Shape Estimation of Continuum Arms Based on Inverse Kinematics Data-Driven Modeling

In recent decades, soft robotic arms inspired by biological organisms with flexible structures have enabled significant advancements in robotics, particularly by allowing robots to navigate in complex and unknown environments. Tendon-driven continuum robotic arms, as a type of such systems, are characterized by their infinite degrees of freedom and the inherent complexity of their nonlinear systems, while the analysis of their configurations is highly challenging. In this study, a data-driven inverse kinematics model is developed to estimate the shape of a two-segment continuum robot called RoboArm in ARAS Lab. The robot's final configuration was recorded using two cameras, and the shape estimation was performed by tracking the position of each segment's end along a given dynamic path. The experimental results were compared with software simulations which demonstrate high accuracy in predicting the arm's configuration. By comparing the estimated shapes from the simulations with the robot's actual shape during movement, we validated the effectiveness of this approach for accurately modeling the robot's configuration in real-time.