COOPERATIVE LEARNING METHOD FOR ROAD INFRASTRUCTURE DETECTION AND CHARACTERIZATION

• US 20150300828A1
• Filed: 04/17/2014
• Published: 10/22/2015
• Est. Priority Date: 04/17/2014
• Status: Abandoned Application

• Current Assignee: Ford Global Technologies, LLC (Ford Motor Company)
• Original Assignee: Ford Global Technologies, LLC (Ford Motor Company)
• Inventors: TSENG, Fling, YU, Hai
• Application Number:
  US14/255,484
• Publication Number:
  US 20150300828A1
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G01C 21/3815 Road data
• CPC Class Codes: G01C 21/3841 Data obtained from two or more sources, e.g. probe vehicles
• CPC Class Codes: G01C 21/3848 Data obtained from both position sensors and additional sensors

Abstract

A method for generating street map data includes collecting acceleration, turning, and geolocation data. The data is collected from acceleration sensors, turning sensors, and geolocation systems in at a least one vehicle across a plurality of vehicle drive cycles. The method additionally includes aggregating the acceleration, turning, and geolocation data. The method further includes predicting the presence of a traffic control device in response to an identified repetitive pattern in the aggregated data. The method further includes updating street map data to include the predicted traffic control device.

Claims

1. A method of generating street map data comprising:

collecting acceleration, turning, and geolocation data from acceleration sensors, turning sensors, and geolocation systems in at least one vehicle across a plurality of vehicle drive cycles;

predicting a presence of a traffic control device at a geolocation in response to an identified repetitive pattern in the data; and

updating street map data to include the traffic control device at the geolocation.

2. The method of claim 1, wherein the identified repetitive pattern comprises a plurality of vehicle stops at the geolocation across a plurality of drive cycles.

3. The method of claim 2, further comprising defining a first time interval corresponding to a vehicle stop at a stop sign, a second time interval corresponding to a vehicle stop at a crosswalk, a third time interval corresponding to a yield sign, and a fourth time interval corresponding to a traffic light, and wherein predicting a presence of a traffic control device at a geolocation comprises correlating the plurality of vehicle stops with one of the first, second, third, and fourth time intervals.

4. The method of claim 2, further comprising defining a first stop probability corresponding to a vehicle stop at a stop sign, a second stop probability corresponding to a vehicle stop at a crosswalk, a third stop probability corresponding to a yield sign, and a fourth stop probability corresponding to a traffic light, and wherein predicting a presence of a traffic control device at a geolocation comprises correlating a percentage of vehicle trips through the geolocation that are vehicle stops at the geolocation with one of the first, second, third, and fourth stop probability.

5. The method of claim 1, wherein predicting a presence of a traffic control device at a geolocation in response to an identified repetitive pattern in the data comprises predicting the presence of a stoplight at an intersection in response to an identified geolocation at which a pattern of first and second driving modes occurs, the first mode including driving toward and through the intersection at a first heading without stopping and the second mode including driving toward the intersection at the first heading and stopping before driving through.

6. A mapping system comprising:

one or more computing devices configured to

aggregate collected data including driver actuations of vehicle controls at corresponding geolocations;

infer a presence of a discrepancy in mapping data in response to an identified repetitive driving pattern at one of the geolocations among the aggregated data; and

update the mapping data to correct the discrepancy.

7. The system of claim 6, wherein the driver actuations of vehicle controls include accelerator pedal actuation, brake pedal actuation, or steering wheel rotation.

8. The system of claim 6, wherein inferring a presence of a discrepancy in mapping data includes predicting a presence of a traffic control device in response to a plurality of vehicle stops at the one of the geolocations at which no traffic control device is indicated in the mapping data.

9. The system of claim 6, wherein inferring a presence of a discrepancy in mapping data includes predicting a presence of a road in response to a plurality of vehicle turns from a first vehicle heading to a second vehicle heading at the one of the geolocations at which no road is indicated in a direction of the second vehicle heading.

10. A method of generating mapping data comprising:

generating mapping data including a predicted geolocation of a traffic control device predicted in response to a repeated driving pattern at the geolocation, the repeated driving pattern being identified among collected acceleration, turning, and geolocation data from acceleration sensors, turning sensors, and geolocation systems in at least one vehicle across a plurality of vehicle drive cycles.

11. The method of claim 10, wherein the repeated driving pattern comprises a plurality of vehicle stops at the geolocation across the plurality of drive cycles.

12. The method of claim 11, further comprising defining a first time interval corresponding to a vehicle stop at a stop sign, a second time interval corresponding to a vehicle stop at a crosswalk, a third time interval corresponding to a yield sign, and a fourth time interval corresponding to a traffic light, and wherein the geolocation of the traffic control device is predicted in response to correlating the plurality of vehicle stops at the geolocation with one of the first, second, third, and fourth time intervals.

13. The method of claim 11, further comprising defining a first stop probability corresponding to a vehicle stop at a stop sign, a second stop probability corresponding to a vehicle stop at a crosswalk, a third stop probability corresponding to a yield sign, and a fourth stop probability corresponding to a traffic light, and wherein the geolocation of the traffic control device is predicted in response to correlating a percentage of vehicle trips through the geolocation that are vehicle stops with one of the first, second, third, and fourth stop probabilities.

14. The method of claim 10, wherein the traffic control device is a stoplight, and wherein the repeated driving pattern at the geolocation includes a pattern of first and second driving modes, the first mode including driving toward and through an intersection at a first heading without stopping and the second mode including driving toward the intersection at the first heading and stopping before driving through.