Cao, Jia (2023) Design of Random Access Protocols with Neural Networks for the Internet of Things PRE - Research Project, ENSTA.
![]() | PDF Restricted to Registered users only 1002Kb |
Abstract
This internship investigates the optimization of wireless network performance through modern random access techniques, focusing on the Irregular Repetition Slotted Aloha (IRSA) protocol within the Internet of Things (IoT) systems. The primary objective is to enhance the throughput in short-frame-size scenarios. This study introduces two key enhancements: the consideration of multiple transmission power levels (Multi-Power IRSA, MP-IRSA) and the introduction of the sensing of neighboring nodes’ transmissions (Deep-Sensing Multi-Power IRSA, DS-MP-IRSA). The research encompasses a literature review, mathematical analyses of simplified scenarios, and simulation-based studies utilizing Deep Reinforcement Learning techniques (Proximal Policy Optimization). MP-IRSA explores multiple power levels to exploit the capture effect, demonstrating significant throughput improvements even with only two power levels. DS-MP-IRSA introduces a sensing phase before transmission, enabling informed decisions based on observed channel energy and showcasing throughputs exceeding one packet per slot for short frame sizes. This study improves wireless communication protocols and suggests approaches to boost network capacity in short-frame-size IRSA scenarios. By using DRL techniques and techniques and advanced sensing, we demonstrate how to enhance the performance of IoT networks
Item Type: | Thesis (PRE - Research Project) |
---|---|
Uncontrolled Keywords: | Internet of Things, Communications, Protocols, Random Access |
Subjects: | Information and Communication Sciences and Technologies |
ID Code: | 9580 |
Deposited By: | Jia CAO |
Deposited On: | 24 août 2023 15:15 |
Dernière modification: | 30 août 2023 10:17 |
Repository Staff Only: item control page