tsta

Home / سی و دومین کنفرانس ملی و دهمین کنفرانس بین المللی مهندسی زیست پزشکی ایران

Deep Neural Network–Based Adaptive Global Logarithmic Sliding Mode Control for Lower-Limb Rehabilitation Exoskeletons

Authors :

Masoud Shirzadeh¹ Ghoncheh Zand² Samim Kamyab³

1- Department of Electronics, Amirkabir University of Technology 2- Amirkabir University of Technology 3- College of Interdisciplinary Science and Technology University of Tehran

Keywords :

Deep neural networks،Global logarithmic sliding mode control،Rehabilitation exoskeletons

Abstract :

Lower-limb rehabilitation exoskeletons have emerged as promising tools to help people with lower-limb disorders. However, designing effective controllers remains challenging due to the coupled nonlinear dynamics, system uncertainties, and external disturbances. While classical Sliding Mode Control (SMC) is robust, it suffers from chattering, which can decrease system functionality and performance. To overcome these limitations, we propose a novel control framework, Deep Neural Network (DNN)-Global Logarithmic (GLog)-SMC, which combines GLog-SMC with an adaptive DNN. The GLog-SMC helps the system reach the sliding surface smoothly and minimizes chattering, while the adaptive DNN approximates unknown system dynamics online. A Lyapunov-based adaptation law is used to ensure closed-loop stability. Results on a three-degree-of-freedom exoskeleton show that the suggested approach improves trajectory tracking accuracy, increases robustness, and produces smoother control torques compared to classical SMC.