Joint probabilistic modeling and inference of intersection structure

• US 10,013,508 B2
• Filed: 10/07/2014
• Issued: 07/03/2018
• Est. Priority Date: 10/07/2014
• Status: Active Grant

• Current Assignee: Toyota Jidosha Kabushiki Kaisha (Toyota Motor Corporation)
• Original Assignee: Toyota Motor Engineering & Manufacturing North America Incorporated (Toyota Motor Corporation)
• Inventors: Joshi, Avdhut S., James, Michael R.
• Primary Examiner(s): Roni, Syed
• Application Number:
  US14/508,653
• Publication Number:
  US 20160098496A1
• Time in Patent Office: 1,365 Days
• Field of Search: None
• CPC Class Codes: E01C 1/02 Crossings, junctions or interconnections between roads on the same level construction of traffic islands E01F1/00
• CPC Class Codes: G01C 21/3815 Road data
• CPC Class Codes: G01C 21/3867 Geometry of map features, e.g. shape points, polygons or for simplified maps
• CPC Class Codes: G06F 30/13 Architectural design, e.g. computer-aided architectural design [CAAD] related to design of buildings, bridges, landscapes, production plants or roads

Abstract

A method of modeling an intersection structure of a roadway. The method includes receiving a first data set including road lane information, and receiving a second data set including vehicle trajectory information for an intersection structure of a roadway. The method includes determining lane node locations from at least one of the first and second data sets. A set of potential links between the lane node locations may be compiled. The method may further include assessing, for each link, a probability that the link is a valid link, and assigning each link with a probability value. The links may be filtered based on a predetermined threshold probability value and a set of valid links is generated. A model of the intersection structure is created based on the set of valid links.

Claims

1. A method of modeling an intersection structure of a roadway to provide a detailed intersection map for a fully automated driving system, the method comprising:

receiving, at a computing device, a first data set including road lane information;

receiving, at the computing device, a second data set including vehicle trajectory information for an intersection structure of a roadway;

determining, using the computing device, lane node locations from at least one of the first and second data sets;

compiling, using the computing device, a set of potential links between the lane node locations;

assessing, for each link, a probability that the link is a valid link by using a tracking algorithm to obtain a plurality of complete tracks from the second data set, wherein each complete track represents a single vehicle tracked through the intersection;

correlating each complete track with a link;

assigning each link with a probability value, with each link that is correlated with a complete track being assigned a probability value of 1;

filtering the links, using the computing device, based on a predetermined threshold probability value and generating a set of valid links;

creating, using the computing device, a model of the intersection structure based on the set of valid links;

providing a detailed intersection map with valid connections between lanes, based on the model of the intersection, for a fully automated driving system; and

utilizing the detailed intersection map by the fully automated driving system.

2. The method according to claim 1, wherein each complete track represents a single vehicle tracked passing two node locations.

3. The method according to claim 1, further comprising:

using the tracking algorithm to obtain a plurality of partial tracks from the second data set, each partial track representing a single vehicle tracked through at least a portion of the intersection; and

correlating each partial track with a link.

4. The method according to claim 3, wherein the partial track passes through at least one node location.

5. The method according to claim 3, wherein the step of assigning the probability value to a link that is correlated with a partial track comprises using a Bayesian Model Averaging technique to calculate a probability value that is less than 1.

6. The method according to claim 5, wherein the Bayesian Model Averaging technique is governed by an equation:

p(λ

_q|T)=Σ

<sub>m∈

M</sub>p(λ

_q|m)p(m|T)

wherein p(λ

_q|T) is the probability value of the link (λ

_q), given a set of partial tracks (T), where m is an intersection model and a set of intersection models (M) is parsimonious with data used with the equation.

7. The method according to claim 1, wherein the road lane information is received from a first source, and the vehicle trajectory information is received from a second source.

8. The method according to claim 7, wherein the first source comprises a preexisting map database.

9. The method according to claim 7, wherein the first source comprises data obtained from a lane estimation algorithm using a combination of light detection and ranging (LIDAR) data and road map data.

10. The method according to claim 7, wherein the second source comprises at least one of LIDAR data, radar data, still image data, video data, tabular data, and combinations thereof.

11. The method according to claim 1, wherein the predetermined threshold probability value is assigned a value specific to the intersection, and filtering the links comprises excluding all links having a probability value of less than the predetermined threshold probability value from the set of valid links.

12. A method of modeling an intersection structure of a roadway to provide a detailed intersection map for a fully automated driving system, the method comprising:

identifying, using a computing device, a set of links between lane node locations for an intersection structure of a roadway;

using vehicle trajectory information from the intersection structure to generate, using the computing device, a set of vehicle tracks that passed through the intersection;

categorizing the vehicle tracks into complete tracks and partial tracks;

assessing, using the computing device, for each partial track, a probability that the partial track correlates to a link, wherein the probability is determined using a Bayesian Model Averaging technique;

generating a set of valid links by;

assigning each complete vehicle track a probability value of 1;

assigning each partial vehicle track a probability value that is less than 1; and

filtering the links based on a predetermined threshold probability value;

creating, using the computing device, a model of the intersection structure based on the set of valid links;

providing a detailed intersection map with valid connections between lanes, based on the model of the intersection, for a fully automated driving system; and

utilizing the detailed intersection map by the fully automated driving system.

13. The method according to claim 12, wherein the set of valid links excludes all links having an assigned probability value of less than the predetermined threshold probability value.

14. The method according to claim 12, wherein the Bayesian Model Averaging technique is governed by an equation:

p(λ

_q|T)=Σ

<sub>m∈

M</sub>p(λ

_q|m)p(m|T)

wherein p(λ

_q|T) is the probability value of the link (λ

_q), given a set of partial tracks (T), where m is an intersection model and a set of intersection models (M) is parsimonious with data used with the equation.