Neural network based vehicle dynamics model

Patent metadata

Field	Value
Publication number	US-12007778-B2
Application number	US-202318094363-A
Country	US
Kind code	B2
Filing date	Jan 8, 2023
Priority date	Aug 8, 2017
Publication date	Jun 11, 2024
Grant date	Jun 11, 2024

Abstract

A system and method for implementing a neural network based vehicle dynamics model are disclosed. A particular embodiment includes: training a machine learning system with a training dataset corresponding to a desired autonomous vehicle simulation environment; receiving vehicle control command data and vehicle status data, the vehicle control command data not including vehicle component types or characteristics of a specific vehicle; by use of the trained machine learning system, the vehicle control command data, and vehicle status data, generating simulated vehicle dynamics data including predicted vehicle acceleration data; providing the simulated vehicle dynamics data to an autonomous vehicle simulation system implementing the autonomous vehicle simulation environment; and using data produced by the autonomous vehicle simulation system to modify the vehicle status data for a subsequent iteration.

First claim

What is claimed is:

1. A system comprising:
a data processor; and
a memory storing a vehicle dynamics modeling module, executable by the data processor to:
receive vehicle control command data and vehicle status data;
generate a first set of simulated vehicle dynamics data based on the vehicle control command data and the vehicle status data by use of a machine learning system trained on a first training dataset comprising recorded historical vehicle driving data captured from real world vehicle operations;
generate a second set of simulated vehicle dynamics data based on the vehicle control command data and the vehicle status data by use of the machine learning system trained on a second training dataset comprising recorded historical vehicle driving data captured from real world vehicle operations, the second set of simulated vehicle dynamics data being different from the first set of simulated vehicle dynamics data, the first training dataset and the second training dataset each representing a different particular vehicle simulation environment with particular types of vehicles having a defined set of characteristics;
provide the first and the second set of simulated vehicle dynamics data to an autonomous vehicle simulation system implementing one of the particular vehicle simulation environments;
modify, by use of the autonomous vehicle simulation system, the vehicle status data by use of either the first or the second set of simulated vehicle dynamics data; and
generate predicted simulated vehicle acceleration data for simulated autonomous vehicles based on the vehicle control command data and the modified vehicle status data.

2. The system of claim 1 wherein the vehicle control command data comprises throttle, brake, and steering control information corresponding to simulated autonomous vehicles.

3. The system of claim 1 wherein the vehicle status data comprises speed and pitch information corresponding to simulated autonomous vehicles.

4. The system of claim 1 wherein the first training dataset includes training data different from training data of the second training dataset.

5. The system of claim 1 wherein the first training dataset or the second training dataset include training data corresponding to one of the particular vehicle simulation environments.

6. The system of claim 1 wherein the first training dataset or the second training dataset include training data corresponding to the recorded historical vehicle driving data captured from real world vehicle operations or simulated vehicle movements.

7. The system of claim 1 wherein the vehicle dynamics modeling module is further configured to use the simulated vehicle dynamics data to generate validation data to validate the first training dataset or the second training dataset.

8. The system of claim 1 wherein the vehicle dynamics modeling module is further configured to generate validation data to validate an accuracy of the first training dataset or the second training dataset.

9. The system of claim 1 wherein the simulated vehicle dynamics data comprises predicted vehicle dynamics data, wherein the predicted vehicle dynamics data is generated based on the vehicle control command data and the vehicle status data, by use of the machine learning system.

10. The system of claim 1 wherein the simulated vehicle dynamics data comprises predicted vehicle dynamics data, the predicted vehicle dynamics data comprises at least one of predicted acceleration data and predicted torque data, wherein the at least one of the predicted acceleration data and the predicted torque data is generated by use of the trained machine learning system based on the vehicle control command data and the vehicle status data.

11. A method comprising:
receiving vehicle control command data and vehicle status data;
generating a first set of simulated vehicle dynamics data based on the vehicle control command data and the vehicle status data by use of a machine learning system trained on a first training dataset comprising recorded historical vehicle driving data captured from real world vehicle operations;
generating a second set of simulated vehicle dynamics data based on the vehicle control command data and the vehicle status data by use of the machine learning system trained on a second training dataset comprising recorded historical vehicle driving data captured from real world vehicle operations, the second set of simulated vehicle dynamics data being different from the first set of simulated vehicle dynamics data, the first training dataset and the second training dataset each representing a different particular vehicle simulation environment with particular types of vehicles having a defined set of characteristics;
providing the first and the second set of simulated vehicle dynamics data to an autonomous vehicle simulation system implementing one of the particular vehicle simulation environments;
modifying, by use of the autonomous vehicle simulation system, the vehicle status data by use of either the first or the second set of simulated vehicle dynamics data; and
generating predicted simulated vehicle acceleration data for simulated autonomous vehicles based on the vehicle control command data and the modified vehicle status data.

12. The method of claim 11 wherein the machine learning system comprises artificial neural networks or connectionist systems.

13. The method of claim 11 wherein the vehicle control command data comprises throttle and brake control information corresponding to simulated autonomous vehicles.

14. The method of claim 11 wherein the vehicle status data comprises speed information corresponding to simulated autonomous vehicles.

15. The method of claim 11 wherein the first training dataset includes training data different from training data of the second training dataset.

16. The method of claim 11 wherein the simulated vehicle dynamics data comprises predicted vehicle torque data generated by use of the trained machine learning system based on the vehicle control command data and the vehicle status data, wherein the vehicle control command data does not comprise steering control data, wherein the vehicle status data does not comprise pitch information.

17. The method of claim 11 wherein the simulated vehicle dynamics data comprises predicted vehicle torque data generated by use of the trained machine learning system based on the vehicle control command data and the vehicle status data, wherein vehicle speed data is generated based on the predicted vehicle torque data.

18. The method of claim 11 including generating validation data to validate an accuracy of the first training dataset or the second training dataset.

19. A non-transitory machine-useable storage medium embodying instructions which, when executed by a machine, cause the machine to:
receive vehicle control command data and vehicle status data;
generate a first set of simulated vehicle dynamics data based on the vehicle control command data and the vehicle status data by use of a machine learning system trained on a first training dataset comprising recorded historical vehicle driving data captured from real world vehicle operations;
generate a second set of simulated vehicle dynamics data based on the vehicle control command data and the vehicle status data by use of the machine learning system trained on a second training dataset comprising recorded historical vehicle driving data captured from real world vehicle operations, the second set of simulated vehicle dynamics data being different fro

Assignees

• Tusimple Inc

Inventors

• Chen Jun
• Lin Wutu
• Liu Liu

Classifications

• G06N3/08 Primary

  Learning methods · CPC title

• G06N3/09

  Supervised learning · CPC title

• G06N3/0499

  Feedforward networks · CPC title

• G05D1/246

  using environment maps, e.g. simultaneous localisation and mapping [SLAM] · CPC title

• G05D1/247

  using signals provided by artificial sources external to the vehicle, e.g. navigation beacons · CPC title