@article{Deisenroth:2014:10.1109/TPAMI.2013.218,
author = {Deisenroth, MP and Fox, D and Rasmussen, CE},
doi = {10.1109/TPAMI.2013.218},
journal = {IEEE Transactions on Pattern Analysis and Machine Intelligence},
title = {Gaussian Processes for Data-Efficient Learning in Robotics and Control},
url = {http://dx.doi.org/10.1109/TPAMI.2013.218},
year = {2014}
}

Download

TY  - JOUR
AB  - Autonomous learning has been a promising direction in control and robotics for more than a decade since data-drivenlearning allows to reduce the amount of engineering knowledge, which is otherwise required. However, autonomous reinforcementlearning (RL) approaches typically require many interactions with the system to learn controllers, which is a practical limitation in realsystems, such as robots, where many interactions can be impractical and time consuming. To address this problem, current learningapproaches typically require task-specific knowledge in form of expert demonstrations, realistic simulators, pre-shaped policies, orspecific knowledge about the underlying dynamics. In this article, we follow a different approach and speed up learning by extractingmore information from data. In particular, we learn a probabilistic, non-parametric Gaussian process transition model of the system.By explicitly incorporating model uncertainty into long-term planning and controller learning our approach reduces the effects of modelerrors, a key problem in model-based learning. Compared to state-of-the art RL our model-based policy search method achieves anunprecedented speed of learning. We demonstrate its applicability to autonomous learning in real robot and control tasks.
AU  - Deisenroth,MP
AU  - Fox,D
AU  - Rasmussen,CE
DO  - 10.1109/TPAMI.2013.218
PY  - 2014///
SN  - 0162-8828
TI  - Gaussian Processes for Data-Efficient Learning in Robotics and Control
T2  - IEEE Transactions on Pattern Analysis and Machine Intelligence
UR  - http://dx.doi.org/10.1109/TPAMI.2013.218
UR  - http://hdl.handle.net/10044/1/12277
ER  - 

Download