Title:	Embodied AI for Autonomous Robots
DNr:	Berzelius-2024-307
Project Type:	LiU Berzelius
Principal Investigator:	Olov Andersson <olovand@kth.se>
Affiliation:	Kungliga Tekniska högskolan
Duration:	2024-09-06 – 2025-04-01
Classification:	10207
Homepage:	http://www.kth.se/profile/olovand
Keywords:

Abstract

We will train neural network models to advance robot learning and embodied AI with real robots. Machine learning approaches are increasingly used in robotic applications whether it be for manipulation of objects (e.g., part assembly, warehouses) or navigation of autonomous vehicles and other mobile robots. Advances in large pre-trained “foundation models” for perception and planning in addition to large-language models (LLM) have lead to large improvements in learning capability by enabling robots to draw on common-sense knowledge and reason about an open set of objects it has not been trained on. Additionally, the rise of pre-trained models and multi-modal LLM adapted to robots, so called “embodied AI” that reason over robot perception and actions, are becoming an increasingly feasible alternative to modular, engineered approaches in robotics. So far most of the work in embodied AI is for static manipulation problems but here we will adapt and explore their use in autonomous mobile robots, including real-world experiments with our Spot quadruped from Boston Dynamics.