EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

519. Learning optimal courier assignments in on-demand delivery platforms

Invited abstract in session TB-3: Machine Learning in Applied Optimization, stream Data Science Meets Optimization.

Tuesday, 10:30-12:00
Room: 1005 (building: 202)

Authors (first author is the speaker)

1. Gökhan Ceyhan
Operations Research, Just Eat Takeaway.com
2. Pol Arias

Abstract

On-demand delivery platforms are characterised by uncertain demand with spatiotemporal patterns and short delivery time targets. Platform operators manage a courier fleet aiming to deliver the orders as quickly as possible and in a cost-efficient way. A core component of the platform is the courier assignment algorithm assigning incoming orders to couriers in real-time and determining a sequence of pickups and deliveries for each courier. Column generation is commonly used to solve these types of problems in practice: First, a set of feasible columns is generated via sequencing algorithms. Then, a set-partitioning problem is solved with the generated set of columns. As the size of the problem increases with the number of columns included in the problem, it is essential to include only the columns that are likely to be optimal in the problem. Our study aims to learn the optimal columns via machine learning utilising the characteristics of the problem instance and the feasible decision space. We then improve the efficiency and the solution quality of the algorithm by favouring the columns that are more likely to be part of the optimal solution. We conduct computational studies on a set of historical instances utilising the optimality predictions generated by the trained machine learning model and evaluate in terms of solution quality and run times.

Keywords

Status: accepted

Back to the list of papers