A Semi Adaptive Large Neighborhood Search for the Maintenance Scheduling and Routing Problem

This work presents a Semi Adaptive Large Neighborhood Search (SALNS) for the Maintenance Scheduling and Routing Problem. A new removal method based on the behavior of risk in maintenance is proposed. It is combined with several destroy and repair operators. A semi adaptive mechanism that ensures effective mix between diversification and learning is proposed. We conduct a comparative analysis with the solver and with adapted algorithms schemes from the literature: classical ALNS and ALNS with learning automata (LA-ALNS). All the algorithms consider the same choices related to problem's specifications and use the same local search procedure. The proposed metaheuristic chooses the most suitable heuristics while alternating between learning and diversification to obtain high quality solutions.