TY - JOUR
T1 - Energy-Efficient Robust Computation Offloading for Fog-IoT Systems
AU - Wu, Zhikun
AU - Li, Bin
AU - Fei, Zesong
AU - Zheng, Zhong
AU - Li, Bin
AU - Han, Zhu
N1 - Publisher Copyright:
© 1967-2012 IEEE.
PY - 2020/4
Y1 - 2020/4
N2 - As the computing nodes of a fog computing system are located at the network edge, it can provide low-latency and reliable computing services to Internet of Things (IoT) mobile devices (MDs). By wirelessly offloading all/part of the computational tasks from MDs to the infrastructure fog nodes, it addresses the contradiction between the limited battery capacity of MDs and their long-lasting operation requirement. Different from previous works, the uncertainty caused by the channel measurements is taken into account in this paper, which yields a robust offloading strategy against realistic channel estimation errors. For this system, we design an energy-efficient computation offloading strategy, while satisfying the delay constraint. By using the Conditional Value-at-Risk (CVaR) framework, the original offloading problem is transformed into a Mixed Integer Nonlinear Programming (MINLP) problem, which is complicated and very challenging to solve. To overcome this issue, we apply Benders decomposition to find the optimal offloading solution. Numerical results show that proposed offloading strategy efficiently achieves obtain the optimal solution of the MINLP problem, and is robust to channel estimation errors.
AB - As the computing nodes of a fog computing system are located at the network edge, it can provide low-latency and reliable computing services to Internet of Things (IoT) mobile devices (MDs). By wirelessly offloading all/part of the computational tasks from MDs to the infrastructure fog nodes, it addresses the contradiction between the limited battery capacity of MDs and their long-lasting operation requirement. Different from previous works, the uncertainty caused by the channel measurements is taken into account in this paper, which yields a robust offloading strategy against realistic channel estimation errors. For this system, we design an energy-efficient computation offloading strategy, while satisfying the delay constraint. By using the Conditional Value-at-Risk (CVaR) framework, the original offloading problem is transformed into a Mixed Integer Nonlinear Programming (MINLP) problem, which is complicated and very challenging to solve. To overcome this issue, we apply Benders decomposition to find the optimal offloading solution. Numerical results show that proposed offloading strategy efficiently achieves obtain the optimal solution of the MINLP problem, and is robust to channel estimation errors.
KW - Internet of Things
KW - benders decomposition
KW - conditional value-at-risk
KW - offloading
KW - robust
UR - http://www.scopus.com/inward/record.url?scp=85083818109&partnerID=8YFLogxK
U2 - 10.1109/TVT.2020.2975056
DO - 10.1109/TVT.2020.2975056
M3 - Article
AN - SCOPUS:85083818109
SN - 0018-9545
VL - 69
SP - 4417
EP - 4425
JO - IEEE Transactions on Vehicular Technology
JF - IEEE Transactions on Vehicular Technology
IS - 4
M1 - 9003203
ER -