ROBOTICS & MACHINE LEARNING

Daniel Dugas

I did my PhD in Machine Learning and Robotics from 2018 to 2022 at the ETH Zürich, under Prof. Siegwart.

I'm interested in embodied AI - How can robotics help us better understand the ingredients of intelligence? My main research question during the PhD was how algorithms can create powerful, predictive models of the world from unlabeled data, and use them in decision making. I'm willing to bet that big breakthroughs will result from 1) deeper understanding of large fundamental models, 2) better world-models, and 3) research on integrating high-level reasoning (for example large multi-modal models) into low level systems (Perception, SLAM, Controls).

I now work at mesh.ch, where I apply robotics and ML research to help make construction less material and labor intensive.

Publications

PhD Thesis: Do Androids Dream of Interactive Navigation?

Daniel Dugas

ETH Zürich Research Collection

NavDreams: Towards Camera-Only RL Navigation Among Humans

Daniel Dugas, Olov Andersson, Roland Siegwart, and Jen Jen Chung

IEEE International Conference on Intelligent Robots and Systems (IROS), 2022

NavRep: Unsupervised Representations for Reinforcement Learning of Robot Navigation in Dynamic Human Environments

Daniel Dugas, Juan Nieto, Roland Siegwart, Jen Jen Chung

IEEE International Conference on Robotics and Automation (ICRA), 2021

IAN: Multi-Behavior Navigation Planning for Robots in Real, Crowded Environments

Daniel Dugas, Juan Nieto, Roland Siegwart, Jen Jen Chung

IEEE International Conference on Intelligent Robots and Systems (IROS), 2020

(co-author) SegMap: Segment-based mapping and localization using data-driven descriptors

R. Dubé, A. Cramariuc, D. Dugas, H. Sommer, M. Dymczyk, J. Nieto, R. Siegwart, and C. Cadena

The International Journal of Robotics Research (IJRR), 2019

More publications...

Projects