Vulnerable next to the raging inferno that is an out-of-control bushfire, human responders need all the help they can get. That’s why engineers are looking to robotic swarms to lend a hand.
For close to 100 years, firefighters in Australia have included air power in their arsenal. It was February 1930 when the Australian Air Force sent a sole Westland Waipiti biplane to monitor territory in Victoria’s Dandenong Ranges, with these patrols rapidly becoming standard firefighting practice.
Australia has since embraced the advantages of aerial technology in bushfire detection, mapping and modelling. By February 1967, planes were dropping retardant on outbreaks in Victoria, and in 2014, the New South Wales Rural Fire Service began trialling drones to gather thermal and visual data that helped ground crew respond to blazes more effectively.
Other fire-prone countries around the world are similarly looking to make unmanned aerial vehicles (UAVs) a more effective part of their disaster response. NASA, for instance, is seeking to integrate onboard computing methods into fleets of drones that can monitor and respond to fires in high-risk areas with minimal human interaction.
“The goal is using UAVs for early detection of fire in forest areas and developing AI models for predicting how the fire is going to spread,” Dr Fatemeh Afghah, the principal investigator on the project, told create.
A professor of electrical and computer engineering at Clemson University, Afghah has worked with fire departments in states across the US, including Florida, California, Washington, Oregon, South Carolina and Georgia.
Drone swarms, she said, offer disaster responders two advantages over traditional UAVs. The first is simply cost: a sophisticated UAV can set fire departments back US$200,000 to US$300,000, several times more than an off-the-shelf model.
“You can have a fancy, expensive drone or robot that can do multiple tasks, but when you’re operating during a disaster, it’s very likely for you to lose that agent,” she said.
Distributing resources, such as cameras and sensors, between multiple affordable drones makes the swarm more durable, as well as more flexible, than a single UAV.
Tough working conditions
Having so many drones engaged at once introduces new problems, particularly considering the intense wind, heat and smoke that firefighting drones must operate around.
“One of the challenges we always have and will have is communication: communication between the agents, and communication between the agents and the controller,” Afghah said.
“Especially operating in disaster zones, the default source of communication is usually 4G, LTE or Wi-Fi.”
Satellite communication can provide essential coverage, but it may require specialised terminals, additional power and an unobstructed view of the sky. Depending on the satellite system, it may also offer higher or less predictable latency.
A more responsive option could come via 4G LTE and, eventually, 5G and 6G networks. Because aerial drones often have a clearer line of sight to distant cellular towers, they can maintain connectivity across a wider area.
In particular, 5G Ultra-Reliable Low-Latency Communication, or URLLC, is designed to support time-critical links with very low delay and high reliability. These capabilities could enable rapid transmission of control commands, sensor data, and video between firefighting drones, swarm controllers, and human operators.
But unless a swarm works in a fully distributed manner, it must also be designed so as to avoid a single point of failure, where the loss of a lead drone results in the entire swarm becoming disconnected from its operator.
Uncertainty during operations also compounds the challenges. The environment itself could be affected by a fire in ways that even a drone equipped with a variety of sensors might not correctly interpret — multispectral imaging and thermal sensors, for instance, cannot detect a complex column of wind shaping the direction of an outbreak.
To help address this issue, Afghah hopes to develop drones controlled by more sophisticated and better trained AI models, which are able to approach problems proactively and creatively.
“We know the factors that are contributing to the environment are unobservable, so we develop models that can estimate … what those factors are and how they are changing,” she said. “The UAV prediction models and the UAV navigation models can take that into account.”
Laying the groundwork
Where Afghah’s drone swarms use their lofty position to give firefighters a perspective otherwise unavailable to them, in Australia, CSIRO’s Data 61 research network wants to send out robotic firefighting agents that are able to provide more forensic detail than can be gathered from the air.
That means ground-based robots, working together using different approaches to explore at-risk areas and returning information about fuel load.
Related article: Australian engineers in line to win lucrative XPrize for bushfire detection
“If you’re above … you can only see the tree crowns,” Data 61 senior research scientist Nicholas Lawrance told create.
“You can do some amount of tree counting, but it’s pretty hard to estimate how dense it is underneath.”
A ground-based robot might be slower, but it’s able to gather a much richer data set. And a swarm of ground-based bots adds further benefits.
“But if you have a limited amount of time to get something done and you really need to collect [that data], then having more sensors out there that can be distributed makes a difference.”
It also means sensors and tactics can be designed to complement one another. One robot can capture large amounts of information, allowing it to identify points of interest, before communicating those locations to another in the swarm, which will follow up with more precise measurements.
The goal for CSIRO’s contribution to the Silvanus Project, a from effort by the European Union, was to build relatively high levels of autonomy into the robots. Using a combination of legs and tracks, the machines incorporated LiDAR sensors that permitted them to negotiate different terrain types, even in the presence of smoke.
Lawrance knows, however, that technology must follow the needs of the human crew first and foremost. Firefighters are more likely to care about a rich data set than a sophisticated AI able to make decisions without human interaction.
“The closer to a fire it is, the more they want and need control,” he said.
“There are a lot of things happening during a fire, and they’re not interested in a lot of complicated systems adding to their workload.”
Batteries are essential for RE to store energy. Long term, batteries will have longer cycles, longer life. More efficient. Australia to invest in building battery factory is good investment, gradually the world demand efficeint and long life batteries