Multiple Access Binary Computation Offloading via Reinforcement Learning

Computation offloading enables energy-limited mobile devices to expand the range of applications that they can execute. When multiple devices each seek to execute a latency-constrained indivisible task, the problem of device energy minimization involves jointly making binary decisions on whether or...

Full description

Saved in:

Bibliographic Details
Published in:	2019 16th Canadian Workshop on Information Theory (CWIT) pp. 1 - 6
Main Authors:	Salmani, Mahsa, Sohrabi, Foad, Davidson, Timothy. N., Yu, Wei
Format:	Conference Proceeding
Language:	English
Published:	IEEE 01.06.2019
Subjects:	Closed-form solutions Computational efficiency Computational modeling Energy consumption Minimization Resource management Task analysis
Online Access:	Get full text
Tags:	Add Tag No Tags, Be the first to tag this record!

Description
Summary:	Computation offloading enables energy-limited mobile devices to expand the range of applications that they can execute. When multiple devices each seek to execute a latency-constrained indivisible task, the problem of device energy minimization involves jointly making binary decisions on whether or not each user should offload its task along with the allocation resources to the offloading users. It has been shown that for a K-user system that employs a multiple access scheme that exploits the full capabilities of the channel, when the binary decisions are given, a closed-form expression for the optimal resource allocation can be obtained. In this paper, we propose a reinforcement learning-based algorithm for finding offloading decisions that takes advantage of this closed-form expression for the resource allocation. Our numerical experiments illustrate that the proposed algorithm can achieve a better trade-off between performance and computational cost as compared to the existing approaches in the literature.
DOI:	10.1109/CWIT.2019.8929930