Scholarlink

<p>High-precision and robust path tracking is a critical challenge for autonomous vehicles. To address insufficient accuracy and poor robustness in path tracking control, this paper proposes an innovative method based on Active Disturbance Rejection Control (ADRC) optimized by a Radial Basis Function Neural Network (RBFNN). The proposed controller utilizes an Extended State Observer (ESO) to estimate and compensate for internal model uncertainties and external disturbances in real-time. Notably, the RBFNN performs online adaptive tuning of the ESO’s key gain parameters, superseding the challenging parameter tuning and addressing the limited adaptability of conventional ADRC. The effectiveness of the proposed RBFNN-ADRC was validated through comprehensive co-simulations on the CarSim-Simulink platform under typical and challenging operating conditions. Simulation results demonstrate that at 18&#xa0;km/h and 36&#xa0;km/h, the RBFNN-ADRC significantly outperforms conventional PID, Model Predictive Control (MPC), and fixed-parameter ADRC. Specifically, at 36&#xa0;km/h, the maximum lateral tracking error was reduced to only 0.103&#xa0;m, representing improvements of 70.7%, 68.8%, and 48.5% over PID, MPC, and standard ADRC, respectively. These findings confirm the proposed method’s effectiveness in enhancing tracking accuracy and robustness, providing a valuable reference for the design of advanced control systems for autonomous vehicles.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

An RBFNN-Optimized Active Disturbance Rejection Controller for Path Tracking of Autonomous Vehicles

  • 期刊论文

  • 发表日期：

    2026年1月5日

查看全文
  • Honggang Li,
  • Yurun Li,
  • Pengfei Du,
  • Caixia Song,
  • Zhichao Ming

摘要

High-precision and robust path tracking is a critical challenge for autonomous vehicles. To address insufficient accuracy and poor robustness in path tracking control, this paper proposes an innovative method based on Active Disturbance Rejection Control (ADRC) optimized by a Radial Basis Function Neural Network (RBFNN). The proposed controller utilizes an Extended State Observer (ESO) to estimate and compensate for internal model uncertainties and external disturbances in real-time. Notably, the RBFNN performs online adaptive tuning of the ESO’s key gain parameters, superseding the challenging parameter tuning and addressing the limited adaptability of conventional ADRC. The effectiveness of the proposed RBFNN-ADRC was validated through comprehensive co-simulations on the CarSim-Simulink platform under typical and challenging operating conditions. Simulation results demonstrate that at 18 km/h and 36 km/h, the RBFNN-ADRC significantly outperforms conventional PID, Model Predictive Control (MPC), and fixed-parameter ADRC. Specifically, at 36 km/h, the maximum lateral tracking error was reduced to only 0.103 m, representing improvements of 70.7%, 68.8%, and 48.5% over PID, MPC, and standard ADRC, respectively. These findings confirm the proposed method’s effectiveness in enhancing tracking accuracy and robustness, providing a valuable reference for the design of advanced control systems for autonomous vehicles.