On the effect of using sensors and dynamic forecasts in inventory-routing problems

Maximiliano Cubillos
Aarhus university, Denmark
DEIB - Alpha Room (Building 24)
June 29th, 2022
10.30 am
Contacts:
Ola Jabali
Research Line:
Operations research and discrete optimization
Aarhus university, Denmark
DEIB - Alpha Room (Building 24)
June 29th, 2022
10.30 am
Contacts:
Ola Jabali
Research Line:
Operations research and discrete optimization
Sommario
On June 29th, 2022 at 10.30 am, Dr. Maximiliano Cubillos, Aarhus university, Denmark, will give a seminar on "On the effect of using sensors and dynamic forecasts in inventory-routing problems" in DEIB Alpha Room.
In this paper, we study an inventory-routing problem with stochastic demand, in which knowledge of the demands of customers can be updated by the use of sensor information, and used to plan delivery decisions in a given planning period. We consider the case in which a limited number of sensors can be placed, and investigate what simple rules can best be applied to decide on their allocation.
To evaluate these simple sensor allocation rules, we propose a Variable Neighborhood Search algorithm for an inventory-routing problem in a rolling horizon framework to solve the problem which uses both sensor and historical data to update demand forecasts.
We perform extensive computational experiments in which we generate random instances and consider different demand generation scenarios to test different sensor allocation rules. Results show that simple allocation rules, such as placing sensors at customers with high demand or far from the depot, can significantly reduce the total cost, particularly if combined with dynamic forecast information.
In this paper, we study an inventory-routing problem with stochastic demand, in which knowledge of the demands of customers can be updated by the use of sensor information, and used to plan delivery decisions in a given planning period. We consider the case in which a limited number of sensors can be placed, and investigate what simple rules can best be applied to decide on their allocation.
To evaluate these simple sensor allocation rules, we propose a Variable Neighborhood Search algorithm for an inventory-routing problem in a rolling horizon framework to solve the problem which uses both sensor and historical data to update demand forecasts.
We perform extensive computational experiments in which we generate random instances and consider different demand generation scenarios to test different sensor allocation rules. Results show that simple allocation rules, such as placing sensors at customers with high demand or far from the depot, can significantly reduce the total cost, particularly if combined with dynamic forecast information.