MIT Research News' Journal
 
[Most Recent Entries] [Calendar View]

Monday, September 10th, 2018

    Time Event
    12:00a
    Robots can now pick up any object after inspecting it

    Humans have long been masters of dexterity, a skill that can largely be credited to the help of our eyes. Robots, meanwhile, are still catching up.

    Certainly there’s been some progress: For decades, robots in controlled environments like assembly lines have been able to pick up the same object over and over again. More recently, breakthroughs in computer vision have enabled robots to make basic distinctions between objects. Even then, though, the systems don’t truly understand objects’ shapes, so there’s little the robots can do after a quick pick-up.  

    In a new paper, researchers from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL), say that they’ve made a key development in this area of work: a system that lets robots inspect random objects, and visually understand them enough to accomplish specific tasks without ever having seen them before.

    The system, called Dense Object Nets (DON), looks at objects as collections of points that serve as sort of visual roadmaps. This approach lets robots better understand and manipulate items, and, most importantly, allows them to even pick up a specific object among a clutter of similar — a valuable skill for the kinds of machines that companies like Amazon and Walmart use in their warehouses.

    For example, someone might use DON to get a robot to grab onto a specific spot on an object, say, the tongue of a shoe. From that, it can look at a shoe it has never seen before, and successfully grab its tongue.

    "Many approaches to manipulation can’t identify specific parts of an object across the many orientations that object may encounter,” says PhD student Lucas Manuelli, who wrote a new paper about the system with lead author and fellow PhD student Pete Florence, alongside MIT Professor Russ Tedrake. “For example, existing algorithms would be unable to grasp a mug by its handle, especially if the mug could be in multiple orientations, like upright, or on its side."

    The team views potential applications not just in manufacturing settings, but also in homes. Imagine giving the system an image of a tidy house, and letting it clean while you’re at work, or using an image of dishes so that the system puts your plates away while you’re on vacation.

    What’s also noteworthy is that none of the data was actually labeled by humans. Instead, the system is what the team calls “self-supervised,” not requiring any human annotations.

    Two common approaches to robot grasping involve either task-specific learning, or creating a general grasping algorithm. These techniques both have obstacles: Task-specific methods are difficult to generalize to other tasks, and general grasping doesn’t get specific enough to deal with the nuances of particular tasks, like putting objects in specific spots.

    The DON system, however, essentially creates a series of coordinates on a given object, which serve as a kind of visual roadmap, to give the robot a better understanding of what it needs to grasp, and where.

    The team trained the system to look at objects as a series of points that make up a larger coordinate system. It can then map different points together to visualize an object’s 3-D shape, similar to how panoramic photos are stitched together from multiple photos. After training, if a person specifies a point on a object, the robot can take a photo of that object, and identify and match points to be able to then pick up the object at that specified point.

    This is different from systems like UC-Berkeley’s DexNet, which can grasp many different items, but can’t satisfy a specific request. Imagine a child at 18 months old, who doesn't understand which toy you want it to play with but can still grab lots of items, versus a four-year old who can respond to "go grab your truck by the red end of it.”

    In one set of tests done on a soft caterpillar toy, a Kuka robotic arm powered by DON could grasp the toy’s right ear from a range of different configurations. This showed that, among other things, the system has the ability to distinguish left from right on symmetrical objects.

    When testing on a bin of different baseball hats, DON could pick out a specific target hat despite all of the hats having very similar designs — and having never seen pictures of the hats in training data before.

    “In factories robots often need complex part feeders to work reliably,” says Florence. “But a system like this that can understand objects’ orientations could just take a picture and be able to grasp and adjust the object accordingly.”

    In the future, the team hopes to improve the system to a place where it can perform specific tasks with a deeper understanding of the corresponding objects, like learning how to grasp an object and move it with the ultimate goal of say, cleaning a desk.

    The team will present their paper on the system next month at the Conference on Robot Learning in Zürich, Switzerland.

    4:00p
    Interns at the forefront of new technology

    MIT Materials Research Laboratory (MRL) interns covered a wide gamut of challenges this summer, working with materials as soft as silk to as hard as iron and at temperatures from as low as that of liquid helium (-452.47 degrees Fahrenheit) to as high as that of melted copper (1,984 F). 

    Summer Scholars and other interns participated on the MIT campus through the MRL’s Materials Research Science and Engineering Center, with support from the National Science Foundation, the AIM Photonics Academy, the MRL Collegium, and the Guided Academic Industry Network (GAIN) program. 

    Mid-infrared detectors

    Simon Egner, from the University of Illinois at Urbana-Champaign, made samples of lead tin telluride to detect mid-infrared light at wavelengths from 4 to 7 microns for integrated photonic applications. Egner measured several materials properties of the samples, including the concentration and mobility of electrons. “One thing we have come up with recently is adding lead oxide to try to decrease the amount of noise we get when sensing light with our detectors,” Egner says.

    Lead tin telluride is an alloy of lead telluride and tin telluride, explains Peter Su, a materials science and engineering graduate student in the lab of MIT Materials Research Laboratory Principal Research Scientist Anuradha Agarwal. “If you have a lot of carriers already present in your material, you get a lot of extra noise, a lot of background signal, above which it’s really hard to detect the new carriers generated by the light striking your material,” Su says. “We’re trying to lower that noise level by lowering the carrier concentration and we’re trying to do that by adding lead oxide to that alloy.”

    Thin films for photonics

    Summer Scholar Alvin Chang, from Oregon State University, created chalcogenide thin films with non-linear properties for photonics applications. He worked with postdoc Samuel Serna in the lab of associate professor of materials science and engineering Juejun Hu. Chang varied the thickness of two different compositions, one of germanium, antimony and sulfur (GSS) and the other of germanium, antimony, and selenium (GSSE), creating a gradient, or ratio, between the two across the length of the film.

    “The GSS and GSSE both have different advantages and disadvantages,” Chang explains. “We're hoping that by merging the two together in a film we can sort of optimize both their advantages and disadvantages so that they would be complementary with each other.”

    These materials, known as chalcogenide glasses, can be used for infrared sensing and imaging. Anyone interested in learning more about Chang's work can watch this video.

    Nanocomposite assembly

    Both Roxbury Community College chemistry and biotechnology Professor Kimberly Stieglitz and Roxbury Community College student Credoritch Joseph worked in the lab of assistant professor in materials science and engineering Robert J. Macfarlane. The Macfarlane Lab grafts DNA to nanoparticles, which enable precise control over self-assembly of molecular structures. The lab is also creating a new class of chemical building blocks that it alls Nanocomposite Tectons, or NCTs, which present new opportunities for self-assembly of composite materials.

    Joseph learned the multi-step process of creating self-assembled DNA-nanoparticle aggregates, and used the ones he preparted to study the stability of the aggregates when exposed to different chemicals. Stieglitz created NCTs consisting of clusters of gold nanoparticles with attached polymers and examined their melting behavior in polymer solutions. "They're actually nanoparticles that are linked together through hydrogen bonding networks," Stieglitz explains.

    Strengthening aerospace composites

    Abigail Nason, from the University of Florida, studied the potential benefits of incorporating carbon nanotubes into carbon fiber reinforced plastic [CFRP] via a process termed “nanostitching” in the lab of Brian L. Wardle, professor of aeronautics and astronautics.

    Bundles of carbon microfibers, which are known as tows, are used to make sheets of aerospace-grade carbon fiber reinforced plastic. Working with graduate student Reed Kopp, Nason took 3-D scans of composite laminate samples to reveal their structure. Areas between sheets of the laminate are called the interlaminar region. Traditional composites have no reinforcement in this interlaminar region, and carbon nanotubes provide nano-scale fiber reinforcement in the nano-stitch version.

    Kopp notes that despite the high level of resolution required to elucidate an intricate architecture of micro-scale features, the 3-D scans can’t distinguish the carbon nanotubes from the epoxy resin because they have similar density and elemental composition. “Since they absorb X-rays similarly, we can’t actually detect X-ray interaction differences that would indicate the locations of reinforcing carbon nanotube forests, but we can visualize how they affect the shape of the interlaminar region, such as how they may push fibers apart and change the shape of inherent resin-rich regions caused during carbon fiber reinforced plastic layer manufacturing.”

    Nason adds: “It’s really interesting to see that there isn’t a lot of information out there about how composites fail and why they fail the way they do. But it’s really cool and interesting to be at the forefront of seeing this new technology and being able to look so closely at the composite layers and quantifying critical micro-scale material features that influence failure.” 

    Synthesizing electronic materials

    Summer Scholar Michael Molinski, from the University of Rhode Island, and Roxbury Community College student Bruce Quinn worked in the lab of assistant professor of materials science and engineering Rafael Jaramillo. Working with graduate students Stephen Filippone and Kevin Ye, both Molinski and Quinn made solid materials, producing powders of compounds such as barium zirconium sulfide, which are desireable for their optical and electrical properties. 

    The process involves mixing together the chemical ingredients to produce the powders in a quartz tupe in the absense of air and sealing it. The first GAIN program participant, Quinn hot pressed the powders into pellets. Molinski also grew crystals, and both examined their powders with X-ray diffraction.

    Developing multiple sclerosis models

    Summer Scholar Fernando Nieves Muñoz, from the University of Puerto Rico at Mayagüez, worked in the lab of Krystyn Van Vliet, the Michael (1949) and Sonja Koerner Professor of Materials Science and Engineering, to develop mechanical models of multiple sclerosis (MS) lesions. Nieves Muñoz worked closely with research scientist Anna Jagielska and chemical engineering graduate student Daniela Espinosa-Hoyos.

    “We are trying to find a way to stimulate repair of myelin in MS patients so that neurological function can be restored. To better understand how remyelination works, we are developing polymer-based materials to engineer models of MS lesions that mimic mechanical stiffness of real lesions in the brain,” Jagielska explains. 

    Nieves Muñoz used stereolithography 3-D printing to create cross-linked polymers with varying degrees of mechanical stiffness and conducted atomic force microscopy studies to determine the stiffness of his samples. “Our long-term goal is to use these models of lesions and brain tissue to develop drugs that can stimulate myelin repair,” Nieves Muñoz says. “As a mechanical engineering major, it has been exciting to work and learn from people with diverse backgrounds.”

    Other MIT Materials Research Laboratory interns tackled projects including superconducting thin films, quantum dots for solar, spinning particles with magnetism, carbon-activated silk fibers, water-based iron flow batteries, and polymer-based neuro fibers. 

    A version of this post, including additional MRL summer intern success stories, originally appeared on the Materials Research Laboratory website.

    << Previous Day 2018/09/10
    [Calendar]
    Next Day >>

MIT Research News   About LJ.Rossia.org