Posture control of tensegrity manipulator based on kinematic model using kernel ridge regression

Biological bodies have numerous redundant soft muscles. In our laboratory, we pursue both the softness and redundancy of biological bodies in robots. We propose the suitability of the “tensegrity” structure for enhancing softness and redundancy in robots. Tensegrity is a structure composed of struts and cables without strut connections. This structure is soft and contains numerous actuators. Herein, we develop a tensegrity manipulator driven by 20 pneumatic cylinders with 20 springs to enhance the softness and redundancy. The robot moves the changing tensile forces of its cables using pneumatic cylinders. The complex relationship between the actuator inputs and the robot’s postures to control this robot must be modeled. In this study, the forward kinematics is modeled by machine learning. The inverse kinematics is solved numerically by deriving the Jacobi matrix from the derived forward kinematic model. This study reports on this simple kinematic model and discusses ways to improve it.