Adaptive dynamic programming for robust path tracking in an agricultural robot using critic neural networks
Alireza Azimi | Redmond R. Shamshiri | Aliakbar Ghasemzadeh
Trajectory tracking control for agricultural mobile robots poses unique challenges due to inherent non-holonomic constraints and external disturbances, which can cause deviations from the desired path, affecting the robot‘s performance and operational efficiency. This paper presents an advanced learning-based control framework for robust path tracking in agricultural robots with Ackermann-steering mechanisms. Using Adaptive Dynamic Programming (ADP) and a Critic Neural Network, the proposed method handles external disturbances, including wheel slippage, which is common in agricultural environments. The Critic Neural Network the Hamilton-Jacobi-Isaacs (HJI) equation, allowing the controller to learn the near-optimal control policy in real time and adapt to environmental disturbances. The critic network‘s weights are updated online through an adaptive law, ensuring continuous learning and adaptation throughout the operation. Furthermore, the paper presents comprehensive simulation studies to evaluate the effectiveness of the proposed framework. The results demonstrate significant improvements in trajectory tracking performance compared to existing control methods, particularly in scenarios with substantial uncertainties and disturbances.
Показать больше [+] Меньше [-]Библиографическая информация
Эту запись предоставил Directory of Open Access Journals
