Realeasy: Real-Time capable Simulation to Reality Domain Adaptation
(2021) 2021 IEEE 17th International Conference on Automation Science and Engineering (CASE) p.2009-2016- Abstract
- We address the problem of insufficient quality of robot simulators to produce precise sensor readings for joint positions, velocities and torques. Realistic simulations of sensor readings are particularly important for real time robot control laws and for data intensive Reinforcement Learning of robot movements in simulation. We systematically construct two architectures based on Long Short-Term Memory to model the difference between simulated and real sensor readings for online and offline application. Our solution is easy to integrate into existing Robot Operating System frameworks and its formulation is neither robot nor task specific. The collected data set, the plug-and-play Realeasy model for the Panda robot and a reproducible... (More)
- We address the problem of insufficient quality of robot simulators to produce precise sensor readings for joint positions, velocities and torques. Realistic simulations of sensor readings are particularly important for real time robot control laws and for data intensive Reinforcement Learning of robot movements in simulation. We systematically construct two architectures based on Long Short-Term Memory to model the difference between simulated and real sensor readings for online and offline application. Our solution is easy to integrate into existing Robot Operating System frameworks and its formulation is neither robot nor task specific. The collected data set, the plug-and-play Realeasy model for the Panda robot and a reproducible real-time docker setup are shared alongside the code. We demonstrate robust behavior and transferability of the learned model between individual Franka Emika Panda robots. Our experiments show a reduction in torque mean squared error of at least one order of magnitude. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/record/edb5551d-e691-4c92-85db-d95acf7e9e25
- author
- Dürr, Alexander LU ; Neric, Liam ; Krueger, Volker LU and Topp, Elin Anna LU
- organization
- publishing date
- 2021
- type
- Chapter in Book/Report/Conference proceeding
- publication status
- published
- subject
- keywords
- Simulation and control, Machine learning, Model learning and control, Deep learning, Robotics, Industrial robots, LSTM Neural Network, Adaptation models, Torque, Robot sensing systems, Data models, Real-time systems, Time series models
- host publication
- 2021 IEEE 17th International Conference on Automation Science and Engineering (CASE)
- pages
- 8 pages
- publisher
- IEEE - Institute of Electrical and Electronics Engineers Inc.
- conference name
- 2021 IEEE 17th International Conference on Automation Science and Engineering (CASE)
- conference location
- Lyon, France
- conference dates
- 2021-08-23 - 2021-08-27
- external identifiers
-
- scopus:85117032558
- ISBN
- 978-1-6654-1873-7
- 978-1-6654-4809-3
- DOI
- 10.1109/CASE49439.2021.9551626
- project
- RobotLab LTH
- Robot Skill Learning based on Interactively Acquired Knowledge-based Models
- Reinforcement Learning in Continuous Spaces with Interactively Acquired Knowledge-based Models
- language
- English
- LU publication?
- yes
- additional info
- © 2021 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
- id
- edb5551d-e691-4c92-85db-d95acf7e9e25
- date added to LUP
- 2021-09-14 10:44:11
- date last changed
- 2024-03-23 09:37:26
@inproceedings{edb5551d-e691-4c92-85db-d95acf7e9e25, abstract = {{We address the problem of insufficient quality of robot simulators to produce precise sensor readings for joint positions, velocities and torques. Realistic simulations of sensor readings are particularly important for real time robot control laws and for data intensive Reinforcement Learning of robot movements in simulation. We systematically construct two architectures based on Long Short-Term Memory to model the difference between simulated and real sensor readings for online and offline application. Our solution is easy to integrate into existing Robot Operating System frameworks and its formulation is neither robot nor task specific. The collected data set, the plug-and-play Realeasy model for the Panda robot and a reproducible real-time docker setup are shared alongside the code. We demonstrate robust behavior and transferability of the learned model between individual Franka Emika Panda robots. Our experiments show a reduction in torque mean squared error of at least one order of magnitude.}}, author = {{Dürr, Alexander and Neric, Liam and Krueger, Volker and Topp, Elin Anna}}, booktitle = {{2021 IEEE 17th International Conference on Automation Science and Engineering (CASE)}}, isbn = {{978-1-6654-1873-7}}, keywords = {{Simulation and control; Machine learning; Model learning and control; Deep learning; Robotics; Industrial robots; LSTM Neural Network; Adaptation models; Torque; Robot sensing systems; Data models; Real-time systems; Time series models}}, language = {{eng}}, pages = {{2009--2016}}, publisher = {{IEEE - Institute of Electrical and Electronics Engineers Inc.}}, title = {{Realeasy: Real-Time capable Simulation to Reality Domain Adaptation}}, url = {{https://lup.lub.lu.se/search/files/108980221/CASE_PAPER_2021_Sim2Real_LSTM_final_submission.pdf}}, doi = {{10.1109/CASE49439.2021.9551626}}, year = {{2021}}, }