An Efficient Offloading Technique using DQN for MEC-IoT Networks

In today’s day there are billions of interconnected smart devices present which are called Internet-of-Things (IoT). It contains software, sensors and a bunch of other technologies trying to connect the systems with the purpose of sharing information over the internet. This system is producing massive amounts of data over the internet. This massive amount of data requires an appropriate spectrum to transmit the data to their server. so that it benefits us in an economical way. Machine learning has been successful in fields of speech recognition, graphics, prediction and decision making. Machine learning is used in our research for optimally allocating the resources. Edge computing can be a possible solution for IoT networks because it puts computer capabilities closer to the user. IoT users have recently been created to have more powerful computing abilities, allowing them to run few tasks locally. The article tries to use edge computing with Machine learning for optimal allocation of resources. K-means clustering algorithm is used to cluster the IoT user into different clusters depending on the distance and feature of their user priority. The cluster of higher priority is executed at the edge level while the lower cluster is executed at the local level. The challenge is to provide a long-term solution for the economical, effective resource allocation in the available spectrum with minimum energy usage. A deep Q-network (DQN) based computational offloading technique is also created to gain knowledge of the appropriate computational offload plan. Furthermore, the DQN-based computational offloading technique surpasses the baseline systems. Through the simulation results, it is concluded that the proposed approach outperforms with the state-of-the-art models.