Synergizing Fuzzy-based Task Offloading with Machine Learning-driven Forecasting for IoT

Nowadays IoT applications play an increasing role in our lives, which require adaptive solutions to meet an acceptable level of quality. To support this need, IoT is often coupled with fog and cloud services, especially for real-time applications, where short latency and fast data integration matter the most. To manage such IoT-Fog-Cloud systems effectively, task offloading methods are needed, which represent an NP-hard problem, to be addressed by scheduling algorithms using some sort of heuristics. In this paper, we present a fuzzy-based offloading algorithm improved by machine learning-based time-series prediction to boost its decision making, and thus to reach efficient system usage. We also extend and use the DISSECT-CF-Fog simulator to evaluate our proposed approach on a real-world-based IoT use case. Our results showed improvements by reducing execution time of the considered IoT application with essentially the same utilisation cost, energy consumption and network usage.