Controlling an autonomous vehicle based on independent driving decisions

Patent metadata

Field	Value
Publication number	US-10503172-B2
Application number	US-201816149219-A
Country	US
Kind code	B2
Filing date	Oct 2, 2018
Priority date	Oct 18, 2017
Publication date	Dec 10, 2019
Grant date	Dec 10, 2019

Abstract

A computer-readable medium stores instructions executable by one or more processors to implement an aggregate self-driving control architecture (SDCA) for controlling an autonomous vehicle. The aggregate SDCA includes a plurality of SDCAs each including a different motion planner. Each motion planner is configured to receive signals descriptive of a current state of an environment through which the autonomous vehicle is moving, and each SDCA is configured to generate candidate decisions for controlling the autonomous vehicle by using the respective motion planner to process the received signals. The aggregate SDCA also includes a decision arbiter configured to receive the candidate decisions generated by the SDCAs, generate decisions for controlling the autonomous vehicle by processing the candidate decisions, and provide signals indicative of the generated decisions to one or more operational subsystems of the vehicle to effectuate maneuvering of the vehicle.

First claim

What is claimed is:

1. A non-transitory computer-readable medium storing thereon instructions executable by one or more processors to implement an aggregate self-driving control architecture for controlling an autonomous vehicle, the aggregate self-driving control architecture comprising:
a plurality of self-driving control architectures each including a different one of a plurality of motion planners, wherein each of the motion planners is configured to receive signals descriptive of a current state of an environment through which the autonomous vehicle is moving, and wherein each of the plurality of self-driving control architectures is configured to generate candidate decisions for controlling the autonomous vehicle by using the motion planner included in the self-driving control architecture to process the received signals; and
a decision arbiter configured to (i) receive the candidate decisions generated by the self-driving control architectures, (ii) generate decisions for controlling the autonomous vehicle by processing the received candidate decisions, and (iii) provide signals indicative of the generated decisions to one or more operational subsystems of the autonomous vehicle to maneuver the autonomous vehicle in accordance with the generated decisions, wherein the decision arbiter includes an arbitration machine learning (ML) model trained to dynamically weight the candidate decisions of different self-driving control architectures based on observed or expected circumstances of the autonomous vehicle.

2. The non-transitory computer-readable medium of claim 1 , wherein the candidate decisions generated by one or more of the self-driving control architectures indicate one or both of:
desired operational parameters for the autonomous vehicle, the desired operational parameters including one or more of (i) braking parameters, (ii) acceleration parameters, (iii) speed parameters, or (iv) direction parameters; and
desired maneuvers for the autonomous vehicle.

3. The non-transitory computer-readable medium of claim 2 , wherein the candidate decisions generated by another one or more of the self-driving control architectures indicate one or both of:
one or more ranges of disallowed operational parameters for the autonomous vehicle; and
one or more disallowed maneuvers for the autonomous vehicle.

4. The non-transitory computer-readable medium of claim 3 , wherein the candidate decisions generated by a further one or more of the self-driving control architectures indicate one or both of:
one or more ranges of allowed operational parameters for the autonomous vehicle; and
one or more allowed maneuvers for the autonomous vehicle.

5. The non-transitory computer-readable medium of claim 1 , wherein the self-driving control architectures include two or more of:
a machine learning based planner;
a search based planner;
a sampling based planner; and
a predictive control based planner.

6. The non-transitory computer-readable medium of claim 5 , wherein at least one of the self-driving control architectures includes the machine learning based planner.

7. The non-transitory computer-readable medium of claim 1 , wherein the plurality of self-driving control architectures further includes one or more perception components each configured to:
receive sensor data;
segment the received sensor data into objects;
classify the segmented objects according to object types;
track movement of the classified objects over time; and
generate, based on the classified and tracked objects, at least a portion of the signals descriptive of the current state of the environment through which the autonomous vehicle is moving.

8. The non-transitory computer-readable medium of claim 7 , wherein:
segmenting the received sensor data comprises segmenting lane markings located in the environment through which the autonomous vehicle is moving;
classifying the segmented objects comprises classifying the segmented lane markings into lane-marking types; and
tracking the movement of the classified lane markings comprises tracking a geometric property of the lane markings over time.

9. The non-transitory computer-readable medium of claim 7 , wherein the sensor data includes, or is generated based on, one or more of (i) data generated by one or more lidar devices, (ii) data generated by one or more camera devices, (iii) data generated by one or more radar devices, (iv) data generated by one or more thermal sensor devices, (v) data generated by one or more inertial measurement units (IMUs), and (vi) data generated by one or more global positioning system (GPS) units.

10. The non-transitory computer-readable medium of claim 7 , wherein the one or more perception components include a plurality of perception components, and wherein each of the self-driving control architectures includes a different one of the plurality of perception components.

11. The non-transitory computer-readable medium of claim 1 , wherein one or more of the self-driving control architectures include a prediction component configured to estimate future positions of tracked objects.

12. The non-transitory computer-readable medium of claim 1 , wherein:
the candidate decisions generated by two or more of the self-driving control architectures are indicative of desired maneuvers for the autonomous vehicle; and
the decision arbiter is configured to generate decisions for controlling the autonomous vehicle at least by determining which desired maneuver is indicated by more of the self-driving control architectures than any other desired maneuver.

13. The non-transitory computer-readable medium of claim 1 , wherein:
the candidate decisions generated by two or more of the self-driving control architectures are indicative of desired operational parameters for the autonomous vehicle, the desired operational parameters including one or more of (i) braking parameters, (ii) acceleration parameters, (iii) speed parameters, or (iv) direction parameters; and
the decision arbiter is configured to generate decisions for controlling the autonomous vehicle at least by performing a mathematical operation on the desired operational parameters.

14. The non-transitory computer-readable medium of claim 13 , wherein the mathematical operation is configured to combine or reduce the desired operational parameters into a single set of operational parameters.

15. The non-transitory computer-readable medium of claim 1 , wherein the decision arbiter is configured to generate decisions for controlling the autonomous vehicle at least by removing from consideration one or more candidate decisions that are statistical outliers relative to other candidate decisions.

16. The non-transitory computer-readable medium of claim 1 , wherein:
the decision arbiter includes a safety module configured to provide safety signals indicative of one of both of (i) allowable maneuvers for the autonomous vehicle, and (ii) allowable operational parameters for the autonomous vehicle; and
the decision arbiter is configured to generate decisions for controlling the autonomous vehicle by (i) generating provisional decisions for controlling the autonomous vehicle by processing the received candidate decisions, and (ii) comparing the provisional decisions to the safety signals.

17. The non-transitory computer-readable medium of claim 1 , wherein the arbitration ML model is trained to, in some instances, dynamically select the candidate decisions of specific self-driving control architectures based on observed or expected circumstances of the autonomous vehicle.

Assignees

• Luminar Tech Inc

Inventors

• Englard Benjamin
• Augenbraun Joseph

Classifications

• B60W60/0015 Primary

  specially adapted for safety · CPC title

• B60W40/02

  related to ambient conditions · CPC title

• G01C21/3492

  employing speed data or traffic data, e.g. real-time or historical (traffic control systems for road vehicles involving transmission of navigation instructions to the vehicle G08G1/0968) · CPC title

• G06N5/022

  Knowledge engineering; Knowledge acquisition · CPC title

• G06F16/285

  Clustering or classification · CPC title