Optimization Model and Algorithm for Dockless Bike-Sharing Systems Considering Unusable Bikes in China

Ying Teng,Xiaohui Li,Xupeng Liang,Hua Zhang

doi:10.1109/access.2020.2967398

Abstract

The damaged dockless shared-bikes are increasingly common in large cities of China, which not only reduce users' satisfactions but also harm the public environment. It is a merging issue for dockless bike-sharing systems to timely recall unusable bikes and replenish new/usable bikes. This paper addresses a maintenance network for a dockless bike-sharing system within a specific region in which the random demands and returns of shared-bikes are considered and the returned bikes have a certain probability to be unusable. The problem is to jointly determine the numbers of unusable and usable bikes to be recalled and replenished, respectively, and the corresponding routes of maintenance vehicles via the maintenance network. The problem is formulated as a stochastic mixed-integer programming model with the objective to maximize the expected revenue of the maintenance network within a operational period. To resolve the such complicated integration-optimization problem efficiently, the formulated model is decomposed into a multi-location newsvendor subproblem and a multiple vehicle routing subproblem with simultaneous delivery and pick-up. An efficient two-stage heuristics is then proposed, which resolves the two subproblems in the two stages of the algorithm, respectively. The computational results based on an real numerical case and a large number of numerical instances validate the performance of the proposed model and algorithm.

Highlights

Bike-sharing systems provide a green and efficient alternative for travelers’ first- and last-mile connections to other transit modes
This paper addresses a maintenance network for dockless bike-sharing systems considering the unusable bikes
To capture the characteristics of the random demands and returns of shared bikes, we formulate the problem as a stochastic mixed-integer programming (MIP) model with the combination of a multilocation newsvendor subproblem with a capacity constraints [33], [34] and a multiple vehicle routing problem with simultaneous delivery and pick-up (MVRPSDP) subproblem [26]

Summary

INTRODUCTION

Bike-sharing systems provide a green and efficient alternative for travelers’ first- and last-mile connections to other transit modes. In the problem addressed in this paper, the locations where users intensively use and return bikes are usually variable from morning to afternoon of each day It implies that the maintenance network needs a fast and efficient clustering method to detect the spatial locations of hotspots for each period throughout an infinite operation horizon. To capture the characteristics of the random demands and returns of shared bikes, we formulate the problem as a stochastic MIP model with the combination of a multilocation newsvendor subproblem with a capacity constraints [33], [34] and a multiple vehicle routing problem with simultaneous delivery and pick-up (MVRPSDP) subproblem [26]. The second one is a MVRPSDP subproblem to minimize the total travel cost of vehicles with the determined Xt and Yt. this section proposes an efficient two-stage heuristics to generate a near-optimal solution to the problem P. We resolve the MVRPSDP subproblem and obtain the optimal routes Zt for vehicles to deliver and pick up bikes at hotspots simultaneously

SUBALGORITHM IN THE FIRST STAGE

NUMERICAL EXPERIMENTS

CONCLUSION