EURO 2024 Copenhagen
Abstract Submission

EURO-Online login

2495. Transit Estimation Models for Transportation Planning

Invited abstract in session WC-57: Forecasting for the middle mile, stream Optimization at Amazon.

Wednesday, 12:30-14:00
Room: S06 (building: 101)

Authors (first author is the speaker)

1. Carlos Sanchez Sanchez
Amazon
2. Mounir Boujrad
3. Tim Forbes
4. Olle Green
Machine Learning & Engineering, Amazon
5. Madhavan Sriram
Amazon Transportation Services, Amazon
6. Chris George
ATS, Amazon

Abstract

Decision making in Transportation, including network design, reactive adjustments during execution, or post-run analysis, requires an understanding of the vehicle transit time, carbon impact, and risks. In this talk, we explore a unified approach for the estimation of such factors that leverages the information available from vehicle sensors and other geospatial data at each time horizon. We propose a model that combines routing and predictive algorithms to provide transit estimates for both connections between Amazon locations as well as unseen routes between third-party vendor connections. We describe how we deal with noisy telematic signals and sparse ground truth, the machine learning approaches to work with road-level data, and the mechanisms to represent the variability inherent to a prediction that depends on the road conditions and driver behaviour.

Keywords

Status: accepted


Back to the list of papers