Generating testing instances for autonomous vehicles

Patent metadata

Field	Value
Publication number	US-11086319-B2
Application number	US-201916271401-A
Country	US
Kind code	B2
Filing date	Feb 8, 2019
Priority date	Nov 2, 2018
Publication date	Aug 10, 2021
Grant date	Aug 10, 2021

Abstract

Sensor data collected from an autonomous vehicle can be labeled using sensor data collected from an additional vehicle. Labeled sensor data can generate targeted testing instances for a trained machine learning model, where the trained machine learning model is used in generating control signals for an autonomous vehicle. In many implementations, targeted training instances can generate an accuracy value for the trained neural network model. Additionally or alternatively, the sensor suite on the additional vehicle can include a removable hardware pod which can be mounted on a variety of vehicles.

First claim

What is claimed:

1. A method of generating testing instances for a machine learning model trained for autonomous vehicle control, the method comprising:
receiving autonomous vehicle data using an autonomous vehicle sensor suite on an autonomous vehicle;
receiving additional vehicle data using an additional vehicle sensor suite on an additional vehicle,
wherein the additional vehicle is in an environment with the autonomous vehicle,
wherein the additional vehicle is a non-autonomous vehicle and the sensor suite on the additional vehicle is a removable hardware pod, and
wherein the removable hardware pod is mounted onto the additional vehicle;

determining an instance of autonomous vehicle data wherein at least one of the sensors of the autonomous vehicle sensor suite detect the additional vehicle in the environment;
determining an instance of additional vehicle data which temporally corresponds with the instance of autonomous vehicle data where the autonomous vehicle detects the additional vehicle; and
generating a plurality of testing instances for a trained machine learning model, wherein each testing instance includes the instance of autonomous vehicle data where the autonomous vehicle detects the additional vehicle and a label indicating a current state of at least one attribute of the additional vehicle using the determined corresponding instance of additional vehicle data.

2. The method of claim 1 , further comprising:
testing the trained machine learning model, by applying the autonomous vehicle data portion of the testing instance as input to the trained machine learning model to generate output of the machine learning model; and
determining an accuracy value of the machine learning model by determining a difference between the output of the machine learning model and the label portion of the testing instance from the corresponding instance of the additional vehicle data.

3. The method of claim 1 , further comprising:
receiving additional vehicle data collected from a plurality of additional vehicles in the environment with the autonomous vehicle;
determining an instance of the autonomous vehicle data where at least one of the sensors in the autonomous vehicle sensor suite detects each additional vehicle in the plurality of additional vehicles;
determining an instance in the additional vehicle data corresponding to each additional vehicle which temporally correlates with the instance of autonomous vehicle data where each of the additional vehicles are detected by the autonomous vehicle; and
generating a plurality of testing instances for the trained machine learning model, wherein each testing instance includes the instance in the autonomous vehicle data where the autonomous vehicle detects each of the additional vehicles and a plurality of labels, each label indicating a current state of at least one attribute of the additional vehicle using the determined corresponding instance of additional vehicle data.

4. The method of claim 1 , further comprising:
determining a location of the autonomous vehicle in the environment in each instance of the autonomous vehicle data and a location of the additional vehicle in the environment in each instance of the additional vehicle data by utilizing GPS data from each instance in the autonomous vehicle data and GPS data from each instance in the additional vehicle data.

5. The method of claim 1 , further comprising:
determining a location of the autonomous vehicle in the environment in each instance in the autonomous vehicle data and a location of the additional vehicle in the environment by localizing the autonomous vehicle and the additional vehicle using one or more common landmarks in the environment identified in the instance of the autonomous vehicle data and instance of the additional vehicle data.

6. The method of claim 1 , wherein the additional vehicle is a second autonomous vehicle.

7. The method of claim 1 , wherein the sensor suite on the autonomous vehicle comprises at least a Global Positioning System (GPS) unit, a radio direction and ranging (RADAR) unit, a light detection and ranging (LIDAR) unit, one or more cameras, and an inertial measurement (IMU) unit.

8. The method of claim 7 , wherein autonomous vehicle data comprises at least GPS data, RADAR data, LIDAR data, one or more images from the one or more cameras, and IMU data.

9. The method of claim 1 , wherein the additional vehicle is selected from a group consisting of a car, a van, a truck, a bus, a motorcycle, and a tractor trailer.

10. The method of claim 1 , wherein the removable hardware pod comprises at least a Global Positioning System (GPS) unit, a light detection and ranging (LIDAR) unit, and one or more cameras.

11. The method of claim 10 , wherein additional vehicle data comprises at least GPS data, LIDAR data, one or more images from the one or more cameras, IMU data, CAN bus data, and known additional vehicle data.

12. The method of claim 11 , wherein the known additional vehicle data is selected from a group consisting of a vehicle make, a vehicle model, a vehicle color, a vehicle year, one or more vehicle dimension measurements, and a position of where the removable hardware pod is mounted on the additional vehicle, and combinations thereof.

13. The method of claim 1 , wherein the sensor suite on the additional vehicle includes one or more additional removable hardware pods, and wherein each of the one or more additional removable hardware pods are mounted onto the additional vehicle.

14. The method of claim 13 , wherein receiving the additional vehicle data comprises receiving corresponding additional vehicle data from each of the one or more additional removable hardware pods mounted onto the additional vehicle.

15. A system comprising one or more processors and a non-transitory memory operably coupled with the one or more processors, wherein the memory stores instructions that, in response to the execution of the instructions by the one or more processors, cause the one or more processors to perform the following operations:
receiving autonomous vehicle data using an autonomous vehicle sensor suite on an autonomous vehicle;
receiving additional vehicle data using an additional vehicle sensor suite on an additional vehicle,
wherein the additional vehicle is in an environment with the autonomous vehicle,
wherein the additional vehicle is a non-autonomous vehicle and the sensor suite on the additional vehicle is a removable hardware pod, and
wherein the removable hardware pod is mounted onto the additional vehicle;

determining an instance of autonomous vehicle data wherein at least one of the sensors of the autonomous vehicle sensor suite detect the additional vehicle in the environment;
determining an instance of additional vehicle data which temporally corresponds with the instance of autonomous vehicle data where the autonomous vehicle detects the additional vehicle; and
generating a plurality of testing instances for a trained machine learning model, wherein each testing instance includes the instance of autonomous vehicle data where the autonomous vehicle detects the additional vehicle and a label indicating a current state of at least one attribute of the additional vehicle using the determined corresponding instance of additional vehicle data.

16. At least one non-transitory computer-readable medium comprising instructions that, in response to execution of the instructions by one or more processors, cause the one or more processors to perform the following operations:
receiving autonomous vehicle data using an autonomous vehicle sensor

Assignees

• Aurora Innovation Inc
• Aurora Operations Inc

Inventors

• Valois Jean-Sebastien
• Munoz Daniel

Classifications

• G06N3/0464

  Convolutional networks [CNN, ConvNet] · CPC title

• G06N3/0895

  Weakly supervised learning, e.g. semi-supervised or self-supervised learning · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/08

  Learning methods · CPC title

• G05B13/0265

  the criterion being a learning criterion · CPC title