An Adaptive Finite-Time Convergent Neural Network With Enhanced Joint Limit Constraints for 3-DoF Visual Servoing of Rigid-Flexible Coupling Robots

Recently, researchers have employed visual servoing to automate rigid-flexible coupling robots and develop a quadratic programming (QP) framework to manage their behavior, thereby demonstrating great potential in industry and other fields. However, fast real-time solutions for such redundant systems remain challenging. Therefore, we propose a novel adaptive finite-time convergent neural network (AFTCNN) to address time-varying redundancy problems with joint limit constraints, which is successfully applied to visual servoing of automated rigid-flexible coupling robots (VSARFCR). First, a QP-based 3-DoF visual tracking control scheme that considers tracking error, joint limit constraints, and joint-drifting issues is constructed based on the VSARFCR kinematics model. A new enhanced joint limit handling technology is specifically designed for VSARFCR with a prismatic joint. Then, a novel AFTCNN is proposed to address the aforementioned QP-based control scheme with various objectives and constraints. Next, the finite-time convergence of the proposed AFTCNN is rigorously proven using Lyapunov theory. Finally, simulations and physical experiments demonstrate the efficacy of the proposed control scheme for 3-DoF visual tracking under diverse constraints and objectives. Compared with existing methods, the proposed AFTCNN exhibits faster convergence and improved solution accuracy for redundancy resolution in VSARFCR.