3D printed soft robotic arm with integrated tool-transfer channels enables diverse interventional functions

Compared with rigid robotic arms, soft robotic arms have shown better continuous deformation capability, environmental adaptability, and flexible contact characteristics. Soft robotic arms can realize a variety of complex postures through active deformation, and can adaptively intervene in complex and narrow spaces for target tasks, increasingly demonstrating valuable potential in real-world applications. However, insufficient capability for multifunctional integration and a lack of complete motion perception and control methods severely hinder the practical applicability of existing soft robotic arms. Here, we report a multi-segmented soft robotic arm with integrated stiffness-tunable function, end-effector high-precision deformation function, internal tool-transfer function, multi-degree-of-freedom flexibility function, and deformation self-sensing function. Using the handle shank, we achieve deep intervention of continuously bending narrow pipes, extraction of waste liquid, and laser sintering and cutting inside the pipes. Afterward, we demonstrate the tool-transfer functions in the situations where the soft robotic arm with a smaller diameter cannot integrate all the required tools. Using flexible sensors, we realize the 3D visualization function of soft robotic arm. We use a long short-term memory network (LSTM) to implement the sensory perception of the end trajectories of the soft robotic arm. By combining the established inverse kinematics models with LSTM neural network and a feedback controller, we achieve the trajectory tracking verification for perception-trained soft robotic arms that shows good agreement between calculated and actual motion trajectories.