Data-driven optimization for rebalancing shared electric scooters

Abstract

Shared electric scooters have become a popular and flexible transportation mode in recent years. However, managing these systems, especially the rebalancing of scooters, poses significant challenges due to the unpredictable nature of user demand. To tackle this issue, we developed a stochastic optimization model (M0) aimed at minimizing transportation costs and penalties associated with unmet demand. To solve this model, we initially introduced a mean-value optimization model (M1), which uses average historical values for user demand. Subsequently, to capture the variability and uncertainty more accurately, we proposed a data-driven optimization model (M2) that uses the empirical distribution of historical data. Through computational experiments, we assessed both models’ performance. The results consistently showed that M2 outperformed M1, effectively managing stochastic demand across various scenarios. Additionally, sensitivity analyses confirmed the adaptability of M2. Our findings offer practical insights for improving the efficiency of shared electric scooter systems under uncertain demand conditions.