An End-to-End Proprioception Framework for Soft Continuum Robot

Soft robots are gaining more and more attention owing to their inherent compliance and excellent flexibility, enabling new potentials for robots to address real-world chal-lenges in the manner of their natural counterparts. The soft con-tinuum robot is one typical example that reflects the successful inspiration achievement from biomimetics to robotics. However, due to the nonlinearity and ultra-high degrees of freedom of soft continuum robots, traditional kinematic modeling methods obtain unsatisfying precision, especially when robots interact with their surroundings. To precisely estimate the kinematic model of soft continuum robots, we propose a novel end-to-end proprioception method. Our method inputs time series data from the soft continuum robot actuators and outputs the end-tip position in three dimensions. The experimental results show that the proposed method can improve the kinematic accuracy compared to the constant curvature model-based method either with contact or in noncontact conditions.