Title:	Geometric robot learning
DNr:	Berzelius-2026-60
Project Type:	LiU Berzelius
Principal Investigator:	Noémie Jaquier <jaquier@kth.se>
Affiliation:	Kungliga Tekniska högskolan
Duration:	2026-03-02 – 2026-10-01
Classification:	10210
Keywords:

Abstract

This project aims at developing novel machine learning techniques for learning expressive robot motion policies for manipulation. The developed robot learning algorithms will lead to sound theoretical guarantees by introducing inductive bias in the form of geometry and physics. The different project members will work on the design of visuomotor deep generative policies for learning manipulation skills with guarantees for both rigid and deformable objects, as well as on the dynamic modeling of deformable objects via physics-inspired and geometric neural networks.