Reliability Versus Latency in IIoT Visual Applications: A Scalable Task Offloading Framework

Junchao Ma,Bodong Shang,Hao Song,Yongming Huang,Pingzhi Fan

DOI: https://doi.org/10.1109/JIOT.2022.3148115

IF: 10.6

2022-01-01

IEEE Internet of Things Journal

Abstract:In Industrial Internet of Things (IIoT), reliability and latency are two important performance indicators. However, these two performance indicators are contradictory with each other, which are difficult to be enhanced simultaneously. In reality, many IIoT applications are in video or image format with critical requirements of both reliability and latency. In this article, we propose a scalable task offloading scheme for IIoT visual applications considering the unique scalable feature compared with general content. The proposed scheme demonstrates that partial content can be adaptively offloaded to a specific computing node to meet the reliability and latency requirements, meanwhile obtaining an excellent tradeoff between them. For optimization, a utility function is defined to characterize the tradeoff between reliability and latency. Then, a greedy algorithm is introduced to solve the problem of maximizing the utility function, where a near-optimal scheduling policy is adopted to achieve offloading association and properly offload data volume. Simulation results reveal that the proposed scalable task offloading scheme performs better than other benchmark schemes in balancing reliability and latency in IIoT visual applications, especially when the network traffic is moderate and channel conditions are undesirable.