Research on high-precision control algorithm for robots based on deep neural networks
Year: 2025
Type: conference-paper
Abstract: This paper presents an adaptive sliding mode control (SMC) strategy enhanced by reinforcement learning for continuum robots in space debris removal missions. A data-driven neural network model first learns the system's nonlinear dynamics with high prediction accuracy (±1% error). Then, a reinforcement learning agent based on the Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithm dynamically optimizes the SMC parameters online. Compared to conventional fixed-parameter SMC, our method demonstrates superior tracking performance with lower overshoot, faster convergence, and enhanced robustness against disturbances. Simulations confirm the effectiveness of this approach for high-precision control of continuum manipulators in challenging space environments.
Author: Jiefei Chen
Institutions: Shanghai Open University, Shanghai Jingan District College
Language: English
Subfield: Control and Systems Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: gold
OpenAlex ID: W7126441909