Reinforcement learning based task offloading of IoT applications in fog computing: algorithms and optimization techniques

In recent years, fog computing has become a promising technology that supports computationally intensive and time-sensitive applications, especially when dealing with Internet of Things (IoT) devices with limited processing capability. In this context, offloading can push resource-intensive tasks closer to the end devices at the network edge. This allows user equipment to profit from the fog computing environment by offloading their tasks to fog resources. Thus, computation offloading mechanisms can overcome the resource constraints of devices and enhance the system's performance by minimizing delay and extending the battery lifetime of devices. In this regard, designing an algorithm to decide which tasks to offload and where to execute them is crucial. Recently, there has been a growing interest in utilizing Reinforcement Learning (RL) and deep reinforcement learning (DRL), to address computation offloading mechanisms in the context of fog computing. This paper reviews the research conducted on Reinforcement learning (RL) and Deep Reinforcement Learning (DRL) based computation offloading mechanisms for IoT applications in the fog environment. We provide a comprehensive and detailed survey, analyzing and classifying the research paper in terms of RL techniques, objectives, architecture, and use cases. Then, in particular, we identify the advantages and weaknesses of each paper. After that, We systematically elaborate on open issues and future research directions that are crucial for the next decade.