Skip to main content

Real-time Flood Monitoring System Employing Telecommunications Networks and Machine Learning

PhD project title and outline, including interdisciplinary dimension:
Real-time Flood Monitoring System Employing Telecommunications Networks and Machine Learning

Around the UK, flooding has cost an average of £1,400 million annually, as towns and cities continue to sustain damage
from the rising waters. In this context, this proposal is focused on solving the pressing problems in the face of increases
in flood-prone areas in the UK, by using advanced technologies in wireless communications and artificial intelligence.
This project proposes to develop a real-time flood monitoring system to improve prediction accuracy and response time
related to flood hazards. An innovative framework is proposed whose aim is to achieve the above goals by identifying
two major work packages:

WP1: Real-time Wireless Monitoring (RWM) – A RWM system consisting of wireless sensors (to provide data e.g. on
rainfall, stream level, temperature, humidity and wind speed) and amateur unmanned aerial vehicles (UAVs, to capture
high resolution images of the areas affected for flood extent delineation and damage assessments) will transmit the
observed data to a Ground Control Station (GCS). The aim is to provide high quality-of-experience communications,
energy efficiency, and optimal deployment. The RWM will be designed as a self-configured network, tolerant in holistic
environments and severe weather, without human interception. The RWM also has the function of self-healing and selfforming
to intelligently estimate situations and perform desired actions.

WP2: Real-time Artificial Intelligence - The measurements collected by RWM are inherently uncertain (e.g. noisy, or
missing due to device malfunction), heterogeneous, high-dimensional, and from various sources spreading through
many locations. To deal with this problem of very large data-set at the GCS, we will propose new algorithms to better
assess and predict how these events are shifting, through the use of state-of-the-art deep learning and data-driven
approaches such as embedding network layers, convolutional neural networks and recurrent neural network.

Primary supervisor: Dr. Trung Q. Duong (EEECS)
Secondary supervisor: Dr. Vien Ngo (EEECS/ECIT/DSSC)
External Partner/Organisation: Nokia Bell Labs