Due to the non-linearity and variability of materials

the complexity of calculation and the instability of load

the establishment of kinematics of soft actuators is still a challenging task. A machine learning approach is proposed to obtain the kinematic mapping of a soft robot in 3D space to solve the problem of the difficulty of the traditional modeling approach. A finite element model (FEM) is also used to generate a large amount of training data to ensure the reliability of the trained model. The optimal solution is obtained by comparing the optimized BP neural network and the polynomial regression model. The results show that the average position error is within 1 mm and the average pressure error is within 1.6 kPa. The machine learning approach for solving the kinematics of soft actuators is simpler and more accurate than the usual analysis methods.