Battery swapping, vehicle rebalancing, and staff routing for electric scooter sharing systems

Electric scooter (e -scooter) sharing systems provide on -demand electric scooter rental services. E -scooters are equipped with swappable batteries managed by staff who visit each scooter and replace depleted batteries with charged ones. The e -scooters in this system are freefloating; they can be located anywhere without having to be returned to designated stations. Due to this characteristic of the scattered location of e -scooters, operation decisions with associated staff routing are more challenging than station -based services. Thus, it is critical to implement the efficient management of e -scooter redistribution and charging decisions with proper staff routing to successfully prepare for user demand within a limited operation time while minimizing the total operating cost. To this end, we introduce a battery swapping and vehicle rebalancing problem with staff routing for e -scooter sharing systems, formulated as a mixed integer programming (MIP) model. To derive solutions efficiently for a large-scale instance in practice, we propose a clustered iterative construction approach where the problem is decomposed into two phases. The first phase clusters regions using an approximation of intraregion and inter -region operation costs with the minimum spanning tree approach. The second phase efficiently derives multiple candidate regional sequences by our partial permutation procedure and following the sequences, iteratively solves a significantly reduced size of the problem to construct operation assignments. Our numerical experiments on the generated instances and real -world instances demonstrate that the proposed two-phase algorithm shows significantly superior performance in practically large scale instances than the benchmarks without clustering or the iterative procedures of our algorithm.