This dissertation introduces the concept of decentralization to address the rebalancing challenges in bike-sharing systems and proposes a model known as the Decentralized Route Assignment Problem (DRAP) under specific assumptions. The primary contributions of this research include the formulation of the DRAP and the derivation of theoretical results that facilitate its transformation into a lower-dimensional global optimization problem. This transformation enables efficient exploration using modern search methods. An extended version of DRAP, called DRAP-EA, is also proposed for further analysis by introducing more agents into the system.

Various solution approaches, such as branch-and-cut, hill-climbing, and simulated annealing, are explored and customized to enhance their performance in the context of rebalancing. Two simulated annealing methods, Gurobi with warm-start, and an extension of the local search algorithm are implemented on 24 instances derived from a comprehensive case study for experimental evaluation. The experimental results consistently demonstrate the superior performance of the simulated annealing methods. Furthermore, a comparison between SA and SA-PS is conducted, and the obtained solutions are visualized to help further explore the spatial patterns and traffic flows within the bike-sharing system.