A Shape Reconstruction Method for Space Continuum Robot Based on Neural Network

Continuum robots have the characteristics of degrees of freedom and flexible shapes, and can complete space operation tasks by avoiding obstacles in narrow space. The closed-loop control and accurate arrival of continuum robot are depended on highly accurate form perception. Firstly, the rope-driven continuum robot is designed and its kinematics model is established. Then, a shape reconstruction method based on BP neural network is proposed, and an evaluation index of shape fitting accuracy considering global information is designed. Finally, a space simulation experiment is carried out in two-dimensional ground microgravity environment, the error between the shape reconstructed by the neural network and the model is compared and analyzed. The average relative distance error of the reconstructed shape of the BP neural network is 0.69%, and the maximum relative distance error is 2.17%. This proves the effectiveness of the proposed shape reconstruction method and its potential application in real space tasks.