A Deep Reinforcement Learning based policy gradient for Energy Consumption in Edge Computing

IoT devices gain much attention in today's world, including various applications like natural language processing, face recognition and augmented reality etc. The above applications need huge resources which will ruin the energy of the device. Based upon the increase of the device usage, the task of computation is a critical process. A pioneer solution to solve the problem is Edge computing, which will bring the computation closer to the IoT device. The computation can be done both in the device and in the remote edge, it is based on the need and resource availability of the IoT application. The offloading can be achieved by using the Deep Reinforcement learning algorithm (DRL) to acquire the concept of Nash Equilibrium. The offloading decisions can be taken based on Policy Gradient algorithm to overcome the problem of Markov Decision Process. The various algorithms of policy gradient are compared with the standard algorithm used for offloading.