Design of a Q-learning based Smart Grid and smart Water scheduling model based on Heterogeneous Task Specific Offloading process

Smart cities require deployment & management of multiple Internet-of-Things (IoT) devices that can sense, communicate & perform information-specific actions. One of the most important tasks for deployment of these devices is controlling their scheduling operations. Existing smart city deployment models are either highly complex, or cannot be scaled beyond a certain set of devices, which limits their deployment capabilities. Moreover, most of these deployments are suited for a limited number of nodal traffic, which further affects their scalability levels. To overcome these issues, this text proposes design of a Q-learning based task scheduling Model that can uses Edge-Based Heterogeneous Task Specific Offloading (HTSO) with Genetic Algorithm (GA) based optimizations. The Q-learning model initially clusters tasks based on their importance-levels, which are determined by the number of real-time request types. High-capacity edge nodes and edge computing devices have their sleep patterns determined by these clusters depending on their relative importance levels. All of these choices are made by VMs in the edge devices, which helps to keep all of the edge computers in sync with one another. The Q-Learning Model may determine task-edge device mapping schedules based on priority levels of tasks and the computing efficiency of edge devices. Finding computationally-free edge devices using the GA Model, then assigning them duties depending on priority levels for those activities. Combining Q-Learning with GA allows the model to improve its computing efficiency for Smart Grid & Smart Water scheduling scenarios. This is accompanied with reduced average access latency, average response delay, and energy usage levels. The model was deployed on multiple smart city scenarios, and its performance was compared w.r.t. different standard deployment models. Based on this comparison, it was observed that the proposed model was able to improve scheduling efficiency by 15.5% for smart grids, and 8.3% for smart water scheduling, increase scheduling speed by 3.9% for smart grids, and 4.5% for smart water scheduling, while maintaining high computational efficiency of 99.5% for both the applications. Due to these advantages, the proposed model is capable of deployment for a wide variety of smart city use cases.