Fuzzy Reinforcement Learning Algorithm for Efficient Task Scheduling in Fog-Cloud IoT-Based Systems

In recent years, the number of IoT applications that require low latency has increased greatly. Traditional cloud servers cannot handle these applications due to strict latency requirements. Edge technologies like fog computing meet these applications' latency needs. Computing infrastructure is near end-user devices in fog-cloud environments. There are numerous traditional methods for scheduling IoT applications on heterogeneous and distributed fog-cloud nodes in these fields. Research in machine learning and its applications in many fields has grown tremendously in recent years. Machine learning algorithms such as reinforcement learning (RL) can be used to learn and make decisions based on reward signals from the environment. The purpose of this paper is to present a Task Scheduling algorithm based on Fuzzy Reinforcement Learning (TSFRL) to allocate fog-cloud computing resources so as to meet the deadlines of IoT requests. The scheduling problem is initially formulated to reduce response times, costs, and energy consumption. Fuzzy logic is then used to prioritize tasks. Fog nodes and cloud nodes employ the on-policy reinforcement learning methodology to prioritize delay-sensitive tasks with a higher priority and delay-tolerant ones with a lower priority. The suggested strategy outperforms existing algorithms in response time, cost, energy usage, and percentage of deadlines met.