Modeling of a soft robotic neck using machine learning techniques

In this paper we address the problem of modeling a soft robotic neck by using different neural network architectures, studying the influence on the results of the number of layers of each network and its corresponding activation function. The Tangent Hyperbolic Tangent (TANH) and Exponential Linear Unit (ELU) activation functions are used. The obtained models are compared with a Multi-Layer Perceptron (MLP) with optimized parameters, as well as with the kinematic model of the neck. The experimental results demonstrate the advantage of using machine learning techniques for modeling highly nonlinear systems such as this soft robotic neck, whose elastic characteristics make it difficult to formulate a robust analytical model.