Fighting fires could be done remotely without the need to place firefighting crews directly in potentially dangerous situations by using collaborative teams of artificial intelligence-powered robots with extinguishing equipment on board, with an initial soft trial of the technology proving successful.
Led by Cyborg Dynamics Engineering with Griffith University and funded by Queensland Defence Science Alliance (QDSA), the team demonstrated the technology in both simulated and hybrid simulation-physical demonstrations using an unmanned ground vehicle (UGV), simulating real fires for a team of up to five robots to extinguish.
In the trial, the UGV successfully navigated around physical obstacles and teamed up with its simulated robot team members to locate and work together to extinguish multiple simulated fires.
Dr Zhe Hou, project Lead Chief Investigator from Griffith University's School of Information and Communication Technology, said the results demonstrated a 99.67 per cent success rate in navigating and extinguishing two fires, suggesting its strong potential for real-world deployment.
"We demonstrated that multiple real and simulated UGVs, trained through a structured three-stage AI learning curriculum, could learn to perform both low-level navigation and high-level collaborative tasks," Dr Hou said.
"This confirms the operational potential of our approach for practical case studies such as autonomous navigation and firefighting."
The research team adopted an artificial intelligence (AI) technique called multi-agent reinforcement learning (MARL) to build neural-network-based AI 'agents' trained through a custom-designed curriculum, progressing from simple tasks such as single-robot navigation, to multi-robot navigation around obstacles, then finally to completing a complex firefighting scenario involving multiple robots and fires with obstacles.
The team said the ability of the robots to self-organise and allocate tasks autonomously - such as splitting into teams to handle multiple fire outbreaks - reduced the cognitive load on human operators, offering increased safety and operational efficiency in emergency situations.
"We have developed the control systems for firefighting UGVs that are currently deployed on mine sites across Australia," Cyborg Dynamics Engineering General Manager Ryan Marple said.
"These units are remotely controlled by a human, a bit like an RC car.
"They have been an extremely effective measure in removing human firefighters from dangerous situations and enabling high-value assets to be saved from fires.
"The future of these kinds of vehicles - and the focus of this research - is the automation of low-level control and swarming behaviour across multiple agents. Such autonomous swarms can respond to complex situations in a way that just isn't possible with direct manual control."
Ryan Marple, Cyborg Dynamics Engineering
"By ingesting data from a wide variety of sensors, these systems can make decisions quickly, which just isn't possible by the very limited situational awareness of a human looking at a screen."
Looking ahead, the research team envisioned further advancements in both the design of neural networks and sim-to-real transfer methodologies.
Future work would also explore the adoption of the developed AI technique on other autonomous systems, such as autonomous underwater vehicles and unmanned aerial vehicles, or even a hybrid team of different types of vehicles.
The study 'Multi-agent reinforcement curriculum learning for real unmanned ground vehicles' has been published in Engineering Applications for Artificial Intelligence.
