FloodWatch: Devising an Autonomous Cyber Physical System for Real-Time Flood in an Operational Framework

Flooding has devastating impacts in
Southeast Asia, in part due to the lack of
appropriate infrastructure. As a part of my
involvement in the FloodWatch research
group, I am devising an autonomous cyber
physical system for real-time flood
intelligence that can be leveraged by citizens
in a progressive webapp. FloodWatch is a
smart city service centered around the
development of cyber physical system
infrastructure for disaster intelligence and
forecasting. The service provides a frontend
progressive web application (PWA), live
sensor readings through LoRAWAN router
technology, and predictive insights on flood
susceptibility. To produce insights on
susceptibility, I am working on several
machine learning pipelines to ensure
accuracy, explainability and corroboration.
The current approach combines an existing
hydrological model called HEC-RAS with a
customized Artificial Neural Network
(ANN) to produce geographical flood
inundation that is then interpreted and
overlaid on the map of Vietnam using a
region tileset automation script. Additionally,
I am devising a corroboration mechanism for
crowdsourced flood reports by using LiDAR
water level interpretation, computer vision
for flood label detection of live camera feeds.