A Modeling and Data-Driven Control Framework for Rigid-Soft Hybrid Robot With Visual Servoing

In this letter, a rigid-soft hybrid robot with visual servoing is designed to improve robotic properties of accuracy and safe interaction, where the hybrid robot is connected by a soft robot and six degrees of freedom rigid robot in series. The series structure of rigid and soft parts brings the coupling and complexity to modeling. For synthesis modeling, we develop a state-space equation of the hybrid robot by the combination of the rigid and soft parts, where the rigid model is built by differential kinematics, and Piecewise Constant-Curvature and Lagrange equations are used to model the soft part with consideration of the elastic deformation caused by gravity loading. Based on the synthesis model, a model predictive controller is designed to eliminate the error between the hybrid robot terminal and the target position. Considering the imprecise modeling process disturbs the control performance, a data-driven strategy named input mapping is incorporated into the predictive controller to reduce the inaccuracy model effect and improve the control performance. In addition, the feedback error of the hybrid robot is transferred in the robot terminal coordinate system to avoid hand-eye calibration work. Finally, the experiments on a rigid-soft hybrid robot show the proposed method converges nearly twice as fast as the model predictive control method and two state-of-the-art motion planning methods for the visual servoing of the soft-rigid hybrid robot system.