Condensed FEM Model Learning

Project Description

This plugin for the open-source simulation framework SOFA contains components for learning a condensed FEM model from a soft robot SOFA scene. Authors also provide an implementation for leveraging the learned model for control, embedded control, calibration and design optimization applications.

GitHub repository

The Finite Element Method (FEM) is a powerful modeling tool for predicting the behavior of soft robots. However, its computation time is a limitation when considering robotics applications. T. Navez, E. Ménager et al. propose a learning-based approach based on condensation of the FEM model for quickly handling all kind of constraints and in particular contacts. This plugin is envisioned as a general framework for modeling, control and design of soft robots based on a condensed FEM model. It provides a platform for generating simulation data using SOFA, training Neural Networks for predicting the condensed FEM Model, as well as for leveraging it in several applications: Inverse Control, Embedded Control and Design Optimization. Moreover, several examples are available for illustrating all aforementioned applications.

Related publications

Direct and inverse modeling of soft robots by learning a condensed FEM model

Back to the plugin list

Related Projects

Back to Top