Gaussian Processes for Data-Efficient Learning in Robotics and Control
- Submitting institution
-
University College London
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 14716
- Type
- D - Journal article
- DOI
-
10.1109/TPAMI.2013.218
- Title of journal
- IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE
- Article number
- 2
- First page
- 408
- Volume
- 37
- Issue
- 2
- ISSN
- 0162-8828
- Open access status
- Out of scope for open access requirements
- Month of publication
- February
- Year of publication
- 2015
- URL
-
-
- Supplementary information
-
-
- Request cross-referral to
- -
- Output has been delayed by COVID-19
- No
- COVID-19 affected output statement
- -
- Forensic science
- No
- Criminology
- No
- Interdisciplinary
- No
- Number of additional authors
-
2
- Research group(s)
-
-
- Citation count
- 161
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Published at top-tier journal; changed playing field of reinforcement learning as it can be applied to real systems, e.g., robots and mechanical systems. Key finding: probabilistic modelling and inference facilitate fully autonomous learning of predictive models and controllers with very small data sets. Software freely available with about 4000 downloads (https://mloss.org/software/view/508/, https://github.com/ICL-SML/pilco-matlab). Follow-on work by Bosch GmbH (contact: Bastian Bischoff, bastian.bischoff@googlemail.com) for controlling throttle valves in combustion engines (http://www.ecmlpkdd2013.org/wp-content/uploads/2013/07/153.pdf), automatic tuning of PID controllers (https://arxiv.org/abs/1703.02899), stability of Gaussian process controllers (https://www.jmlr.org/papers/volume18/16-590/16-590.pdf). UK-based start-up PROWLER.io’s initial business model was based on this work.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -