MCOTM: Mobility-aware computation offloading and task migration for edge computing in industrial IoT

Mobility-aware devices are crucial components of Industrial Internet of Things (IIoT). However, they face limitations in terms of battery capacity and computation power, which restrict their ability to provide services requiring broad bandwidth and strong computation power for computation-intensive tasks. While offloading can strengthen device computation power, ineffective offloading decisions result from device mobility and limited adaptability to changes in environmental resources, or are not applicable to the current mobile edge computing (MEC) environment. In this paper, we address these challenges by proposing a mobility-aware computation offloading and task migration approach (MCOTM) based on trajectory and resource prediction to address this issue of mobility offloading, which minimizes task turnaround time and system energy consumption. Simultaneously, our approach enhances the decision agent continuously to decrease task migration rates. MCOTM uses Lagrange interpolation equations to determine the trajectory of mobile devices, and Long Short-Term Memory (LSTM) to track the time-varying resources characteristics in IIoT. These prediction results will be used to assist Deep Deterministic Policy Gradient (DDPG) for making online computation offloading, task migration and resource allocation decisions. Experimental results show that the proposed MCOTM effectively reduces task turnaround time by at least 42% and system energy consumption by 10% while maintaining a low task migration rate of around 50%, even with an increasing number of tasks.