Drones and AI systems developed to detect natural disasters

 More reliable and efficient systems will spot unfolding threats

A new early-warning system using drones and artificial intelligence will be developed to detect a selection of natural disasters in a project led by Manchester Metropolitan University.

Natural disasters – including earthquakes, floods and fires – not only pose a threat to human life and safety, but also threaten global economies and the climate.

Around 1.7 billion people have been affected by a natural disaster over the past decade, resulting in more than 400,000 deaths, according to a recent report published by the British Red Cross.

As most natural disasters occur suddenly, early-warning systems (EWSs) are designed to predict or detect them, helping responders to act quickly and minimise the impact.

However, for events such as flash flooding, wildfires and frost damage – which are not always considered to be as major as disasters like volcanic eruptions and earthquakes – such systems aren't as readily available.

This is despite these weather events causing significant impacts to public safety, the economy and the climate – with flooding affecting more people globally than any other natural disaster.

But a new project, funded by NATO's Emerging Security Challenges Division, aims to change this, focusing on developing a system that can predict this set of extreme-weather events.

Dr Khaled Rabie, Reader in the Department of Engineering at Manchester Met and the academic lead for the project, said: "Flash flooding, wildfires and frost damage are natural disasters, which can often be overlooked for more catastrophic events. However, they cause immeasurable amounts of damage.

"Here in the UK, flooding is the most common natural disaster we encounter, but we have also been confronted with wildfires and land damage from frost.

"And as the world continues to warm as a result of climate change, these extreme weather events will only become more common, which is why it is vital we have systems in place that can help us detect these events and be more prepared."

Efficient solution
Scientists explain that EWSs currently used for natural disaster detection use wireless sensor networks (WSNs) that can measure environmental conditions such as temperature and pollution.

However, the networks require a large number of sensors to be spread out across vast amounts of land in order to work effectively.

Conditions in these areas can be harsh, making it difficult – sometimes impossible – to not only deploy the sensors, but also to recharge and replace batteries.

Because of this, researchers say it is necessary to look for alternative solutions, which are more energy efficient and easier to deploy.

Transmitting warning messages can also be difficult in traditional EWSs, due to the substantial number of sensors covering large areas, which can lead to long distance transmissions and congestion in transmitting messages. This also uses a massive amount of energy.

However, the team behind the 36-month project, named RESCUE, has already begun work on a new system that they say will be efficient, reliable, and practical for natural disaster applications.

Researchers from the UK, along with partners in Jordan and Morocco, are investigating a range of emerging technologies to see how they can be used to improve EWSs.

They believe a hybrid approach, utilising a range of technologies, will be most effective.

Dr Rabie said: "What we want to achieve in this project is to create a system that is reliable enough to prevent any waste of resources spent in the reaction taken as a response to a warning.

"This means that warnings need to be based on accurate sensing results, which makes creating a strategy for detection and sending warnings to be of paramount importance to the overall performance of any EWS.

"On the other hand, energy efficiency is a huge concern as there are limited energy resources in the sensor equipment, and it's difficult to recharge and replace batteries in harsh environments."

Future development
Scientists will be exploring how both radio frequency and optical technologies – which uses light such as infrared and ultraviolet – can be used in different environments to monitor conditions and transmit warnings.

Using different complementary technologies for this will make transmission timelier and more reliable than depending on just one solution.

In addition, they will be exploring the use of Unmanned Aerial Vehicles (UAVs) – also known as drones – that will fly over the at-risk area, collecting and transmitting data – helping to ensure continuous observation.

Passive retroreflectors – devices which reflect light – will be used to minimise the amount of energy used by the sensors.

Artificial intelligence (AI) and machine learning (ML) algorithms will also be developed to further enhance the new solutions, as they will provide rapid detection of any network failures and help to repair them immediately.

https://www.mmu.ac.uk/science-engineering/about-us/news/story/index.php?id=16304