Machine Learning Driven Latency Optimization for Application-aware Edge Computing-based IoTs

Most IoT devices have limited or no computing capability while many emerging IoT applications require both computing and communications services. Moreover, low latency requirements of numerous applications such as autonomous driving and augmented reality are becoming critical. In this paper, we propose a novel framework that can utilize the edge-computing facilities and the full-duplex technique at the edge nodes to address computing and communication services with low latency to IoT terminals for different applications. We then formulate an application-aware edge-computing problem for IoTs with the target to minimize the average latency. We propose a machine learning algorithm to solve this problem by achieving the best user-edge-node assignment and developing an optimal assignment and scheduling algorithm for the communication and computing resources. We evaluate the performance of the proposed machine learning algorithm (via Python and Tensorflow) and present results and comparison with other methods.