborhoods inspires us to apply them to other challenging optimization problems with multi-level decisions and strong constraints.