Robot co-pilots to make helicopter rescue faster, safer
Associate Professor Hanna Kurniawati and her team at the ANU College of Engineering & Computer Science (CECS) have been awarded two grants by the Australian Research Council (ARC). Both research projects stem from her groundbreaking work on robotic decision making in uncertain and dynamic environments.
“I’m grateful for the support ANU provides, from the early engagement with industry, all the way to the grant writing and administration of the grants,” Kurniawati said, adding that she is “excited about the research and the collaborations” that will result.
A virtual co-pilot for helicopter rescue missions
A research project to develop robotic pilot assistance for emergency helicopters was granted $497,630 by the ARC Linkage Program, which fosters long-term alliances between higher education organisations and private industry. The industry partner is Safran Electronics and Defense Australasia Pty Ltd.
ARC announced that the new piloting system “will advance fundamental algorithms for probabilistic planning in partially observable scenarios” to offer “robust, safe, and pilot-aware mission and manoeuvring strategies.” This will have life-saving impacts, particularly in remote areas that are far from hospitals.
A 2005 study in The Medical Journal of Australia found that the accident rate is 28 times higher during emergency medical helicopter missions compared to commercial aircraft.
Helicopters used for rescue missions and aeromedical transport do not typically have co-pilots because space is needed for medical equipment, for medical staff, and for a maximum number of passengers. While commercial flights can be cancelled or postponed due to inclement weather, rescue missions often cannot be. During storms and bushfires, weather and flying conditions can be unpredictable. Landing helicopters in the wrong spot might mean they cannot safely take off again due to the inclination of the land and the added weight of evacuees.
Built-in robotic co-pilots will help make life-and-death decisions and reduce the cognitive load for helicopter pilots sent on emergency missions by enhancing situational awareness without taking up additional space on the aircraft.
“We hope that the system will be adaptable to the pilot, to account for changes,” Kurniawati said. “For instance, if there is a search and rescue mission that has gone on for hours, is the pilot’s ability to do precise maneuvering declining due to fatigue? Is the pilot comfortable with taking risks, or is the pilot more risk averse?” Suggestions made by the robotic co-pilot will be adjusted accordingly.
Kurniawati said that the precise mix between audio and visual communication between the human pilot and the robotic co-pilot, as well as the complexity and precision of the information exchanged, can be finessed once the system is fully-developed and the interface is ready for testing.
Kurniawati describes the project as “a big computation problem” but she is confident that she and her team can deliver. “We have quite a good team with expertise in decision making under uncertainty and in robust decision making. I think ANU is very strong in this domain,” she said of a research team that includes Dr Felipe Werndl Trevizan, Professor Sylvie Thiébaux, and Professor Tom Gedeon.
Robots that help keep our infrastructures safe
A second ARC grant funds the development of intelligent robots that can assist in the building and maintenance of bridges, buildings, offshore oil rigs, and other complex structures.
“If the structure is in a remote area, say underwater or in space, it's dangerous and expensive to send people to do the inspection,” Kurniawati said.
If robots are sent instead, they need to be able to make intelligent decisions to avoid damaging themselves or the structures around which they operate.
Kurniawati offered the example of an undersea robot inspecting the support beams and struts of an offshore oil rig. “They might have the blueprint, but once the robot is put under water, the algorithm needs to account for unexpected things” such as underwater currents, sea life, or obstructions. It may need to alter its mission and take a different route if the planned route is unsafe or parts of structures are occluded.
The $10 million project is funded by The ARC Research Hub in Intelligent Robotic Systems for Real-Time Asset Management. Trevizan serves on this research team as well, along with Associate Professor Patrik Haslum.
Professor Ian Manchester, from the University of Sydney, heads up the project, which also includes Queensland University of Technology (QUT) and multiple private companies. Kurniawati’s ANU team will design the robust planning and decision-making capability to help robots autonomously perform data gathering to build digital twins of infrastructures and maps of the environment.
The same underlying mathematical concepts that make a robot co-pilot possible can be employed to enable undersea robots, flying drones, and robots operating in outer space. “But, to make the concepts work effectively and efficiently in the different domains, we do need additional work to handle all the additional computational issues for each new machine,” Kurniawati said. “If a given robot has especially complex dynamics, it would usually require solving complex differential equations, which can create a bottleneck for computation capability.”
Early breakthroughs stand the ‘Test of Time’
Both of the ARC-funded projects will expand upon more than a decade’s worth of research in robust decision-making for robots — a journey Kurniawati began as a PhD candidate at the National University of Singapore (NUS).
One of her first papers in this domain was recently recognised with the Robotics: Science and Systems 2021 Test of Time Award, assessed based on the impact that research has had in robotics for over a period of ten years.
The 2008 paper is heralded for “pioneering work on developing an effective approximation algorithm that enables robots to operate effectively under uncertain and dynamic situations” which had previously been unattainable due to high computational complexity. In so doing, the Kurniawati overcame a barrier that had hampered robust decision making for more than 4 decades.
Professor David Hsu and Professor Wee Sun Lee were Kurniawati’s co-authors as well as post-doctoral research fellowship supervisors at NUS. Hsu had also served as her PhD advisor.
Kurniawati grew up in Indonesia where she was a fan of the earless robotic cat, Doraemon, from a Japanese manga series, as well as the robots from Star Wars films. She earned her undergraduate degree in Computer Science from the University of Indonesia in 2001.
During her doctoral work at NUS, Kurniawati did an internship at INRIA — France’s National Institute for Research in Digital Science and Technology — and then worked in the United States as a Research Scientist at the Singapore-MIT Alliance for Research and Technology, MIT before coming to Australia to join the School of ITEE, University of Queensland (UQ) in 2012.
It was Thiébaux who recruited Kurniawati to come to ANU in 2019 having met her at the Dagstuhl computer science conference in Germany. “Sylvie contacted me about the [Computer Science Futures Fellowship grant] and asked if I’d be interested. It looked like a good opportunity, and so I applied,” Kurniawati said.
Once Kurniawati had accepted a position and arrived in Canberra, the School of Computing nominated her for the ANU Futures Fellowship grant as well. The two fellowships enabled Kurniawati to set up her robotics lab and maintain a staff, which she said was “vital for the fundamental research I do and for starting a fruitful industry engagement”.
Kurniawati and Thiébaux now work together on the helicopter co-piloting research project, which began when one of Safran’s program managers, Christophe Guettier, visited Canberra from France in the spring of 2019. Chris Dennis, in the CECS industry partnership team at the time, learned of Guettier’s visit from “a connection of a connection”, and arranged for him to meet ANU academics working in the same field, including Kurniawati and Thiébaux.
“Luckily, I was able to get a few of them in the room with Christophe that afternoon, and really the rest is history!” Dennis said. “Hanna’s research really speaks for itself, the folks from Safran could see there was a lot of alignment in the sorts of problems they were trying to tackle.”
The ANU/Safran partnership began soon after. They applied for the ARC Linkage grant in early 2020 but were not successful. Safran began funding the project on its own and they applied again later that year. With preliminary findings to tout and a committed industry partner, the application was well-received by ARC which announced substantial funding in July 2021.
Kurniawati is now an Australian citizen but still misses Indonesian food. She is always on the lookout for new Indonesian restaurants in Canberra.
She said she enjoys teaching at ANU. “It keeps you young,” she said with a grin, adding that her students “ask questions that you might not have thought about before”.
With two newly-funded projects getting underway, Kurniawati is looking to bring on a new Research Fellow and fill 3 PhD positions soon and will open additional positions in the coming years. She said that teaching has kept her in direct contact with a valuable talent pool.
“Some of my PhD students were my honors year students. Once they finish their Phd, some of them might want to continue for a little bit longer as a postdoc with my group before launching their own respective careers,” she said, charting a course for her proteges that she herself took not long ago.