Adaptive QoS-Aware Task Offloading in Dynamic Mobile Edge Computing Environment

Ensuring Quality of Service (QoS) for real-time applications like Augmented Reality in a Mobile Edge Computing (MEC) setting is both vital and demanding in research. In this work, we propose a novel framework of hybrid ML approaches to enable QoS-aware offloading in dynamic MEC. First, we create a method using deep reinforcement learning to figure out the best way to offload tasks for an application in a new MEC environment, even when we don’t have early information about the application’s QoS. Then, we deploy a transfer learning approach in the dynamic MEC that transfers knowledge from previously trained deep reinforcement off-loading policies to new optimal policies. Experimental results show that the proposed framework adapts to the dynamic MEC environments efficiently, reducing ML training time and retaining higher accuracy and precision during the task offloading process.