A hybrid variable neighborhood search algorithm based on grouping strategies for the static bike sharing re-positioning problem

Abstract

This paper considers the static bike sharing repositioning operation that is essential to eliminate inventory imbalance among bike sharing stations caused by stochastic bike renting and returning in modern bike sharing systems. The operation aims to remove excess bike inventories from surplus stations and add needed bikes to insufficient stations in order to minimize both traveling cost and inventory cost. A hybrid variable neighborhood search (VNS) algorithm based on two grouping strategies is proposed to solve the static bike sharing re-positioning problem (s-BSRP). The two grouping strategies, namely, geolocation-based grouping and supply-demand-based grouping, are designed to construct station groups. Vehicle routes within each group and among groups are improved using a variable neighborhood search algorithm with local search, for which several neighborhood structures are designed. Extensive computational experiments are conducted on benchmark instances with various sizes taken from the literature. Performance of the proposed algorithm is compared with that of branch-and-cut and two other state-of-the-art algorithms. Computational results show the superior performance of the proposed grouping strategies and the hybrid VNS algorithm in solving large-scale BSRPs.