Scholarlink

<p>Vision-based trajectory tracking control for high-speed wheeled mobile robots (WMRs) in complex environments faces critical challenges from multi-source noise coupling, nonholonomic motion constraints, and system uncertainties. This paper proposes an integrated control framework addressing these challenges systematically. Firstly, the kinematic model and trajectory tracking error dynamics are established to characterize system behavior under nonholonomic constraints. Secondly, a Dual GRU-enhanced KalmanNet state estimator (DG-KalmanNet) is designed, employing dual GRU architecture to separately process visual measurements and system dynamics while incorporating error state constraints to ensure physical consistency. Thirdly, an adaptive Extended State Observer (ESO) is constructed based on state estimates to provide real-time disturbance estimation, enabling a Nonsingular Terminal Sliding Mode Controller (NTSMC) to achieve robust trajectory tracking with proven global stability and finite-time convergence. Finally, comprehensive simulations demonstrate significant performance improvements over traditional PID control: 80.0% RMSE reduction, 75.4% faster convergence, and 93.2% disturbance rejection rate, validating the effectiveness of the proposed approach.</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

Vision-based trajectory tracking control for wheeled mobile robots using DG-KalmanNet and nonsingular terminal sliding mode control

  • 期刊论文

  • 发表日期：

    2025年9月10日

查看全文
  • Chenmeng Zhang,
  • Jingfeng Mao,
  • Qiuyue Qin,
  • Aihua Wu,
  • Xudong Zhang,
  • Junqiang Zheng

摘要

Vision-based trajectory tracking control for high-speed wheeled mobile robots (WMRs) in complex environments faces critical challenges from multi-source noise coupling, nonholonomic motion constraints, and system uncertainties. This paper proposes an integrated control framework addressing these challenges systematically. Firstly, the kinematic model and trajectory tracking error dynamics are established to characterize system behavior under nonholonomic constraints. Secondly, a Dual GRU-enhanced KalmanNet state estimator (DG-KalmanNet) is designed, employing dual GRU architecture to separately process visual measurements and system dynamics while incorporating error state constraints to ensure physical consistency. Thirdly, an adaptive Extended State Observer (ESO) is constructed based on state estimates to provide real-time disturbance estimation, enabling a Nonsingular Terminal Sliding Mode Controller (NTSMC) to achieve robust trajectory tracking with proven global stability and finite-time convergence. Finally, comprehensive simulations demonstrate significant performance improvements over traditional PID control: 80.0% RMSE reduction, 75.4% faster convergence, and 93.2% disturbance rejection rate, validating the effectiveness of the proposed approach.