Adaptive Hybrid Visual Servo Control with Dynamic Smoothing for Enhanced Robotic Manipulator Control

This paper presents a comparative study of eight visual servoing control schemes for robotic manipulation involving complex star-shaped object tracking. Three main approaches—Image-Based Visual Servoing (IBVS), Position-Based Visual Servoing (PBVS), and hybrid methods are analysed under fixed and adaptive gain configurations. Novel techniques include smoothing, adaptive blending, and damped least squares (DLS) inversion. Experiments were conducted in MATLAB using a 6-DOF PUMA-560 manipulator with an eye-in-hand camera tracking a 16-point asymmetric star-shaped target. Evaluation metrics covered convergence, computational efficiency, positioning accuracy, trajectory smoothness, and stability. Results show that the hybrid adaptive smoothing method outperforms others, achieving 23 − 31% faster convergence, lowest final error (0.0008 m vs. 0.0015–0.0025 m), 45% lower velocity variation, 38% fewer control discontinuities, and a 35% lower condition number, while maintaining computation time within 5% of fixed-gain schemes. The improvement stems from sigmoidbased adaptive blending—transitioning from IBVS (adaptive smoothing factor − 𝛼 = 0 ) under large errors to PBVS (𝛼 > 0.8) near convergence and exponential moving average filtering of control velocities. These results advance adaptive visual servoing for high-precision industrial automation.