Robust Visual Servoing for Quadruped Robots: ISS-Based Target Tracking With Secondary Formation Objectives

This work presents the design, implementation, and experimental validation of a visual servo control scheme formulated at the task-space level and applied to a quadruped robot equipped with an active two degree of freedom (2DoF) vision system for continuous tracking of moving targets. The proposed controller is based on a differential kinematic model that maps the robot's task-space velocity references to the kinematic evolution of the active vision system, incorporating a target velocity estimator that enables dynamic compensation of the target motion projected onto the image plane. The control architecture exploits the kinematic redundancy of the robotic system to simultaneously address multiple control objectives within a hierarchical framework. The primary task ensures target centering in the image plane, thereby maintaining visibility within the robot's field of view (FoV), while the secondary task regulates relative formation by controlling the distance and orientation between the robot and the target. Quadruped locomotion is handled by a low-level gait controller and is not modified by the proposed scheme. Closed-loop stability is rigorously analyzed using Lyapunov theory, demonstrating Input-to-State Stability (ISS) properties in the presence of bounded disturbances. Simulation and real-world experimental results obtained with the Unitree Go2 robot confirm the robustness and stability of the proposed approach for dynamic visual tracking in realistic scenarios.