One shot learning and generation of dexterous grasps for novel objects
- Submitting institution
-
The University of Birmingham
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 24113127
- Type
- D - Journal article
- DOI
-
10.1177/0278364915594244
- Title of journal
- The International Journal of Robotics Research
- Article number
- -
- First page
- 959
- Volume
- 35
- Issue
- 8
- ISSN
- 0278-3649
- Open access status
- Out of scope for open access requirements
- Month of publication
- September
- 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
-
5
- Research group(s)
-
-
- Citation count
- 45
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- The paper proposes a novel approach to dexterous grasping which addresses the challenge of achieving generalisation of grasps across object categories with high degrees-of-freedom hands. The main novelty comes from one-shot learning of a contact model and a hand configuration model, and recombining them when inferring a new grasp. This is significant as it achieves generalisability to novel objects and novel grasp types as confirmed by extensive experimental evaluation. This research has contributed to a successful grant proposal (EU-CHIST-ERA) and the main researcher (PhD student) now works for a major UK industry research unit (Dyson).
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -