TY - JOUR
T1 - Handle Heavy Workload in Penalty-Based Machine-Type Communication
T2 - Using ResNet
AU - Feng, Zhipeng
AU - Du, Changhao
AU - An, Jianping
AU - Zhang, Zhongshan
N1 - Publisher Copyright:
© 2012 IEEE.
PY - 2024
Y1 - 2024
N2 - Although machine-type communication (MTC) will play a crucial role in massive-terminal communication in the future, existing MTC systems often overlook the impact of workload on system performance, leading to the underestimation of the effects of high workload and link retransmission in existing power allocation schemes, ultimately resulting in a deterioration of network capacity. This letter quantifies the penalty caused by retransmission and employs the penalty-weighted sum-rate (PWSR) to evaluate the system performance under different workload conditions, thus studying penalty-based MTC (pMTC) networks. Meanwhile, to obtain the optimal PWSR, we consider both computational cost and performance when designing a power scheme. The residual network (ResNet) adds skip connections on the basis of the deep neural network (DNN), which not only retains the advantages of simple structure and considerable performance but also has the advantages of fast convergence and effective mitigation of gradient vanishment. Thus, we adopt a ResNet-based scheme to allocate the power of each device in the pMTC. The numerical results indicate that under heavy workload conditions, the performance loss generated by the ResNet-based scheme is much lower than that of other investigated schemes, and the former has a better "penalty-robustness"and a lower computational complexity.
AB - Although machine-type communication (MTC) will play a crucial role in massive-terminal communication in the future, existing MTC systems often overlook the impact of workload on system performance, leading to the underestimation of the effects of high workload and link retransmission in existing power allocation schemes, ultimately resulting in a deterioration of network capacity. This letter quantifies the penalty caused by retransmission and employs the penalty-weighted sum-rate (PWSR) to evaluate the system performance under different workload conditions, thus studying penalty-based MTC (pMTC) networks. Meanwhile, to obtain the optimal PWSR, we consider both computational cost and performance when designing a power scheme. The residual network (ResNet) adds skip connections on the basis of the deep neural network (DNN), which not only retains the advantages of simple structure and considerable performance but also has the advantages of fast convergence and effective mitigation of gradient vanishment. Thus, we adopt a ResNet-based scheme to allocate the power of each device in the pMTC. The numerical results indicate that under heavy workload conditions, the performance loss generated by the ResNet-based scheme is much lower than that of other investigated schemes, and the former has a better "penalty-robustness"and a lower computational complexity.
KW - machine-type communication
KW - penalty
KW - power scheme
KW - residual network
KW - workload
UR - http://www.scopus.com/inward/record.url?scp=85213488474&partnerID=8YFLogxK
U2 - 10.1109/LWC.2024.3520993
DO - 10.1109/LWC.2024.3520993
M3 - Article
AN - SCOPUS:85213488474
SN - 2162-2337
JO - IEEE Wireless Communications Letters
JF - IEEE Wireless Communications Letters
ER -