Deep Imitation Learning for 3D Navigation Tasks
- Submitting institution
-
Birmingham City University
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 11Z_OP_D0026
- Type
- D - Journal article
- DOI
-
10.1007/s00521-017-3241-z
- Title of journal
- Neural Computing and Applications
- Article number
- -
- First page
- 389
- Volume
- 29
- Issue
- 7
- ISSN
- 0941-0643
- Open access status
- Compliant
- Month of publication
- -
- Year of publication
- 2018
- URL
-
https://link.springer.com/article/10.1007/s00521-017-3241-z
- 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
-
-
- Research group(s)
-
-
- Citation count
- 8
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- The paper proposes a novel deep active imitation learning system, bringing active learning and learning from demonstrations together for building effective navigation systems in 3D environments. The paper inspired work in path planning in navigation systems . Driverless vehicles are among the many real-life examples of applications that can directly benefit from the proposed system. Extensive experiments using MASH simulator on four challenging 3D navigation tasks proved the effectiveness and superiority of active learning as a component, and the whole imitation learning system compared to state-of-the-art reinforcement methods.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -