System and method for estimating arrival time

Patent metadata

Field	Value
Publication number	US-10527436-B2
Application number	US-201916372118-A
Country	US
Kind code	B2
Filing date	Apr 1, 2019
Priority date	Nov 23, 2017
Publication date	Jan 7, 2020
Grant date	Jan 7, 2020

Abstract

Systems and methods are provided for estimating an arrival time associated with a trip. The exemplary method receives trip information including an origin and a destination of the trip, and determines a route connecting the origin and the destination. The route includes a plurality of road segments. The method then receives transportation information associated with the road segments of the route and extracts global features and local features from the transportation information. Each global feature is indicative of characteristics involving at least two of the road segments, and each local feature is indicative of characteristics related to an individual one of the road segments. The method applies a machine learning model to estimate the arrival time, which includes a first neural network dedicated to process the global features and a second neural network dedicated to process the local features. The first neural network is distinct from the second neural network.

First claim

What is claimed is:

1. A method for estimating an arrival time associated with a trip, comprising:
receiving, through a network interface, trip information including an origin and a destination of the trip;
determining, by a processor, a route connecting the origin and the destination, the route including a plurality of road segments;
receiving, through the network interface, transportation information associated with the road segments of the route;
extracting, by the processor, global features and local features from the transportation information, each global feature indicative of characteristics involving at least two of the road segments, each local feature indicative of characteristics related to an individual one of the road segments; and
applying, by the processor, a machine learning model to estimate the arrival time, wherein the machine learning model comprises a first neural network dedicated to process the global features and a second neural network dedicated to process the local features, wherein the first neural network is distinct from the second neural network.

2. The method of claim 1 , wherein extracting the global features includes extracting features uniform to the at least two road segments.

3. The method of claim 1 , wherein the global features include a shape encompassing the route.

4. The method of claim 1 , wherein the local features include real-time traffic data associated with the respective road segments.

5. The method of claim 1 , wherein the first neural network is a feedforward neural network and the second neural network is a recurrent neural network.

6. The method of claim 1 , wherein the second neural network comprises a sequence of layers corresponding to a sequence of the road segments, wherein each layer is configured to process local features of a road segment in the corresponding sequence of road segments.

7. The method of claim 1 , wherein the machine learning model further comprises a third neural network dedicated to process the global features, the third neural network is configured to obtain a feature product between two global features.

8. The method of claim 1 , wherein the machine learning model further comprises a multilayer perceptron network configured to process data obtained from the first neural network and the second neural network.

9. The method of claim 1 , wherein the trip information is received from a device used by a user to make a ride order for transportation from the origin to the destination.

10. The method of claim 1 , wherein the first neural network and the second neural network of the machine learning model are collectively trained using historical trips and corresponding historical travel times.

11. A system for estimating an arrival time associated with a trip, comprising:
a network interface configured to receive trip information including an origin and a destination of the trip, and transportation information associated with a route connecting the origin and the destination, the route including a plurality of road segments;
a storage device configured to store a machine learning model; and
a processor configured to:
extract global features and local features from the transportation information, each global feature indicative of characteristics involving at least two of the road segments, each local feature indicative of characteristics related to an individual one of the road segments; and
apply a machine learning model to estimate the arrival time, wherein the machine learning model comprises a first neural network dedicated to process the global features and a second neural network dedicated to process the local features, wherein the first neural network is distinct from the second neural network.

12. The system of claim 11 , wherein to extract the global features, the processor is further configured to extract features uniform to the at least two road segments.

13. The system of claim 11 , wherein the local features include real-time traffic data associated with the respective road segments.

14. The system of claim 11 , wherein the first neural network is a feedforward neural network and the second neural network is a recurrent neural network.

15. The system of claim 11 , wherein the second neural network comprises a sequence of layers corresponding to a sequence of the road segments, wherein each layer is configured to process local features of a road segment in the corresponding sequence of road segments.

16. The system of claim 1 , wherein the machine learning model further comprises a multilayer perceptron network configured to process data obtained from the first neural network and the second neural network.

17. The system of claim 1 , wherein the network interface is further configured to receive the trip information from a device used by a user to make a ride order for transportation from the origin to the destination.

18. The system of claim 11 , wherein the processor is further configured to:
determine a plurality of routes connecting the origin and the destination;
estimating an arrival time for each route by applying the machine learning model; and
make a recommendation of a route among the plurality of routes associated with the shortest arrival time.

19. The system of claim 11 , wherein the first neural network and the second neural network of the machine learning model are collectively trained using historical trips and corresponding historical travel times.

20. A non-transitory computer-readable storage medium storing instructions that, when executed by a processor, cause the processor to perform a method for estimating an arrival time associated with a trip, the method comprising:
receiving trip information including an origin and a destination of the trip;
determining a route connecting the origin and the destination, the route including a plurality of road segments;
receiving transportation information associated with the road segments of the route;
extracting global features and local features from the transportation information, each global feature indicative of characteristics involving at least two of the road segments, each local feature indicative of characteristics related to an individual one the road segments; and
applying a machine learning model to estimate the arrival time, wherein the machine learning model comprises a first neural network dedicated to process the global features and a second neural network dedicated to process the local features, wherein the first neural network is distinct from the second neural network.

Assignees

• Beijing Didi Infinity Technology & Dev Co Ltd

Inventors

• Fu Kun
• Wang Zheng

Classifications

• G06N3/045

  Combinations of networks · CPC title

• G06N3/044

  Recurrent networks, e.g. Hopfield networks · CPC title

• G01C21/3484 Primary

  Personalized, e.g. from learned user behaviour or user-defined profiles · CPC title

• G06N3/08

  Learning methods · CPC title

• G06N3/04

  Architecture, e.g. interconnection topology · CPC title