Claire chatted to Dr Maria Bauza Villalonga from DeepMind all about robot learning, transferable skills, and general AI.
Maria Bauza Villalonga is a research scientist at DeepMind. In 2022, she earned her PhD in Robotics at the Massachusetts Institute of Technology, working with Prof. Alberto Rodriguez. Her research focuses on achieving precise robotic generalization by learning probabilistic models of the world that allow robots to reuse their skills across multiple tasks with high success. Maria has received several fellowships including Facebook, NVIDIA, and LaCaixa.
Researchers from Bochum and Saarbrücken have detected security vulnerabilities, some of them serious, in several drones made by the manufacturer DJI. These enable users, for example, to change a drone's serial number or override the mechanisms that allow security authorities to track the drones and their pilots. In special attack scenarios, the drones can even be brought down remotely in flight.
As technology advances rapidly, the gap between the speed of policy development and technological change is becoming more prominent. This is particularly true in robotics, where current legislation often lags behind and fails to adequately frame robot technologies. This gap increases legal uncertainty and poses safety risks, as developers may not know which regulatory frameworks to follow. The resulting technology may thus perform poorly and lead to biases and discrimination.
If animals and insects can jump across grass and sand, why can't robots? Sarah Bergbreiter, Professor of Mechanical Engineering, has found that researchers don't have to look far to enable this in robots. Existing latch mechanisms that were once thought of as an 'on' or 'off' switch to release stored energy can also be used to control jump performance across a wide range of terrains.
Are drones really energy efficient? That depends on the average wind speed. An Argonne study compares drone energy usage to diesel trucks and electric vehicles and aims to help industry determine if they could save money with drone delivery.
The use of robots in automotive manufacturing processes enables a very high degree of repeatability and precision, resulting in process optimization and increased productivity. This is how the increasingly high industry standards demanded by manufacturers are met.
Researchers at Istituto Italiano di Tecnologia (IIT-Italian Institute of Technology) in Genoa has realized a new soft robot inspired by the biology of earthworms, which is able to crawl thanks to soft actuators that elongate or squeeze, when air passes through them or is drawn out.
These systems can be configured and supplied as individual components or as complete integrated motion systems with custom interfaces, cable management solutions, vacuum feedthroughs and application-optimized integrated granite and aluminum structures.
The final episode of our RoboHouse Interview Trilogy: ‘The Working Life of the Robotics Engineer’ interviews Srimannarayana Baratam. Sriman, as he is also called, co-founded the company Perciv.ai just two months after graduating. Rens van Poppel explores his journey so far.
Perciv.ai claims that AI-driven machine perception could become affordable to everyone. When was this vision formed, and how did it come about? Sriman points to the period right after his graduation. He says it was pivotal for building trust with partners, and consensus with effective communication. Because starting your own company comes with a lot of challenges.
Srimannarayana Baratam was the first to graduate from the MSc Robotics at the TU Delft. The Master’s degree programme was newly launched in 2020 and aims to train students who can guide the industry towards a kind of robotisation that promotes and reinforces workplace attractiveness.
“It is important for to find partners you can trust,” says Sriman. “You need to understand each other’s motivation and commitment. You need to assess what real value does this person add to the team.”
Coming from an automotive background in India, Sriman’s master’s thesis investigated the use of radar and cameras to protect vulnerable people in urban environments. He co-founded the start-up with his supervisor, Dr András Pálffy, and Balazs Szekeres, another robotics student who heard about the project. In the two months after his graduation, Sriman and his co-founders came together to focus full-time on their vision for Perciv.ai.
“In July and August we sat down and discussed the vision between the three of us,” he says. This period also led to tough conversations, ranging from finance to market strategy. “When finally the main questions were sorted out, you just got to take that leap of faith together. This leap of faith proved fruitful, seeing that the high level of trust resulted in a high level of productivity over the past five months.”
Since then Perciv.ai went on to win the NWO take-off phase 1 grant, got their own office and workspace in RoboHouse, signed a contract with an unmanned aerial vehicle (UAV) company and in doing so, generated their first sales revenue.
“This does not mean that there are no more heavy debates,” Sriman adds. “We all share the same vision, but in order to reach our goal of a sustainable and affordable product, we sometimes have different ideas on what that final product should look like.”
Sriman’s passion for robotics and the company’s goals is palpable: “We want to make machine perception technology available to all.”
While robots have become increasingly advanced over the past few years, most of them are still unable to reliably navigate very crowded spaces, such as public areas or roads in urban environments. To be implemented on a large-scale and in the smart cities of the future, however, robots will need to be able to navigate these environments both reliably and safely, without colliding with humans or nearby objects.
