An artificial intelligent approach to the bicycle repositioning problems

Abstract

Purpose Many bike rental organizations permit customers to pick-up bikes from one bike station and return them at a different one. However, this service may result in bike imbalance, as bikes may accumulate in stations with low demand. To overcome the imbalance problem, this paper aims to develop a decision model to minimize the total costs of unmet demand and empty bike transport by determining bike fleet size, deployments and the vehicle routing schedule for bike transports. Design/methodology/approach This paper developed a constrained mixed-integer programming model to deal with this bike imbalance problem. The proposed model belongs to the non-deterministic polynomial-time (NP)-hard problem. This paper developed a two-phase heuristic approach to solve the model. In Phase 1, the approach determines fleet size, deployment level and the number of satisfied demands. In Phase 2, the approach determines the routing schedule for bike transfers. Findings Computational results show the following results that the proposed approach performs better than General Algebraic Modeling System (GAMS) in terms of solution quality, regardless of problem size. The objective values and the fleet size of rental bikes allocated increase as the number of rental stations increases. The cost of transportation is not directly proportional to the number of bike stations. Originality/value The authors provide an integrated model to simultaneously deal with the problems of fleet sizing, empty-resource repositioning and vehicle routing for bike transfer in multiple-station systems, and they also present an algorithm that can be applied to large-scale problems which cannot be solved by the well-known commercial software, GAMS/CPLEX.