Open Deep Learning Toolkit for Robotics (OpenDR)
WebsiteProject type
EU Horizon 2020 program, call H2020-ICT-2018-2020 (Information and Communication Technologies); 2019 – 2022
Abstract
The aim of OpenDR is to develop a modular, open and non-proprietary deep learning toolkit for robotics. We will provide a set of software functions, packages and utilities to help roboticists develop and test robotic applications that incorporate deep learning. OpenDR will enable linking robotics applications to software libraries such as tensorflow and the ROS operating environment. We focus on the AI and cognition core technology in order to give robotic systems the ability to interact with people and environments by means of deep-learning methods for active perception, cognition and decisions making. OpenDR will enlarge the range of robotics applications making use of deep learning, which will be demonstrated in the applications areas of healthcare, agri-food and agile production.
TUD project members
ir. Bas van der Heijden, ir. Jelle Luijkx, Dr. Osama Mazhar, Dr. Laura Ferranti, Dr.-Ing. Jens Kober, prof.dr. Robert Babuška,
Project consortium
Aristotle University of Thessaloniki, Tampere University of Technology, Aarhus University, Delft University of Technology, Albert-Ludwigs-Universität Freiburg, Cyberbotics Ltd., PAL Robotics S.L., Agro Intelligence ApS
Publications with videos
Efficient Parallelized Simulation of Cyber-Physical Systems. Transactions on Machine Learning Research, 2024. Reproducibility Certification. .
PARTNR: Pick and place Ambiguity Resolving by Trustworthy iNteractive leaRning. In NeurIPS 2022 - 5th Robot Learning Workshop: Trustworthy Robotics, 2022. .
A Deep Learning Framework for Recognizing Both Static and Dynamic Gestures. Sensors, 21(6), 2021. .
Publications without videos
ExploRLLM: Guiding Exploration in Reinforcement Learning with Large Language Models. arXiv:2403.09583 [cs.RO], 2024. .
Engine Agnostic Graph Environments for Robotics (EAGERx): A Graph-Based Framework for Sim2real Robot Learning. IEEE Robotics & Automation Magazine, :2-15, 2024. Early Access. .
Prioritizing States with Action Sensitive Return in Experience Replay. In Sixteenth European Workshop on Reinforcement Learning (EWRL), 2023. .
OpenDR: An Open Toolkit for Enabling High Performance, Low Footprint Deep Learning for Robotics. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 12479–12484, 2022. .
GEM: Glare or Gloom, I Can Still See You – End-to-End Multimodal Object Detector. IEEE Robotics and Automation Letters, 6(4):6321–6328, 2021. The contents of this paper were also selected by IROS'21 Program Committee for presentation at the Conference. .
Random Shadows and Highlights: A New Data Augmentation Method for Extreme Lighting Conditions. arXiv:2101.05361 [cs.CV], 2021. .
DeepKoCo: Efficient Latent Planning with an Invariant Koopman Representation. In IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pp. 183–189, 2021. .