Method and system for training automatic driving model

Patent metadata

Field	Value
Publication number	US-2025190860-A1
Application number	US-202418638243-A
Country	US
Kind code	A1
Filing date	Apr 17, 2024
Priority date	Dec 6, 2023
Publication date	Jun 12, 2025
Grant date	—

Abstract

A method for training an automatic driving model comprises: obtaining a sample acquisition strategy of the automatic driving model; transmitting the sample acquisition strategy of the automatic driving model to a vehicle equipment, the vehicle equipment being configured to sort objective driving data from a driving data set based on the sample acquisition strategy; receiving the objective driving data from the vehicle equipment; adding the objective driving data into a training set of the automatic driving model; and training the automatic driving model based on the training set. A system for training the automatic driving model is also disclosed.

First claim

What is claimed is:

1 . A method for training an automatic driving model, comprising:
obtaining a sample acquisition strategy of the automatic driving model; transmitting the sample acquisition strategy of the automatic driving model to a vehicle equipment for sorting objective driving data; receiving the objective driving data from the vehicle equipment; adding the objective driving data into a training set of the automatic driving model; and training the automatic driving model based on the training set.

2 . The method of claim 1 , wherein the sample acquisition strategy comprises a driving scene of the objective driving data, obtaining the sample acquisition strategy of the automatic driving model further comprises:
obtaining a vehicle control effect of the automatic driving model in a test driving scene; and setting the test driving scene as the driving scene of the objective driving data when the vehicle control effect of the automatic driving model in the test driving scene does not meet a preset driving safety standard.

3 . The method of claim 2 , wherein obtaining the vehicle control effect of the automatic driving model in the test driving scene further comprises:
obtaining test samples of the test driving scene; inputting the test samples into the automatic driving model to obtain a vehicle control command corresponding to the test driving scene; and detecting whether an execution result of the vehicle control command in the test driving scene meets the preset driving safety standard, to obtain the vehicle control effect.

4 . The method of claim 2 , wherein after training the automatic driving model based on the training set, the method further comprises:
continuing to obtain the vehicle control effect of the automatic driving model in the test driving scene until the vehicle control effect of the automatic driving model in the test driving scene meets the preset driving safety standard.

5 . The method of claim 1 , further comprising:
deploying the trained automatic driving model to a vehicle, wherein the vehicle operates in an automatic driving mode based on the trained automatic driving model.

6 . A method for training an automatic driving model, comprising:
collecting driving data of a vehicle when the vehicle is being driven; receiving a sample acquisition strategy of the automatic driving model from a computer device; sorting objective driving data from the collected driving data based on the sample acquisition strategy; and transmitting the objective driving data to the computer device, wherein the objective driving data is added into a training set of the automatic driving model, and the automatic driving model is trained based on the training set.

7 . The method of claim 6 , wherein the sample acquisition strategy comprises a driving scene of the objective driving data, sorting the objective driving data from the collected driving data based on the sample acquisition strategy further comprises:
obtaining driving scenes of the collected driving data; and sorting the objective driving data meeting the sample acquisition strategy from the collected driving data based on the driving scenes of the collected driving data.

8 . The method of claim 7 , wherein obtaining the driving scenes of the collected driving data further comprising:
inputting the collected driving data into a pre-trained scene recognition model to obtain the driving scenes of the collected driving data.

9 . The method of claim 6 , further comprising:
deploying the automatic driving model trained by the computer device, wherein the vehicle operates in an automatic driving mode based on the trained automatic driving model.

10 . A system for training an automatic driving model, comprising;
a vehicle comprising a vehicle equipment; and a computer device communicated with the vehicle equipment; wherein the vehicle equipment is configured to:
collect driving data of the vehicle in a driving process,
receive a sample acquisition strategy of the automatic driving model from the computer device,
sort objective driving data from the collected driving data based on the sample acquisition strategy, and
transmit the objective driving data to the computer device;
the computer device is configured to:
obtain the sample acquisition strategy of the automatic driving model,
transmit the sample acquisition strategy of the automatic driving model to the vehicle equipment,
receive the objective driving data from the vehicle equipment,
add the objective driving data into a training set of the automatic driving model, and
train the automatic driving model based on the training set;
wherein the vehicle operates in an automatic driving mode based on the trained automatic driving model.

11 . The system of claim 10 , wherein the sample acquisition strategy comprises a driving scene of the objective driving data, when the computer device obtains the sample acquisition strategy of the automatic driving model, the computer device is further configured to:
obtain a vehicle control effect of the automatic driving model in a test driving scene; and set the test driving scene as the driving scene of the objective driving data when the vehicle control effect of the automatic driving model in the test driving scene does not meet a preset driving safety standard.

12 . The system of claim 11 , wherein when the computer device obtains the vehicle control effect of the automatic driving model in the test driving scene, the computer device is further configured to:
obtain test samples of the test driving scene; input the test samples into the automatic driving model to obtain a vehicle control command corresponding to the test driving scene; and detect whether an execution result of the vehicle control command in the test driving scene meets the preset driving safety standard, to obtain the vehicle control effect.

13 . The system of claim 11 , wherein after the computer device trains the automatic driving model based on the training set, the computer device is further configured to:
continue to obtain the vehicle control effect of the automatic driving model in the test driving scene until the vehicle control effect of the automatic driving model in the test driving scene meets the preset driving safety standard.

14 . The system of claim 10 , wherein the sample acquisition strategy comprises a driving scene of the objective driving data, when the vehicle equipment sorts the objective driving data from the collected driving data based on the sample acquisition strategy, the vehicle equipment is further configured to:
obtain driving scenes of the collected driving data; and sort the objective driving data meeting the sample acquisition strategy from the collected driving data based on the driving scenes of the collected driving data.

15 . The system of claim 14 , wherein when the vehicle equipment obtains the driving scenes of the collected driving data, the vehicle equipment is further configured to:
input the collected driving data into a pre-trained scene recognition model to obtain the driving scenes of the collected driving data.

Assignees

• Hon Hai Prec Ind Co Ltd

Inventors

• Lee Chieh
• Kuo Chin-Pin

Classifications

• G06N20/00 Primary

  Machine learning · CPC title

• B60W60/0015

  specially adapted for safety · CPC title

• B60W50/0098

  Details of control systems ensuring comfort, safety or stability not otherwise provided for · CPC title

• B60W2556/45

  External transmission of data to or from the vehicle · CPC title

• G07C5/02 Primary

  Registering or indicating driving, working, idle, or waiting time only (apparatus forming part of taximeters G07B13/00) · CPC title