Skip to main content
Engineering

How Trip Inferences and Machine Learning Optimize Delivery Times on Uber Eats

June 15, 2018 / Global
Featured image for How Trip Inferences and Machine Learning Optimize Delivery Times on Uber Eats
Figure 1: GPS signals can tell us whether a delivery-partner is driving or parked near a restaurant during an Uber Eats trip.
Figure 2: The Android activity recognition API defines eight types of activity.
Figure 3: This sensor data pipeline shows the data we collect, how it is processed for Uber Eats dispatch, and eventually stored for later analysis.
Figure 4: The sequence model finds change-points amongst a series of noisy input observations.
Figure 5: The Trip State Model adds color to the black-and-white photo of the Uber Eats delivery-partner order pick up flow.
Figure 6: Here is an example distribution inferred from the Trip State Model depicting the average time spent in the waiting, walking, and parking states for a particular restaurant.
Ryan Waliany

Ryan Waliany

Ryan Waliany is a Product Manager on Uber’s Sensing, Inference & Research (SIR) team.

Lei Kang

Lei Kang

Lei Kang is a Data Scientist on the Uber Eats Marketplace Prediction team.

Ernel Murati

Ernel Murati

Ernel Murati is a Mobile Engineer on Uber’s Sensing, Inference & Research (SIR) team.

Mohammad Shafkat Amin

Mohammad Shafkat Amin

Mohammad Shafkat Amin is a Software Engineer on Uber’s Sensing, Inference & Research (SIR) team.

Posted by Ryan Waliany, Lei Kang, Ernel Murati, Mohammad Shafkat Amin, Nikolaus Volk

Category: