Generation of trip estimates using real-time data and historical data

Inventors: Shijing Yao (El Cerrito, CA); Xiao Cai (San Bruno, CA)

Assignee: Uber Technologies, Inc.

G01C21/3484 G01C21/3492 G01C21/3617 G08G1/096811 G08G1/096888

View Patent ↗

Loading inventors, assignments & file history…

Monitor This Case

Get email alerts when status or documents change.

Order Certified Copies

Most orders are placed with the USPTO same day — all within 24 business hours.

Order via The Patent Place →

Pre-filled with this patent's details

Quick Facts

Patent No.: US 10,557,713
App. No.: 15/678,253
Granted: Feb 11, 2020
Kind: B1

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

A system uses machine models to estimate trip durations or distance. The system trains a historical model to estimate trip duration using characteristics of past trips. The system trains a real-time model to estimate trip duration using characteristics of recently completed trips. The historical and real-time models may use different time windows of training data to predict estimates, and may be trained to predict an adjustment to an initial trip estimate. A selector model is trained to predict whether the historical model, the real-time model, or a combination of the historical and real-time models will more accurately estimate a trip duration, given features associated with a trip duration request, and the system accordingly uses the models to estimate a trip duration. In some embodiments, the real-time model and the selector may be trained using batch machine learning techniques which allow the models to incorporate new trip data as trips complete.

Claims (62)

1. A computer-implemented method, comprising:

for each sequential time period, iteratively training a real-time model for estimating a trip duration of a trip within a next time period based on characteristics of the trip and training a selector model for predicting an estimated trip duration of a trip by:

identifying a first training window occurring within a current time period;

identifying a first set of completed trips associated with the first training window, each trip having a known duration and characteristics;

training the real-time model to predict trip duration based on the first set of completed trips;

identifying a second training window occurring after the first training window and within the current time period;

identifying a second set of completed trips associated with the second training window, each trip having a known duration and characteristics;

generating a set of estimated trip durations of the trips in the second set of completed trips using a historical model;

generating a set of estimated trip durations of the trips in the second set of completed trips using the real-time model;

training the selector model to predict an estimated trip duration based on the set of estimated trip durations generated by the real-time model and the set of estimated trip durations generated by the historical model;

receiving a request to estimate a duration of a new trip during the next time period;

generating a first estimated trip duration of the new trip by applying characteristics of the new trip to the real-time model;

generating a second estimated trip duration of the new trip by applying characteristics of the new trip to the historical model; and

determining, by the selector model, the new trip duration based on the first estimated trip duration and the second estimated trip duration.

2. The method of claim 1 , wherein identifying a set of completed trips associated with a training window comprises:

monitoring a set of ongoing trips; and

adding a trip from the set of ongoing trips to the set of completed trips when the trip completes.

3. The method of claim 1 , wherein characteristics of the new trip include an estimated trip duration for the trip predicted by a network-based estimation algorithm.

4. The method of claim 1 , wherein the real-time model is trained to estimate duration of a specific category of trip, the categorization being based on trip length.

5. The method of claim 1 , wherein, in the next time period, the new trip duration is determined by a first processing thread applying the real-time model and the selector model while a second processing thread iteratively trains a second real-time model and a second selector model within the next time period.

6. A non-transitory computer-readable storage medium storing computer program instructions executable by one or more processors of a system to perform steps comprising: