Project Title: Optimization of ATM Replenishment
Student: Maksim Shutov
Course: MSc in Computer Science

Abstract:
This work is aimed at modelling and implementing the real-world problem of ATM replen
ishment optimization. It consists in finding an optimal replenishment schedule, routes for
vehicles and ATM inventory policy to minimize operational costs related to the replenish
ment procedure. Objects of minimization are both transportation and inventory holding
costs which incurs when cash is kept in ATMs.
The problem is tackled as a problem of the Inventory Routing Problem class with speci?c
features of a business case described above. We consider optimization on a planning
horizon of a certain length. The problem at hand features multiple vehicles, time windows,
inventory management. The model does not consider cash pick-ups from ATMs, deliveries
only.
This study provides an algebraic formulation of the problem. The model was implemented
in Python using Gurobi solver API. Model testing has been conducted using a real-world
ATM data which is made available for using in further research.
Experiments with implemented model has shown that the model requires certain simplifications 
to be able to solve instances of moderate size. With these simplifications the model is capable 
of solving instances of moderate sizes. For example, an instance with 50 ATMs can be solved 
within three hours time limit. The influence of several model aspects on the solving time and 
complexity of the problem has been also studies through experimentation. Low initial and 
final inventory levels, high rate of cash withdrawals and high distances between ATMs increase 
solving times of problem instances.