Shared bicycle liberates travelers’ walking load, promoting their commuting speed, and owns the characteristic of convenience and accessibility. However, at the meantime they create chaos since tons of companies flowing into this market. In hence, it is vital to propose novel management strategies for the shared bicycle. This study contributes to proposing a methodological framework to solve the online operation problem of shared bicycles, considering efficiency and security requirements. We build a centralized control system, establish dynamic optimization model and dynamically make repositioning decisions using approximate dynamic programming. Using a neural network as approximate function, it can avoid myopic decisions in that it considers the impact of current actions on future system states. We design a hypothetical network to capture spatial and temporal heterogeneity of demand and supply, and utilize hypothetical network to verify the effectiveness of the proposed model and algorithm.

CICTP