A location and mobility-aware routing framework for improving multimedia streaming performance in mobile ad-hoc networks

Abstract

Disaster telemedicine leverages communications networks to provide remote diagnosis of injured persons in areas affected by disasters such as earthquakes. However, telemedicine relies heavily on infrastructure, and in a disaster scenario there is no guarantee that such infrastructure will be intact. In an ad-hoc network, devices form a network amongst themselves and forward packets for each other without infrastructure. Ad-hoc networks could be deployed in a disaster scenario to enable communications between responders and base camp to provide telemedicine services. However, most ad-hoc routing protocols cannot meet the necessary standards for streaming multimedia because they do not attempt to manage Quality of Service (QoS). Node mobility adds an additional layer of complexity leading to potentially detrimental effects on QoS.

Geographic routing protocols use physical locations to make routing decisions and are typically lightweight, distributed, and require only local network knowledge. They are thus less susceptible to the effects of mobility, but are not impervious. Location-prediction can be used to enhance geographic routing, and counter the negative effects of mobility, but this has received relatively little attention. Machine Learning algorithms have been deployed for predicting locations in infrastructure networks with some success, but such algorithms require modifications for use in ad-hoc networks.

This thesis explores the possibility of using an Artificial Neural Network (NN) to perform location-prediction in an ad-hoc network. These predictions along with other context information can be used to make QoS-aware routing decisions. The aim of Geographic QoS Peer-to-peer Streaming framework (GQP2PS) is to provide a means of managing QoS in an ad-hoc network, allowing the utilisation of telemedicine services in disaster recovery scenarios. A peer-to-peer (P2P) model has been chosen for streaming because it reflects the distributed and decentralised environment in which streaming will be performed.

Date of Award	Jan 2014
Original language	English
Supervisor	Kevin Curran (Supervisor), Jose Santos (Supervisor) & Sandra Moffett (Supervisor)