Abstract
We present TrueAdapt, a model-free method to learn online adaptations of robot trajectories based on their effects on the environment. Given sensory feedback and future waypoints of the original trajectory, a neural network is trained to predict joint accelerations at regular intervals. The adapted trajectory is generated by linear interpolation of the predicted accelerations, leading to continuously differentiable joint velocities and positions. Bounded jerks, accelerations and velocities are guaranteed by calculating the valid acceleration range at each decision step and clipping the network's output accordingly. A deviation penalty during the training process causes the adapted trajectory to follow the original one. Smooth movements are encouraged by penalizing high accelerations and jerks. We evaluate our approach by training a simulated KUKA iiwa robot to balance a ball on a plate while moving and demonstrate that the balancing policy can be directly transferred to a real robot with little impact on performance.
Original language | English |
---|---|
Title of host publication | IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2020 |
Publisher | Institute of Electrical and Electronics Engineers (IEEE) |
Pages | 5387-5394 |
Number of pages | 8 |
ISBN (Electronic) | 978-1-7281-6212-6 |
ISBN (Print) | 978-1-7281-6213-3 |
DOIs | |
Publication status | Published - 1 Mar 2021 |
Event | 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2020) - Las Vegas, United States Duration: 25 Oct 2020 → 29 Oct 2020 |
Publication series
Name | Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems |
---|---|
Publisher | IEEE |
ISSN (Print) | 2153-0858 |
ISSN (Electronic) | 2153-0858 |
Conference
Conference | 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS 2020) |
---|---|
Abbreviated title | IROS 2020 |
Country/Territory | United States |
City | Las Vegas |
Period | 25/10/20 → 29/10/20 |
Bibliographical note
This research was supported by the German FederalMinistry of Education and Research (BMBF) and the IndoGerman Science & Technology Centre (IGSTC) as part of the
project TransLearn (01DQ19007A). We would like to thank
Tamim Asfour for his valuable input and helpful advice.
All real-world experiments were performed at the KUKA
Robot Learning Lab at KIT [23]. Special thanks to Wolfgang
Wiedmeyer for his tremendous effort in building up the lab.
Keywords
- cs.RO