Path Generation for Articulated Robots Using Neural Networks Trained with Random Sampling Methods

Abstract

We propose a method that combines random sampling and machine learning for path generation of articulated robots in environments with obstacles. This method enables fast collection and training of data, as well as rapid and reproducible path inference, by training paths obtained through random sampling while dividing them into multi-layer perceptrons connected in multiple stages. Additionally, the effectiveness of this method is demonstrated through simulations.