Multi-year network level road maintenance programming by genetic algorithms and variable neighbourhood search

Preservation of the value, accessibility and adequate service level of road assets is one of the main tasks of road administrations. Right timing of maintenance works both decrease the road user costs as well as maintenance costs maximising overall benefits to the society measured by Net Present Value (NPV). We present a problem of road maintenance programming as a large-scale optimisation problem, which we optimise with genetic algorithms (GA), a parallel version of GA, and a variable neighbourhood search as a post-processing step on the previous solutions. We also compared all the optimised solutions with a large number of random solutions. A case study in the Sindh Province of Pakistan shows that parallel genetic algorithms with variable neighbourhood search produce 50 percent better results compared to random sampling and 6 percent better compared to regular genetic algorithms. The case study shows that current priorities in recent years are service level upgrading and routine maintenance.