Allocating Resource for Task Groups in MEC IoT Systems with Reinforcement Learning

Abstract

The virtualization of Real World Object (RWO), coupled with its association with mobile edge computing (MEC) server, is gaining popularity in our today's internet of things (IoT) network. We see the prevalence in the savings it brings in the node's resources to execute the task by an application, thereby realizing the IoT vision. This paper looked at the interaction between a task group and the MEC server as a solution to reduce the average cost in the consumed power and delay in processing task and consequently to lead to a high network lifetime. To achieve this, we have added to the functionality of the virtual object (VO) and combined the ideas of the familiar schemes in a Mobile-edge learning and consensus algorithm. Markov's decision process is used to model a task group's choice, or the MEC server and solution got using a reinforcement learning algorithm. The simulations, when compared to some earlier schemes, showed significant improvement in power consumption, task processing delay, and network lifetime.