Calibration model training method and control method based on vehicle lateral and longitudinal coupling relationship

By constructing a calibration model training method based on the vehicle's lateral and longitudinal coupling relationship, the problem of low lateral and longitudinal calibration accuracy in autonomous vehicles is solved, achieving higher control accuracy and adaptability.

CN122087448APending Publication Date: 2026-05-26ARTIFICIAL INTELLIGENCE RES INST OF HEFEI COMPREHENSIVE NAT SCI CENT (ANHUI ARTIFICIAL INTELLIGENCE LAB)
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ARTIFICIAL INTELLIGENCE RES INST OF HEFEI COMPREHENSIVE NAT SCI CENT (ANHUI ARTIFICIAL INTELLIGENCE LAB)
Filing Date
2025-12-30
Publication Date
2026-05-26

AI Technical Summary

Technical Problem

In existing technologies, the calibration methods for the lateral and longitudinal control variables of autonomous vehicles fail to effectively consider the dynamic coupling relationship between the lateral and longitudinal directions, resulting in low calibration accuracy and poor adaptability to complex working conditions, thus reducing control precision.

Method used

A calibration model training method based on the lateral and longitudinal coupling relationship of vehicles is adopted. By acquiring a training sample dataset, a feature extraction network and a calibration network group are used for training. A coupling physical constraint loss function and a consistency loss function are constructed to output the predicted control command error, thereby improving the accuracy of the calibration model.

Benefits of technology

It improves the control precision of autonomous vehicles, enabling them to respond flexibly to different operating conditions and enhancing the adaptability and accuracy of the calibration model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122087448A_ABST
    Figure CN122087448A_ABST
Patent Text Reader

Abstract

This disclosure provides a calibration model training method and control method based on the lateral and longitudinal coupling relationship of a vehicle, which can be applied to the field of autonomous driving technology. The training method includes: acquiring a training sample dataset; training a feature extraction network and a calibration network group using the training sample dataset to obtain a target feature extraction network and a target calibration network group; training a calibration error network using the training sample dataset to obtain a target calibration error network, which is used to output the predicted control command error based on the coupling relationship between longitudinal acceleration and lateral front wheel steering angle; and obtaining a calibration model based on the target feature extraction network, the target calibration network group, and the target calibration error network. This disclosure does not focus on the tracking control algorithm itself, but only calibrates the relationship between the algorithm's output and the vehicle's execution, and considers the lateral and longitudinal coupling relationship during calibration.
Need to check novelty before this filing date? Find Prior Art