Repositioning Shared Urban Personal Transport Units: Considerations of Travel Cost and Demand Uncertainty

Abstract

Operators of personal transport units (PTUs) face the challenge of intelligently balancing the locational demand and supply of PTUs in order to mitigate surpluses or deficits at PTU pickup stations. To accomplish this goal, operators need to be able to reliably predict the spatial distribution of PTU demand and to optimize the distributional allocation of resources to meet this demand. This paper proposes a three-step mathematical programming approach that addresses PTU supply vehicle routing and PTU repositioning that minimize the weighted total travel costs and unmet user demand. The methodology combines discrete wavelet transform (DWT) and artificial neural network (ANN) techniques to predict the demand at PTU stations, considers travel cost and unmet user demand in a multiobjective model and solves it with a multiobjective coevolutionary algorithm (MOCA), and incorporates the demand uncertainty to ensure robustness of the optimal repositioning and routing strategy for all the PTU stations. The paper demonstrated the proposed approach using real-world bicycle-sharing data from Nanjing, China, and showed that the proposed approaches for demand prediction (DWT-ANN) and optimization (MOCA) significantly produce superior results compared with traditional methods. Sensitivity analysis demonstrated the robustness of the proposed approaches.