Fast Q-learning for Improved Finite Length Performance of Irregular Repetition Slotted ALOHA
- Submitting institution
-
The University of Essex
- Unit of assessment
- 11 - Computer Science and Informatics
- Output identifier
- 1396
- Type
- D - Journal article
- DOI
-
10.1109/tccn.2019.2957224
- Title of journal
- IEEE Transactions on Cognitive Communications and Networking
- Article number
- 2
- First page
- 844
- Volume
- 6
- Issue
- 2
- ISSN
- 2332-7731
- Open access status
- Compliant
- Month of publication
- December
- Year of publication
- 2019
- 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
-
1
- Research group(s)
-
C - Communications and Networking (Comms)
- Citation count
- 2
- Proposed double-weighted
- No
- Reserve for an output with double weighting
- No
- Additional information
- Published in IEEE-TCCN, the flagship journal for machine learning for communications, this paper proposes a novel method based on advanced reinforcement learning for optimising the degree distribution of irregular repeated slotted ALOHA(IRSA) protocol in the finite block-length regime. The proposed solution is theoretically analysed and provides guarantees of finding near-optimal transmission policies. The theoretical analysis is widely applicable to resource-optimisation problems suffering from waterfall-effect. Results demonstrated that the proposed method achieves high-throughput in wireless sensor network settings and does not suffer from severe degradation of IRSA protocols (challenging high channel-loads case). This formed part of a TUBerlin requested invited talk.
- Author contribution statement
- -
- Non-English
- No
- English abstract
- -