A Block-structured Optimization Approach for Data Sensing and Computing in Vehicle-assisted Edge Computing Networks
Citation:
Luan N. T. Huynh, Md. Delowar Hossain, Quoc-Viet Pham, Yan Kyaw Tun, Eui-Nam Huh, A Block-structured Optimization Approach for Data Sensing and Computing in Vehicle-assisted Edge Computing Networks, IEEE Sensors Journal, 24, 1, 2024, 952 - 961Abstract:
With the rapid development of IoT applications and
multi-access edge computing (MEC) technology, massive amounts
of sensing data can be collected and transmitted to MEC servers
for rapid processing. On the other hand, as the number of
IoT devices grows, the MEC server cannot perform tremendous
computing tasks because of its limited computation capacity.
This paper introduces a vehicle-assisted MEC framework that
leverages vehicles to provide computational services for IoT
devices and overcome this challenge. The problem of latency
minimization was formulated by optimizing the sensing data rate,
offloading decisions, and resource allocation while considering en-
ergy consumption constraints. Nevertheless, achieving the global
optimal solution in polynomial time is challenging because the
formulated problem is mixed-integer nonlinear and non-convex.
This paper provides an efficient algorithm that adopts the block
coordinate descent technique to decompose the original problem
into four subproblems. These subproblems can be solved using
Lagrangian relaxation and the block successive upper-bound
minimization method. The superiority of the proposed approach
in reducing latency compared to baseline schemes is evident from
the simulation results.
Author's Homepage:
http://people.tcd.ie/phamqDescription:
PUBLISHED
Author: Pham, Viet
Type of material:
Journal ArticleCollections
Series/Report no:
IEEE Sensors Journal24
1
Availability:
Full text availableKeywords:
Computation offloading, Data sensing, Edge computing, Resource allocation, Vehicular networkSubject (TCD):
TelecommunicationsDOI:
10.1109/JSEN.2023.3332230ISSN:
1530-437XMetadata
Show full item recordLicences: