Enabling Energy-Efficient IoT via Learning Assisted Header-Free Communication

With millions of connected devices expected to proliferate across multiple application domains, energy efficiency is a critical factor in IoT solutions. This paper aims to enhance the energy efficiency of networked IoT sensors by transitioning to a header-free communication framework. Novel enhancements to the reception technique based on the stochastic expectation maximization algorithm are proposed. Specifically, in contrast to prior efforts, a combination of compressive sensing principles along with deep learning methodologies are used to improve the performance of header-free sensor communications. Using simulation results, performance & complexity gains relative to the classic approach of up to 95% and 99%, respectively, are achieved.