Efficient Task Assignment for Crowd-Powered Rebalancing in Bike Sharing Systems

Bike Sharing Systems (BSSs), serving the cycling trips, have become an important alternative for addressing the last mile problem in city Intelligent Transportation Systems (ITSs). However, it is difficult to achieve a desired quality of service due to the unbalanced distribution of station inventory, which is caused by dynamic usage patterns. Different from conventional redistribution strategies performed by trucks or trailers, we propose a Prediction-Based Task Assignment (PBTA) scheme to outsource the rebalancing tasks to crowds so as to decrease the maintenance and labor cost. The Long Short-Term Memory (LSTM) is first utilized to predict the future fill-levels of the bike stations. Then, the rebalancing problem is reformulated as a maximum weighted bipartite matching problem, which is solved by the Hungarian method in polynomial time. Extensive simulations with real-world bike-sharing datasets are conducted to show the effectiveness of the proposed scheme.