Mitigating the “black holes”: Periodic repair and maintenance problem of shared bikes

Abstract

This article addresses a periodic repair problem for free-floating shared bikes that incorporates uncertain failure rates and covariate information. We conceptualize a physical landscape resembling black holes in cosmology to represent locations with exceptionally high failure rates. To mitigate this “black hole” effect, we introduce two special strategies: dedicated repair periods and preventive maintenance. The effectiveness of these two strategies is first theoretically validated within a two-region system. We develop the operational data analytics (ODA) framework to generate enhanced data-integrated solutions for the periodic repair problem, improving decision quality under limited data. Within this framework, baseline solutions from existing models, including a scenario-wise distributionally robust optimization (DRO) model with an exact linear decision rule, are evaluated and refined to guide the ODA solution. A real-world case study validates the effectiveness of our approach and offers valuable managerial insights. The ODA framework guides the selection of ambiguity sets in DRO models and enhances solution quality, even when the oracle data-integrated solution underperforms. Notably, the two strategies reduce regional disparities in penalty costs, helping to mitigate the black hole effect, as evidenced by the Gini coefficient in a generalized multi-region system.