Expansion of robot perception | News put

by Brenden Burgess

When you buy through links on our site, we may earn a commission at no extra cost to you. However, this does not influence our evaluations.

The robots have gone a long way from the Roomba. Today, drones are starting to deliver from door to door, autonomous cars sail by certain roads, Robo-Chiens help the first stakeholders, and even more robots make backflips and help the factory floor. However, Luca Carlone thinks that the best is yet to come.

Carlone, who recently received the mandate as an associate professor of the aeronautical and astronautics department of MIT (Aeroastro), directs the Spark Lab, where he and his students fill a key gap between humans and robots: perception. The group does theoretical and experimental research, all to extend the consciousness of a robot from its environment in a way that approaches human perception. And perception, as Carlone often says, is more than detection.

Although robots have grown up by step as a giant in terms of ability to detect and identify objects in their environment, they still have a lot to learn when it comes to giving a level of level higher than their environment. As human beings, we perceive objects with an intuitive sense not only of their forms and labels, but also of their physique – how they could be manipulated and moved – and how they relate to each other, their broader environment and ourselves.

This kind of perception at the level of man is what Carlone and his group hope to transmit to robots, in a way that allows them to interact safely and transparently with people in their homes, their workplaces and other unstructured environments.

Since he joined the Faculty of MIT in 2017, Carlone led her team to develop and apply perception and stage understanding algorithms for various applications, including autonomous research and rescue vehicles, drones that can collect and manipulate objects on the fly and autonomous cars. They could also be useful for domestic robots which follow controls in natural language and have potentially even anticipate the needs of man according to the contextual clues of higher level.

“Perception is a big bottleneck to get robots to help us in the real world,” explains Carlone. “If we can add cognition and reasoning elements to the robot perception, I think they can do a lot of good.”

Expanding horizons

Carlone was born and grew up near Salerne, Italy, near the picturesque coast Amalfaitaine, where he was the youngest of three boys. Her mother is a primary school retirement who taught mathematics, and her father is a retired professor and publisher, who has always adopted an analytical approach to her historical research. The brothers may have unconsciously adopted the mentality of their parents, because all three continued to be an engineers – the two older ones continued electronics and mechanical genius, while Carlone landed on robotics or mecatronics, as we knew at the time.

However, he did not come to the field that at the end of his undergraduate studies. Carlone frequented the University Polytechnique de Turin, where it was initially focused on theoretical work, in particular on control theory – an area that applies mathematics to develop algorithms that automatically control the behavior of physical systems, such as electrical networks, plans, cars and robots. Then, during her last year, Carlone signed up for a course on robotics which explored the progress of manipulation and how robots can be programmed to move and operate.

“It was love at first sight. Using algorithms and mathematics to develop the brain of a robot and make it move and interact with the environment is one of the most fulfilling experiences, ”explains Carlone. “I immediately decided that this is what I want to do in life.”

He continued a double degree program at Polytechnique de Turin University and at Polytechnic University in Milan, where he obtained master's degree in mecatronics and automation engineering, respectively. As part of this program, called Alta Scuola Policnica, Carlone also took management courses, in which he and students of various university horizons had to team up to conceptualize, build and establish a marketing argument for a new product design. The Carlone team has developed a contactless table lamp designed to follow a user's hand controls. The project prompted him to think about engineering from different angles.

“It was like having to speak different languages,” he says. “It was an early exhibition to the need to look beyond the engineering bubble and think about how to create technical work that can have an impact on the real world.”

The next generation

Carlone stayed in Turin to finish her doctorate in mechatronics. Meanwhile, he had the freedom to choose a subject of thesis, which he did, as he remembers: “a little naively”.

“I explored a subject that the community considered it well understood, and for which many researchers thought there was nothing more to say.” Said Carlone. “I underestimated how the subject was established and I thought I could still contribute something new to it, and I had the chance to do it.”

The subject in question was “the location and simultaneous cartography” or Slam – the problem of generation and updating an environment map of a robot while simultaneously keeping the location of the robot in this environment. Carlone found a way to crop the problem, so algorithms could generate more precise cards without having started with an initial assumption, as most of the Slam methods did at the time. His work helped to open a field where most of the roboticians thought that we could not do better than existing algorithms.

“Slam is to understand the geometry of things and how a robot moves among these things,” explains Carlone. “Now, I am part of a community that asks me, what is the next generation of slam?”

In search of an answer, he accepted a postdoctoral position in Georgia Tech, where he plunged into coding and computer vision – an area which, retrospectively, may have been inspired by a brush with blindness: as he finished his doctorate in Italy, he underwent a medical complication that seriously affected his vision.

“For a year, I could have easily lost an eye,” explains Carlone. “It's something that made me think about the importance of vision and artificial vision.”

He was able to receive good medical care, and the condition has completely resolved, so that he could continue his work. To Georgia Tech, his advisor, Frank DellaertHe has shown him ways to code in computer vision and formulate elegant mathematical representations of complex three -dimensional problems. His advisor was also one of the first to develop a library of open source slams, called GTSAMthat Carlone quickly admitted to being an invaluable resource. More broadly, he has seen that making software available to everyone has unlocking enormous potential for progress in robotics as a whole.

“Historically, progress in Slam has been very slow, because people have kept their owner codes, and each group had essentially reached zero,” explains Carlone. “Then, the open source pipelines began to appear, and that has changed the situation, which has largely motivated the progress we have seen in the past 10 years.”

Ai Spatial

After Georgia Tech, Carlone came to MIT in 2015 as a laboratory post-doctorator for information and decision-making systems (LIDS). Meanwhile, he collaborated with Sertac Karaman, professor of aeronautics and astronautics, in the development of software to help drones the size of a palm to navigate in their environment using very little on -board power. A year later, he was promoted to the researcher, then in 2017, Carlone accepted a teacher position in Aeroastro.

“One thing that I fell in love with it is that all the decisions are motivated by questions like: what are our values?” What is our mission? These are never low level gains. Motivation really concerns how to improve society, ”explains Carlone. “As a state -of -the -art state, it was very refreshing.”

Today, the Carlone group develops means to represent the environment of a robot, beyond characterizing their geometric shape and its semantics. It uses in -depth learning and large languages ​​models to develop algorithms that allow robots to perceive their environment through a higher level lens, so to speak. Over the past six years, his laboratory has released more than 60 open source repositorywhich are used by thousands of researchers and practitioners from around the world. Most of his work is part of a larger and emerging field known as “space IA”.

“Spatial AI is like a slam on steroids,” explains Carlone. “In a word, it has to do with robots to think and understand the world as humans do, in a way that can be useful.”

It is a huge company that could have large -scale impacts, in terms of allowing more intuitive and interactive robots to help at home, in the workplace, on the roads and in distant and potentially dangerous areas. Carlone says there will be a lot of work in advance, in order to get closer to the way humans perceive the world.

“I have twin daughters of 2 years, and I see them manipulating objects, carrying 10 different toys at the same time, sailing through the congested parts with ease and quickly adapting to new environments. The robot perception cannot yet correspond to what a toddler can do, ”explains Carlone. “But we have new tools in the arsenal. And the future is brilliant.”

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.