A multiple-level variable neighborhood search approach to the orienteering problem

A multiple-level variable neighborhood search (ML-VNS) approach is developed for the orienteering problem (OP) which maximizes the reward collected from visited sites while satisfying predetermined constraints. The ML-VNS approach remedies situations in which the information accumulated during the search process of an individual instance is not shared with the search processes in other instances with different constraint levels. New large-sized OPs have been defined. Results of the ML-VNS approach show promise when compared with previous tabu search (TS) algorithm and probabilistic solution discovery algorithm (PSDA).