CN-122009220-A - Intelligent driving method, device and storage medium

CN122009220ACN 122009220 ACN122009220 ACN 122009220ACN-122009220-A

Abstract

The present application relates to the field of intelligent driving, and more particularly, to an intelligent driving method, apparatus, and computer storage medium based on an end-to-end model. The method comprises the steps of obtaining perception data from a sensor and navigation information from a navigation system, encoding the perception data and the navigation information to generate a plurality of token, wherein each token is represented by a plurality of types of tensors, stitching tensors of the same type in the plurality of token to generate a plurality of types of stitched tensors, and providing the plurality of types of stitched tensors to an end-to-end model to generate an instruction for controlling vehicle running.

Inventors

REN SHAOQING
ZHANG ZHI
Liu Qikang
CHEN KUNSHENG
LIU GUOYI
CHENG ZHENGXIN
FU XIAOXIN
YE CHAOQIANG
SHE XIAOLI

Assignees

安徽蔚来智驾科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260409

Claims (10)

1. An intelligent driving method, characterized in that the method comprises: acquiring sensing data from a sensor and navigation information from a navigation system; encoding the sensory data and the navigational information to generate a plurality of token, wherein each token is represented by a plurality of types of tensors; splicing tensors of the same type in the plurality of token to generate a plurality of types of spliced tensors, and The plurality of types of stitched tensors are provided to an end-to-end model to generate instructions for controlling vehicle travel.
2. The method of claim 1 wherein the plurality of types of tensors comprises a polygon tensor, a multi-segment line tensor, a mask tensor, and an attribute tensor.
3. The method of claim 1, wherein the navigation information includes road topology information, guidance information, and lane information.
4. The method of claim 3, wherein the road topology information includes a main path and a bifurcation path associated with the main path, the guidance information includes a main action and an auxiliary action associated with the initiative, and the lane information includes a lane type, a lane position number, whether a recommended lane, a traveling direction of the recommended lane, a distance from a next section of lane.
5. The method of claim 3, wherein encoding the road topology information to generate a road token comprises: encoding the road topology information into a multi-segment line tensor and an attribute tensor of the road token, and The mask tensor for the road token is used to indicate that the polygon tensor for the road token is invalid and the polyline tensor is valid.
6. The method of claim 5 wherein encoding the road topology information to generate a road token further comprises interpolating form points on a main path and a bifurcation path in the road topology information to match dimensions of the multi-segment line tensor.
7. The method of claim 3, wherein encoding the inducement information to generate an action token comprises: encoding the inducement information into an attribute tensor of the action token, and And setting values in the polygon tensor, the multi-segment line tensor and the mask tensor of the action token as filling values.
8. The method of claim 3, wherein, for each lane on a road, encoding lane information for that lane to generate a respective lane token comprises: encoding the lane information into an attribute tensor of the lane token, and And setting values in the polygon tensor, the multi-section line tensor and the mask tensor of the lane token as filling values.
9. An intelligent driving apparatus, characterized in that the apparatus comprises a memory, a processor, and a computer program stored on the memory and executable by the processor, the execution of the computer program causing the method according to any one of claims 1-8 to be performed.
10. A computer storage medium comprising instructions which, when executed, perform the method of any one of claims 1-8.

Description

Intelligent driving method, device and storage medium Technical Field The present application relates to the field of intelligent driving, and more particularly, to an intelligent driving method, apparatus, and computer storage medium based on an end-to-end model. Background In intelligent driving solutions, an End-to-End (End-to-End) intelligent driving model is receiving increasing attention. In contrast to traditional modular pipelines (e.g., modules including sensing, prediction, planning, control, etc.), the end-to-end model is able to directly map various raw data into instructions for controlling vehicle travel, thereby completing a full flow decision in a single model. The method avoids information loss and error accumulation among modules, has zero sample learning capability, and can better cope with various driving scenes. It is noted that the information disclosed in the above background section is only for enhancement of understanding of the background of the application and thus may include information that does not form the prior art that is already known to those of ordinary skill in the art. Disclosure of Invention The application provides an intelligent driving method, an intelligent driving device for realizing the method and a computer storage medium. The method can fully utilize navigation information of a standard definition (Standard Definition, SD) map through a standardized token architecture without excessive modification of an end-to-end model architecture. Therefore, the method can reduce redundant learning of the model on the characteristics, reduce the calculation amount of the model and shorten decision delay. According to a first aspect of the application there is provided an intelligent driving method comprising obtaining perception data from a sensor and navigation information from a navigation system, encoding the perception data and the navigation information to generate a plurality of token, wherein each token is represented by a plurality of types of tensors, stitching the same types of tensors in the plurality of token to generate a plurality of types of stitched tensors, and providing the plurality of types of stitched tensors to an end-to-end model to generate instructions for controlling vehicle travel. Alternatively or additionally to the above, in a method according to an embodiment of the application, the plurality of types of tensors include a polygon (polygon) tensor, a polyline (polyline) tensor, a mask (mask) tensor, and a property (property) tensor. Alternatively or additionally to the above, in a method according to an embodiment of the application, the navigation information includes road topology information, guidance information, and lane information. Alternatively or additionally to the above, in the method according to an embodiment of the present application, the road topology information includes a main path (main-path) and a branch path (sub-path) associated with the main path, the guidance information includes a main action (main-action) and an auxiliary action (auxiliary-action) associated with the main action, and the lane information includes a lane type, a lane position number, whether it is a recommended lane, a driving direction of the recommended lane, a distance from a next lane. Alternatively or additionally to the above, in a method according to an embodiment of the application, encoding the road topology information to generate a road token includes encoding the road topology information into a polyline tensor and an attribute tensor of the road token, and using a mask tensor of the road token to indicate that a polygon tensor of the road token is invalid and the polyline tensor is valid. Alternatively or additionally to the above, in a method according to an embodiment of the application, encoding the road topology information to generate a road token further comprises interpolating form points on a main path and a bifurcation path in the road topology information to match dimensions of the multi-segment line tensor. Alternatively or additionally to the above, in a method according to an embodiment of the application, encoding the inducement information to generate the action token comprises encoding the inducement information into an attribute tensor of the action token and setting values in a polygon tensor, a multi-segment line tensor, and a mask tensor of the action token as padding values. Alternatively or additionally to the above, in a method according to an embodiment of the application, encoding lane information for each lane on a road to generate a corresponding lane token comprises encoding the lane information into an attribute tensor for the lane token, and setting values in a polygon tensor, a multi-segment line tensor, and a mask tensor for the lane token as padding values. According to a second aspect of the present application there is provided an intelligent driving apparatus comprising a memory, a processor, and a comp