Virtual vehicle control method, model training method, control device and storage medium

1. A method for controlling a virtual vehicle in a racing game application, performed by a computing device having one or more processors and memory storing a plurality of computer-readable instructions to be executed by the one or more processors, the method comprising:

obtaining a virtual vehicle traveling picture, the virtual vehicle traveling picture comprising a picture of a virtual vehicle traveling on a virtual track of the racing game application;

extracting a virtual track image from the virtual vehicle traveling picture, the virtual track image being an image of a virtual track segment in which the virtual vehicle is located in real time;

inputting the virtual track image into a traveling decision-making model, the traveling decision-making model being generated through training according to a sample track image and a sample traveling policy, the sample track image and the sample traveling policy being extracted from a sample vehicle traveling picture, and the sample vehicle traveling picture comprising a picture of another virtual vehicle controlled by an actual player traveling on the virtual track of the racing game application; and

controlling the virtual vehicle to travel on the virtual track of the racing game application according to a target traveling policy outputted by the traveling decision-making model based on the virtual track image.

2. The method according to claim 1 , wherein the virtual vehicle traveling picture comprises a map display area, the map display area being used for displaying an image of the virtual track segment in which the virtual vehicle is located in real time; and

the extracting a virtual track image from the virtual vehicle traveling picture comprises:

extracting, according to coordinates of the map display area, an image displayed in the map display area; and

determining the image displayed in the map display area as the virtual track image.

3. The method according to claim 2 , wherein the image displayed in the map display area is transparent, and the determining the image displayed in the map display area as the virtual track image comprises:

performing feature extraction on the image displayed in the map display area, to obtain a track image and a vehicle image, the feature extraction comprising at least one of edge feature extraction and color feature extraction;

synthesizing the virtual track image according to the track image and the vehicle image that are extracted; and

scaling the virtual track image to a pre-determined size, the pre-determined size being an input image size specified by the traveling decision-making model.

4. The method according to claim 1 , wherein the virtual vehicle traveling picture further comprises an operation control, the operation control being configured to control the virtual vehicle; and

the controlling the virtual vehicle to travel on the virtual track of the racing game application according to a target traveling policy outputted by the traveling decision-making model based on the virtual track image comprises:

determining a target operation control corresponding to the target traveling policy, the target operation control being configured to trigger execution of the target traveling policy on the virtual vehicle in the racing game application; and

simulating, according to coordinates of the target operation control in the virtual vehicle traveling picture, an operation of triggering the target operation control on the virtual vehicle in the racing game application.

5. The method according to claim 4 , wherein the target traveling policy is represented by using a one-hot coded character string, a length of the one-hot coded character string is n, n being an integer greater than or equal to 2, each character in the one-hot coded character string corresponds to a traveling manner; and

the determining a target operation control corresponding to the target traveling policy comprises:

determining a target traveling manner corresponding to a maximum character in the one-hot coded character string; and

determining an operation control corresponding to the target traveling manner as the target operation control,

the target operation control being a left-turning control in a case that the target traveling manner is turning left, the target operation control being a right-turning control in a case that the target traveling manner is turning right, the target operation control being a left-turning control and a drift control in a case that the target traveling manner is drifting towards left, the target operation control being a right-turning control and a drift control in a case that the target traveling manner is drifting towards right, and the target operation control being a braking control in a case that the target traveling manner is deceleration.

6. The method according to claim 4 , wherein the simulating an operation of triggering the target operation control on the virtual vehicle in the racing game application comprises:

stopping simulating an operation of long pressing a first operation control, and simulating an operation of long pressing a second operation control, in a case that the target operation control is changed from the first operation control to the second operation control; and

simulating, in a case that the target operation control does not change, an operation of long pressing the target operation control on the virtual vehicle in the racing game application.

7. The method according to claim 1 , wherein the traveling decision-making model is generated through training further according to a sample vehicle speed, the sample vehicle speed being extracted from a picture of a sample vehicle running video; and

after the obtaining a virtual vehicle traveling picture, the method further comprises:

performing numeral recognition on the virtual vehicle traveling picture, and extracting a vehicle speed from the virtual vehicle traveling picture; and

after the inputting the virtual track image into a traveling decision-making model, the method further comprises:

inputting the vehicle speed into the traveling decision-making model.

8. The method according to claim 1 , wherein the traveling decision-making model is generated through training further according to sample obstacle information, the sample obstacle information being extracted from the picture of the sample vehicle running video; and

after the obtaining a virtual vehicle traveling picture, the method further comprises:

performing object recognition on the virtual vehicle traveling picture, and extracting obstacle information from the virtual vehicle traveling picture; and

after the inputting the virtual track image into a traveling decision-making model, the method further comprises:

inputting the obstacle information into the traveling decision-making model.

9. A computing device, comprising a memory and one or more processors, the memory storing computer-readable instructions, the computer-readable instructions, when being executed by the one or more processors, causing the computing device to perform a plurality of operations including:

obtaining a virtual vehicle traveling picture, the virtual vehicle traveling picture comprising a picture of a virtual vehicle traveling on a virtual track of a racing game application;

extracting a virtual track image from the virtual vehicle traveling picture, the virtual track image being an image of a virtual track segment in which the virtual vehicle is located in real time;

inputting the virtual track image into a traveling decision-making model, the traveling decision-making model being generated through training according to a sample track image and a sample traveling policy, the sample track image and the sample traveling policy being extracted from a sample vehicle traveling picture, and the sample vehicle traveling picture comprising a picture of another virtual vehicle controlled by an actual player traveling on the virtual track of the racing game application; and

controlling the virtual vehicle to travel on the virtual track of the racing game application according to a target traveling policy outputted by the traveling decision-making model based on the virtual track image.

10. The computing device according to claim 9 , wherein the virtual vehicle traveling picture comprises a map display area, the map display area being used for displaying an image of the virtual track segment in which the virtual vehicle is located in real time; and

the extracting a virtual track image from the virtual vehicle traveling picture comprises:

extracting, according to coordinates of the map display area, an image displayed in the map display area; and

determining the image displayed in the map display area as the virtual track image.

11. The computing device according to claim 10 , wherein the image displayed in the map display area is transparent, and the determining the image displayed in the map display area as the virtual track image comprises:

performing feature extraction on the image displayed in the map display area, to obtain a track image and a vehicle image, the feature extraction comprising at least one of edge feature extraction and color feature extraction;

synthesizing the virtual track image according to the track image and the vehicle image that are extracted; and

scaling the virtual track image to a pre-determined size, the pre-determined size being an input image size specified by the traveling decision-making model.

12. The computing device according to claim 9 , wherein the virtual vehicle traveling picture further comprises an operation control, the operation control being configured to control the virtual vehicle; and

the controlling the virtual vehicle to travel on the virtual track of the racing game application according to a target traveling policy outputted by the traveling decision-making model based on the virtual track image comprises:

determining a target operation control corresponding to the target traveling policy, the target operation control being configured to trigger execution of the target traveling policy on the virtual vehicle in the racing game application; and

simulating, according to coordinates of the target operation control in the virtual vehicle traveling picture, an operation of triggering the target operation control on the virtual vehicle in the racing game application.

13. The computing device according to claim 12 , wherein the target traveling policy is represented by using a one-hot coded character string, a length of the one-hot coded character string is n, n being an integer greater than or equal to 2, each character in the one-hot coded character string corresponds to a traveling manner; and

the determining a target operation control corresponding to the target traveling policy comprises:

determining a target traveling manner corresponding to a maximum character in the one-hot coded character string; and

determining an operation control corresponding to the target traveling manner as the target operation control,

the target operation control being a left-turning control in a case that the target traveling manner is turning left, the target operation control being a right-turning control in a case that the target traveling manner is turning right, the target operation control being a left-turning control and a drift control in a case that the target traveling manner is drifting towards left, the target operation control being a right-turning control and a drift control in a case that the target traveling manner is drifting towards right, and the target operation control being a braking control in a case that the target traveling manner is deceleration.

14. The computing device according to claim 12 , wherein the simulating an operation of triggering the target operation control on the virtual vehicle in the racing game application comprises:

stopping simulating an operation of long pressing a first operation control, and simulating an operation of long pressing a second operation control, in a case that the target operation control is changed from the first operation control to the second operation control; and

simulating, in a case that the target operation control does not change, an operation of long pressing the target operation control on the virtual vehicle in the racing game application.

15. The computing device according to claim 9 , wherein the traveling decision-making model is generated through training further according to a sample vehicle speed, the sample vehicle speed being extracted from a picture of a sample vehicle running video; and the plurality of operations further comprise:

after obtaining a virtual vehicle traveling picture, performing numeral recognition on the virtual vehicle traveling picture, and extracting a vehicle speed from the virtual vehicle traveling picture; and

after inputting the virtual track image into a traveling decision-making model, inputting the vehicle speed into the traveling decision-making model.

16. The computing device according to claim 9 , wherein the traveling decision-making model is generated through training further according to sample obstacle information, the sample obstacle information being extracted from the picture of the sample vehicle running video; and the plurality of operations further comprise:

after obtaining a virtual vehicle traveling picture, performing object recognition on the virtual vehicle traveling picture, and extracting obstacle information from the virtual vehicle traveling picture; and

after inputting the virtual track image into a traveling decision-making model, inputting the obstacle information into the traveling decision-making model.

17. A non-transitory computer-readable storage medium storing a plurality of computer-readable instructions, the computer-readable instructions, when being executed by one or more processors of a computing device, causing the computing device to perform a plurality of operations including:

obtaining a virtual vehicle traveling picture, the virtual vehicle traveling picture comprising a picture of a virtual vehicle traveling on a virtual track of a racing game application;

extracting a virtual track image from the virtual vehicle traveling picture, the virtual track image being an image of a virtual track segment in which the virtual vehicle is located in real time;

inputting the virtual track image into a traveling decision-making model, the traveling decision-making model being generated through training according to a sample track image and a sample traveling policy, the sample track image and the sample traveling policy being extracted from a sample vehicle traveling picture, and the sample vehicle traveling picture comprising a picture of another virtual vehicle controlled by an actual player traveling on the virtual track of the racing game application; and

controlling the virtual vehicle to travel on the virtual track of the racing game application according to a target traveling policy outputted by the traveling decision-making model based on the virtual track image.

18. The non-transitory computer-readable storage medium according to claim 17 , wherein the virtual vehicle traveling picture comprises a map display area, the map display area being used for displaying an image of the virtual track segment in which the virtual vehicle is located in real time; and

the extracting a virtual track image from the virtual vehicle traveling picture comprises:

extracting, according to coordinates of the map display area, an image displayed in the map display area; and

determining the image displayed in the map display area as the virtual track image.

19. The non-transitory computer-readable storage medium according to claim 18 , wherein the image displayed in the map display area is transparent, and the determining the image displayed in the map display area as the virtual track image comprises:

performing feature extraction on the image displayed in the map display area, to obtain a track image and a vehicle image, the feature extraction comprising at least one of edge feature extraction and color feature extraction;

synthesizing the virtual track image according to the track image and the vehicle image that are extracted; and

scaling the virtual track image to a pre-determined size, the pre-determined size being an input image size specified by the traveling decision-making model.

20. The non-transitory computer-readable storage medium according to claim 17 , wherein the virtual vehicle traveling picture further comprises an operation control, the operation control being configured to control the virtual vehicle; and

the controlling the virtual vehicle to travel on the virtual track of the racing game application according to a target traveling policy outputted by the traveling decision-making model based on the virtual track image comprises:

determining a target operation control corresponding to the target traveling policy, the target operation control being configured to trigger execution of the target traveling policy on the virtual vehicle in the racing game application; and

simulating, according to coordinates of the target operation control in the virtual vehicle traveling picture, an operation of triggering the target operation control on the virtual vehicle in the racing game application.