Variable neighbourhood search for resolving routing and wavelength assignment problem in optical wavelength-division multiplexing networks

Since the issue of finding optimal signal routes imposes itself as one of the main challenges within the framework of optimal networks, this paper aims at illustrating the development of a new, mixed-integer mathematical model for solving the static routing and wavelength assignment (RWA) problem in optical WDM networks and presents the comparison of results obtained by heuristics and CPLEX commercial software. The objective function of the mathematical model is to minimise the blocking probability by maximising the number of successfully connected lightpaths. In obtaining the results of simulations, two WDM networks with full wavelength conversion nodes were tested. The proposed heuristics are the Variable neighbourhood search (VNS) and Tabu search (TS). Firstly, we compare the results obtained by these two heuristics and CPLEX on a lowcomplexity WDM network, aiming to determine whether these heuristics may yield results which approximate optimal results. Secondly, having illustrated that the heuristics provide satisfactory results, we run these heuristics on a high-complexity WDM network. The results indicate minor differences in blocking probability and show that in majority cases, VNS spends ten times less CPU time as opposed to TS. In conclusion, we argue that VNS heuristic is appropriate for resolving this static RWA problem.