Tool: Fleet rebalancing

In the case of modes like bike-sharing and scooter-sharing, fleet size decisions depend on rebalancing decisions. Smart and efficient fleet rebalancing algorithms can reduce not only operating costs, but also capital costs as fewer bicycles are needed to serve demand. We therefore tackle the problem of optimal repositioning decisions for bike-sharing systems (bike rebalancing problem – BRP), accounting for the cost of repositioning and the cost of leaving demand unmet due to lack of bicycles. The redistribution of the bikes takes place via a fleet of capacitated vehicles (trucks). A Mixed-Integer Linear Programming (MILP) algorithm, developed by the Technical University of Athens, is proposed aiming to minimize the total costs for a bike-sharing system, consisting of routing costs and unsatisfied user demand. 

In an effort to improve the computational times required to obtain an optimal solution for larger-scale networks, a heuristic algorithm has been developed. The developed algorithm is capable of deriving high-quality solutions in short computational times for medium and large-scale problem instances. The output of the heuristic was then used to generate a warm-starting technique of providing an initial feasible solution to the solver before starting the optimization process. The model takes as input the geographical distribution of stations and the available fleet of trucks, and therefore, it is useful to analyze the influence of different configurations of the geographical distribution of demand, stations and bikes into the performance of the whole system.