CN-122015878-A - Intelligent device, track planning method thereof and storage medium

CN122015878ACN 122015878 ACN122015878 ACN 122015878ACN-122015878-A

Abstract

The application relates to the technical field of intelligent driving, in particular to intelligent equipment, a track planning method thereof and a storage medium, and aims to solve the problem of how to improve the reliability and stability of a planned track obtained by using a track planning model. The method comprises the steps of obtaining target information for track planning, carrying out track planning on the target information by adopting a track planning model to generate a running track, wherein the track planning model is obtained by carrying out multi-round training on a model to be trained according to an information sample until the running track generated by the model to be trained meets track quality requirements, carrying out quality assessment on the running track generated by the model to be trained based on the track quality requirements after the training of each round is completed, and taking an assessment result and the information sample as input data of the next round of training, wherein the assessment result is used for guiding the model to be trained to adjust a training strategy of the next round of training. Based on the method, stable and reliable intelligent equipment running track can be obtained.

Inventors

REN SHAOQING
WANG ZHUO
XIANG ZHENZHEN
XU NING
SONG YU
YE CHAOQIANG
ZENG CHAO
YANG JIN
WANG CHENGFA
LIU GUOYI

Assignees

安徽蔚来智驾科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (10)

1. A track planning method for an intelligent device, the method comprising: The method comprises the steps of obtaining target information for track planning, wherein the target information at least comprises environmental information of an environment where intelligent equipment is located; wherein the trajectory planning model is obtained by: Acquiring an information sample of the target information, and performing multi-round training on a model to be trained according to the information sample until a running track generated by performing track planning on the information sample by the model to be trained meets a preset track quality requirement; For each round of training, after the round of training is completed, based on the track quality requirement, carrying out quality evaluation on the running track generated by the model to be trained in the round of training, obtaining an evaluation result, and taking the evaluation result and the information sample as input data of the next round of training, wherein the evaluation result is used for guiding the model to be trained to adjust a training strategy of the next round of training.
2. The method of claim 1, wherein the step of determining the position of the substrate comprises, The model to be trained is a model which is pre-trained in the completed track planning.
3. The method of claim 2, wherein the training the model to be trained in multiple rounds comprises training the model to be trained in multiple rounds of reinforcement learning; Wherein the training strategy comprises a hyper-parameter of a reward function employed by reinforcement learning training.
4. The method of claim 1, wherein the performing quality evaluation on the running track generated by the training of the model to be trained in the present round and obtaining the evaluation result includes: Determining an evaluation index of the track quality requirement; Acquiring the numerical value of the evaluation index according to the running track generated by the model to be trained; And acquiring the evaluation result according to the numerical value of the evaluation index.
5. The method of claim 1, wherein the preset track quality requirement is a plurality of, and the performing quality evaluation on the running track generated by the training of the model to be trained and obtaining the evaluation result includes: Respectively carrying out quality evaluation on the running tracks generated by the present training according to the quality requirements of each track to obtain a first evaluation result of the quality requirements of each track; And encoding the first evaluation results of all the track quality requirements to form a second evaluation result in a vector form.
6. The method of claim 5, wherein the preset track quality requirements include any or all of the following track quality requirements: the instruction follows the consistency requirement, and the corresponding quality requirement is that the running track generated by the track planning model is consistent with the navigation reference track of the intelligent equipment; The collision risk requirement, the corresponding quality requirement is that the running track generated by the track planning model meets the preset safety constraint condition; The corresponding quality requirement of the kinematic constraint requirement is that the running track generated by the track planning model meets the preset kinematic constraint condition; The track stability requirement, its corresponding quality requirement is that the travel track that track planning model produced satisfies the stability constraint condition of predetermineeing.
7. The method of claim 1, wherein the step of determining the position of the substrate comprises, The intelligent equipment is a vehicle, the target information further comprises task information of a task to which the track planning belongs, and the task at least comprises parking.
8. The method of claim 1, wherein the step of determining the position of the substrate comprises, The trajectory planning model employs an end-to-end trajectory planning model.
9. An intelligent device, comprising: At least one processor; And a memory communicatively coupled to the at least one processor; wherein the memory has stored therein a computer program which, when executed by the at least one processor, implements the trajectory planning method of the smart device of any one of claims 1 to 8.
10. A computer readable storage medium, in which a plurality of program codes are stored, characterized in that the program codes are adapted to be loaded and run by a processor to perform the trajectory planning method of a smart device according to any one of claims 1 to 8.

Description

Intelligent device, track planning method thereof and storage medium Technical Field The application relates to the technical field of intelligent driving, and particularly provides intelligent equipment, a track planning method thereof and a storage medium. Background When intelligent driving control is performed on a vehicle, a track planning model is usually deployed on the vehicle, state information, environment information and the like of the vehicle are input into the track planning model to be processed so as to obtain a planned track of the vehicle, and then the vehicle is controlled to run according to the planned track. In order to ensure safe driving of the vehicle, a stable and reliable driving track is required to be obtained by the track planning model, but in some special scenes, the track planning model may be sporadically abnormal, and an abnormal driving track is planned. For example, in a parking scenario, the trajectory planning model may output a planned trajectory that traverses an obstacle (e.g., through a wall) or curves around a large curve, etc. In addition, the track stability of the track planning model output may also be deteriorated when the environmental information changes, thereby affecting the safe running of the vehicle. When the planned track output by the track planning model has the problem of abnormality or poor stability, the conventional processing method at present mainly carries out offline optimization on the track planning model, or adds post-processing after the output of the track planning model, and optimizes the track output by the track planning model through the post-processing. The offline optimization may be enriching training data or adjusting a training method, and training the trajectory planning model again by using new training data or training method to optimize model parameters of the trajectory planning model. But is limited by the richness of training data, the effectiveness of training methods and post-processing, even if the model parameters or the output tracks of the track planning model are optimized by adopting the method, the track planning model can still output abnormal or poorly stable planning tracks when facing complex scenes, and the safe running of the vehicle is influenced. Accordingly, there is a need in the art for a new solution to the above-mentioned problems. Disclosure of Invention The present application has been made to overcome the above drawbacks, and aims to solve or at least partially solve the technical problem of how to improve the reliability and stability of a planned trajectory obtained by using a trajectory planning model, thereby improving the safety and stability of a vehicle when driving according to the planned trajectory. In a first aspect, the present application provides a track planning method for an intelligent device, the method comprising: The method comprises the steps of obtaining target information for track planning, wherein the target information at least comprises environmental information of an environment where intelligent equipment is located; wherein the trajectory planning model is obtained by: Acquiring an information sample of the target information, and performing multi-round training on a model to be trained according to the information sample until a running track generated by performing track planning on the information sample by the model to be trained meets a preset track quality requirement; For each round of training, after the round of training is completed, based on the track quality requirement, carrying out quality evaluation on the running track generated by the model to be trained in the round of training, obtaining an evaluation result, and taking the evaluation result and the information sample as input data of the next round of training, wherein the evaluation result is used for guiding the model to be trained to adjust a training strategy of the next round of training. In one technical scheme of the track planning method, the model to be trained is a model which is pre-trained in the track planning. In one technical scheme of the track planning method, the training of the model to be trained for multiple rounds comprises the steps of performing reinforcement learning training on the model to be trained for multiple rounds; Wherein the training strategy comprises a hyper-parameter of a reward function employed by reinforcement learning training. In one technical scheme of the track planning method, the performing quality evaluation on the running track generated by the training of the model to be trained in the round of training and obtaining an evaluation result includes: Determining an evaluation index of the track quality requirement; Acquiring the numerical value of the evaluation index according to the running track generated by the model to be trained; And acquiring the evaluation result according to the numerical value of the evaluation index. In one technical scheme