Catherine Iacobo named industry co-director for MIT Leaders for Global Operations
Cathy Iacobo, a lecturer at the MIT Sloan School of Management, has been named the new industry co-director for the MIT Leaders for Global Operations (LGO) program. Read more
2 from MIT Sloan make Forbes 30 Under 30 List
According to Forbes the annual lists spotlight “revolutionaries … changing the course — and the face — of business and society.”
January 10, 2020 | More
Preventing energy loss in windows
“The choice of windows in a building has a direct influence on energy consumption,” says Nicholas Fang, professor of mechanical engineering and current LGO Thesis Advisor. “We need an effective way of blocking solar radiation.”
In the quest to make buildings more energy efficient, windows present a particularly difficult problem. According to the U.S. Department of Energy, heat that either escapes or enters windows accounts for roughly 30 percent of the energy used to heat and cool buildings. Researchers are developing a variety of window technologies that could prevent this massive loss of energy.
January 6, 2020 | More
Sixteen grad students named to the Siebel Scholars class of 2020
LGO ’20 Hans Nowak is among the 2020 cohort of Siebel Scholars hailing from the world’s top graduate programs in bioengineering, business, computer science, and energy science. They were recognized at a luncheon and awards ceremony on campus on Oct. 31.
“You’re among a very select group of students to receive this honor,” Anantha Chandrakasan, dean of the School of Engineering and Vannevar Bush Professor of Electrical Engineering and Computer Science, told the students. “Your department heads obviously think very highly of your accomplishments.”
Honored for their academic achievements, leadership, and commitments to addressing crucial global challenges, the MIT students are among 93 Siebel Scholars from 16 leading institutions in the United States, China, France, Italy, and Japan.
Siebel Scholars each receive an award of $35,000 to cover their final year of study. In addition, they will join a community of more than 1,400 past Siebel Scholars, including about 260 from MIT, who serve as advisors to the Thomas and Stacy Siebel Foundation and collaborate “to find solutions to society’s most pressing problems,” according to the foundation.
Past Siebel Scholars have launched more than 1,100 products, received at least 370 patents, published nearly 40 books, and founded at least 150 companies, among other achievements, according to the Siebel Scholars Foundation, which administers the program.
MIT’s 2020 class of Siebel Scholars includes:
- Katie Bacher, Department of Electrical Engineering and Computer Science
- Alexandra (Allie) Beizer, MIT Sloan School of Management
- Sarah Bening, Department of Biological Engineering
- Allison (Allie) Brouckman, MIT Sloan School of Management
- Enric Boix, Department of Electrical Engineering and Computer Science
- M. Doga Dogan, Department of Electrical Engineering and Computer Science
- Jared Kehe, Department of Biological Engineering
- Emma Kornetsky, MIT Sloan School of Management
- Kyungmi Lee, Department of Electrical Engineering and Computer Science
- Graham Leverick, Department of Mechanical Engineering
- Lauren Milling, Department of Biological Engineering
- Hans Nowak, MIT Sloan School of Management
- Lauren Stopfer, Department of Biological Engineering
- Jon Tham, Sloan School of Management
- Andrea Wallace, Department of Biological Engineering
- Clinton Wang, Department of Electrical Engineering and Computer Science
November 19, 2019 | More
Practicing for a voyage to Mars
If you want to make the long voyage to Mars, you first have to train and rehearse, and MIT LGO alumnus Barret Schlegelmilch SM ’18, MBA ’18 is doing just that. He recently commanded a 45-day practice mission living and working with three other would-be astronauts in a cramped simulated spaceship.
NASA’s Human Exploration Research Analog (HERA) analog mission “departed” last spring for a trip to Phobos, the larger of the two moons of Mars. It was the second of four planned missions to Phobos in the mock spacecraft located at the Johnson Space Center in Houston. The goal is to study the physiological and psychological effects of extended isolation and confinement, team dynamics, and conflict resolution.
While on the mission, Schlegelmilch and three other crew me
November 1, 2019 | More
New leadership for Bernard M. Gordon-MIT Engineering Leadership Program
Olivier de Weck, frequent LGO advisor, professor of aeronautics and astronautics and of engineering systems at MIT, has been named the new faculty co-director of the Bernard M. Gordon-MIT Engineering Leadership Program (GEL). He joins Reza Rahaman, who was appointed the Bernard M. Gordon-MIT Engineering Leadership Program industry co-director and senior lecturer on July 1, 2018.
“Professor de Weck has a longstanding commitment to engineering leadership, both as an educator and a researcher. I look forward to working with him and the GEL team as they continue to strengthen their outstanding undergraduate program and develop the new program for graduate students,” says Anantha Chandrakasan, dean of the MIT School of Engineering and the Vannevar Bush Professor of Electrical Engineering and Computer Science.
A leader in systems engineering, de Weck researches how complex human-made systems such as aircraft, spacecraft, automobiles, and infrastructures are designed, manufactured, and operated. By investigating their lifecycle properties, de Weck and members of his research group have developed a range of novel techniques broadly adopted by industry to maximize the value of these systems over time.
August 1, 2019 | More
Building the tools of the next manufacturing revolution
John Hart, an associate professor of mechanical engineering at MIT, LGO adviser, and the director of the Laboratory for Manufacturing and Productivity and the Center for Additive and Digital Advanced Production Technologies, is an expert in 3-D printing, also known as additive manufacturing, which involves the computer-guided deposition of material layer by layer into precise three-dimensional shapes. (Conventional manufacturing usually entails making a part by removing material, for example through machining, or by forming the part using a mold tool.)
Hart’s research includes the development of advanced materials — new types of polymers, nanocomposites, and metal alloys — and the development of novel machines and processes that use and shape materials, such as high-speed 3-D printing, roll-to-roll graphene growth, and manufacturing techniques for low-cost sensors and electronics.
June 19, 2019 | More
LGO Best Thesis 2019 for Big Data Analysis at Amgen, Inc.
After the official MIT commencement ceremonies, Thomas Roemer, LGO’s executive director, announced the best thesis winner at LGO’s annual post-graduation celebration. This year’s winner was Maria Emilia Lopez Marino (Emi), who developed a predictive framework to evaluate and assess the impact of raw material attributes on the manufacturing process at Amgen. Thesis readers described Marino’s project as an “extremely well-written thesis. Excellent coverage of not only the project, but also the industry as a whole.”
Applying MIT knowledge in the real world
Marino, who earned her MBA and SM in Civil and Environmental Engineering, completed her six-month LGO internship project at Amgen, Inc. For her project, Marino developed a new predictive framework through machine learning techniques to assess the impact of raw material variability on the performance of several commercial processes of biologics manufacturing. Finding this solution represents a competitive advantage for biopharmaceutical leaders. The results from her analysis showed an 80% average accuracy on predictions for new data. Additionally, the framework she developed is the starting point of a new methodology towards material variability understanding in the manufacturing process for the pharmaceutical industry.
Each year, the theses are nominated by faculty advisors and then reviewed by LGO alumni readers to determine the winner. Thesis advisor and Professor Roy Welsch stated Emi “understood variation both in a statistical sense and in manufacturing in the biopharmaceutical industry and left behind highly accurate and interpretable models in a form that others can use and expand. We hope she will share her experiences with us in the future at LGO alumni reunions and on DPT visits.”
Marino, who earned her undergraduate degree Chemical Engineering from the National University of Mar Del Plata in Argentina, has accepted a job offer with Amgen in Puerto Rico.
June 11, 2019 | More
The tenured engineers of 2019
The School of Engineering has announced that 17 members of its faculty have been granted tenure by MIT, including 3 LGO advisors: Saurabh Amin, Kerri Cahoy, and Julie Shah.
“The tenured faculty in this year’s cohort are a true inspiration,” said Anantha Chandrakasan, dean of the School of Engineering. “They have shown exceptional dedication to research and teaching, and their innovative work has greatly advanced their fields.”
This year’s newly tenured associate professors are:
Antoine Allanore, in the Department of Materials Science and Engineering, develops more sustainable technologies and strategies for mining, metal extraction, and manufacturing, including novel methods of fertilizer production.
Saurabh Amin, in the Department of Civil and Environmental Engineering, focuses on the design and implementation of network inspection and control algorithms for improving the resilience of large-scale critical infrastructures, such as transportation systems and water and energy distribution networks, against cyber-physical security attacks and natural events.
Emilio Baglietto, in the Department of Nuclear Science and Engineering, uses computational modeling to characterize and predict the underlying heat-transfer processes in nuclear reactors, including turbulence modeling, unsteady flow phenomena, multiphase flow, and boiling.
Paul Blainey, the Karl Van Tassel (1925) Career Development Professor in the Department of Biological Engineering, integrates microfluidic, optical, and molecular tools for application in biology and medicine across a range of scales.
Kerri Cahoy, the Rockwell International Career Development Professor in the Department of Aeronautics and Astronautics, develops nanosatellites that demonstrate weather sensing using microwave radiometers and GPS radio occultation receivers, high data-rate laser communications with precision time transfer, and active optical imaging systems using MEMS deformable mirrors for exoplanet exploration applications.
Juejun Hu, in the Department of Materials Science and Engineering, focuses on novel materials and devices to exploit interactions of light with matter, with applications in on-chip sensing and spectroscopy, flexible and polymer photonics, and optics for solar energy.
Sertac Karaman, the Class of 1948 Career Development Professor in the Department of Aeronautics and Astronautics, studies robotics, control theory, and the application of probability theory, stochastic processes, and optimization for cyber-physical systems such as driverless cars and drones.
R. Scott Kemp, the Class of 1943 Career Development Professor in the Department of Nuclear Science and Engineering, combines physics, politics, and history to identify options for addressing nuclear weapons and energy. He investigates technical threats to nuclear-deterrence stability and the information theory of treaty verification; he is also developing technical tools for reconstructing the histories of secret nuclear-weapon programs.
Aleksander Mądry, in the Department of Electrical Engineering and Computer Science, investigates topics ranging from developing new algorithms using continuous optimization, to combining theoretical and empirical insights, to building a more principled and thorough understanding of key machine learning tools. A major theme of his research is rethinking machine learning from the perspective of security and robustness.
Frances Ross, the Ellen Swallow Richards Professor in the Department of Materials Science and Engineering, performs research on nanostructures using transmission electron microscopes that allow researchers to see, in real-time, how structures form and develop in response to changes in temperature, environment, and other variables. Understanding crystal growth at the nanoscale is helpful in creating precisely controlled materials for applications in microelectronics and energy conversion and storage.
Daniel Sanchez, in the Department of Electrical Engineering and Computer Science, works on computer architecture and computer systems, with an emphasis on large-scale multi-core processors, scalable and efficient memory hierarchies, architectures with quality-of-service guarantees, and scalable runtimes and schedulers.
Themistoklis Sapsis, the Doherty Career Development Professor in the Department of Mechanical Engineering, develops analytical, computational, and data-driven methods for the probabilistic prediction and quantification of extreme events in high-dimensional nonlinear systems such as turbulent fluid flows and nonlinear mechanical systems.
Julie Shah, the Boeing Career Development Professor in the Department of Aeronautics and Astronautics, develops innovative computational models and algorithms expanding the use of human cognitive models for artificial intelligence. Her research has produced novel forms of human-machine teaming in manufacturing assembly lines, healthcare applications, transportation, and defense.
Hadley Sikes, the Esther and Harold E. Edgerton Career Development Professor in the Department of Chemical Engineering, employs biomolecular engineering and knowledge of reaction networks to detect epigenetic modifications that can guide cancer treatment, induce oxidant-specific perturbations in tumors for therapeutic benefit, and improve signaling reactions and assay formats used in medical diagnostics.
William Tisdale, the ARCO Career Development Professor in the Department of Chemical Engineering, works on energy transport in nanomaterials, nonlinear spectroscopy, and spectroscopic imaging to better understand and control the mechanisms by which excitons, free charges, heat, and reactive chemical species are converted to more useful forms of energy, and on leveraging this understanding to guide materials design and process optimization.
Virginia Vassilevska Williams, the Steven and Renee Finn Career Development Professor in the Department of Electrical Engineering and Computer Science, applies combinatorial and graph theoretic tools to develop efficient algorithms for matrix multiplication, shortest paths, and a variety of other fundamental problems. Her recent research is centered on proving tight relationships between seemingly different computational problems. She is also interested in computational social choice issues, such as making elections computationally resistant to manipulation.
Amos Winter, the Tata Career Development Professor in the Department of Mechanical Engineering, focuses on connections between mechanical design theory and user-centered product design to create simple, elegant technological solutions for applications in medical devices, water purification, agriculture, automotive, and other technologies used in highly constrained environments.
June 7, 2019 | More
MIT team places second in 2019 NASA BIG Idea Challenge
An MIT student team, including LGO ’20 Hans Nowak, took second place for its design of a multilevel greenhouse to be used on Mars in NASA’s 2019 Breakthrough, Innovative and Game-changing (BIG) Idea Challenge last month.
Each year, NASA holds the BIG Idea competition in its search for innovative and futuristic ideas. This year’s challenge invited universities across the United States to submit designs for a sustainable, cost-effective, and efficient method of supplying food to astronauts during future crewed explorations of Mars. Dartmouth College was awarded first place in this year’s closely contested challenge.
“This was definitely a full-team success,” says team leader Eric Hinterman, a graduate student in MIT’s Department of Aeronautics and Astronautics (AeroAstro). The team had contributions from 10 undergraduates and graduate students from across MIT departments. Support and assistance were provided by four architects and designers in Italy. This project was completely voluntary; all 14 contributors share a similar passion for space exploration and enjoyed working on the challenge in their spare time.
The MIT team dubbed its design “BEAVER” (Biosphere Engineered Architecture for Viable Extraterrestrial Residence). “We designed our greenhouse to provide 100 percent of the food requirements for four active astronauts every day for two years,” explains Hinterman.
The ecologists and agriculture specialists on the MIT team identified eight types of crops to provide the calories, protein, carbohydrates, and oils and fats that astronauts would need; these included potatoes, rice, wheat, oats, and peanuts. The flexible menu suggested substitutes, depending on astronauts’ specific dietary requirements.
“Most space systems are metallic and very robotic,” Hinterman says. “It was fun working on something involving plants.”
Parameters provided by NASA — a power budget, dimensions necessary for transporting by rocket, the capacity to provide adequate sustenance — drove the shape and the overall design of the greenhouse.
Last October, the team held an initial brainstorming session and pitched project ideas. The iterative process continued until they reached their final design: a cylindrical growing space 11.2 meters in diameter and 13.4 meters tall after deployment.
An innovative design
The greenhouse would be packaged inside a rocket bound for Mars and, after landing, a waiting robot would move it to its site. Programmed with folding mechanisms, it would then expand horizontally and vertically and begin forming an ice shield around its exterior to protect plants and humans from the intense radiation on the Martian surface.
Two years later, when Earth and Mars orbits were again in optimal alignment for launching and landing, a crew would arrive on Mars, where they would complete the greenhouse setup and begin growing crops. “About every two years, the crew would leave and a new crew of four would arrive and continue to use the greenhouse,” explains Hinterman.
To maximize space, BEAVER employs a large spiral that moves around a central core within the cylinder. Seedlings are planted at the top and flow down the spiral as they grow. By the time they reach the bottom, the plants are ready for harvesting, and the crew enters at the ground floor to reap the potatoes and peanuts and grains. The planting trays are then moved to the top of the spiral, and the process begins again.
“A lot of engineering went into the spiral,” says Hinterman. “Most of it is done without any moving parts or mechanical systems, which makes it ideal for space applications. You don’t want a lot of moving parts or things that can break.”
The human factor
“One of the big issues with sending humans into space is that they will be confined to seeing the same people every day for a couple of years,” Hinterman explains. “They’ll be living in an enclosed environment with very little personal space.”
The greenhouse provides a pleasant area to ensure astronauts’ psychological well-being. On the top floor, just above the spiral, a windowed “mental relaxation area” overlooks the greenery. The ice shield admits natural light, and the crew can lounge on couches and enjoy the view of the Mars landscape. And rather than running pipes from the water tank at the top level down to the crops, Hinterman and his team designed a cascading waterfall at
May 24, 2019 | More
MIT team places first in U.S. Air Force virtual reality competition
When the United States Air Force put out a call for submissions for its first-ever Visionary Q-Prize competition in October 2018, a six-person team of 3 MIT students and 3 LGO alumni took up the challenge. Last month, they emerged as a first-place winner for their prototype of a virtual reality tool they called CoSMIC (Command, Sensing, and Mapping Information Center).
The challenge was hosted by the Air Force Research Labs Space Vehicles Directorate and the Wright Brothers Institute to encourage nontraditional sources with innovative products and ideas to engage with military customers to develop solutions for safe and secure operations in space.
April 12, 2019 | More
How to manage the hidden risks in remote work
For many people, COVID-19 began as a distant story. Now it’s a threat filling every corner of life. Markets have plunged, settled uneasily, and plunged again. Sports seasons are suspended. Schools are closing, along with restaurants and bars. People are huddled at home. And companies, amid it all, are struggling to balance employee and public health with the maintenance of basic operations.
March 23, 2020 | More
6 steps to handle supply chain disruption
The rapid spread of the novel coronavirus is crippling supply chains around the world, with companies from Apple to Nintendo announcing shipping and manufacturing delays. Other unanticipated events have caused supply chain upheaval in the last few decades, including the Fukushima nuclear meltdown, 2011 floods in Thailand, SARS and MERS pandemics, and Hurricane Katrina.
But it is inaccurate to compare the coro
March 19, 2020 | More
Writing a new leadership playbook
What will define great leadership in the new digital economy? A new report from MIT Sloan senior lecturer
finds it is a combination of emerging and enduring behaviors, while simultaneously rejecting what Ready calls eroding behaviors.
But it doesn’t stop there: recognizing the behaviors is only the first, small step. The real challenge is in confronting cultural inertia and making new behaviors the norm throughout an organization.
March 9, 2020 | More
How to master two different digital transformations
To thrive in the digital age, companies must undergo two distinct digital transformations. They must become digitized by incorporating digital technology into their core operations, like accounting and invoicing. They also need to be become digital, which means developing a digital platform for the company’s digital offerings.
March 3, 2020 | More
What keeps supply chain execs up at night? Lots of things
Managing supply chains in 2020 is hard. It means always questioning what will happen next, as the globalized economy puts forth an overlapping series of new opportunities and critical disruptions.
We asked leaders in industries enmeshed in global supply chains – apparel, venture capital, computing hardware, and international theaters – about their greatest challenges, the technology changing everything, and supply chain visibility.
February 26, 2020 | More
5 supply chain technologies that deliver competitive advantage
Facing globalization, increased product complexity, and heightened customer demands, companies are taking up advanced technologies to transform their supply chain from a pure operations hub into the epicenter of business innovation.
Using sensors and ever-improving internet connectivity, forward-thinking companies are collecting data at every checkpoint, from the status of raw materials flow to the condition and location of finished goods.
Machine learning, artificial intelligence (AI), and advanced analytics help drive automation and deliver insights that promote efficiencies — making on-the-fly route changes to accelerate product delivery, for example, or swapping out materials to take advantage of better pricing or availability.
February 20, 2020 | More
Supply chain transparency, explained
Understanding your supply chain is more important than ever.
In the wake of reports about slave labor, food contamination, and human rights abuses, consumers are concerned about how their purchases impact their health, their communities, and the world at large.
February 20, 2020 | More
Supply chain resilience in the era of climate change
Climate change can upend supply chains in obvious ways — sudden floods, flash fires — as well as through secondary repercussions like a migrating workforce or infrastructure in need of a retrofit.
And those scenarios and others stand to directly affect a company’s bottom line. According to recent research from the United Nations Development Programme, productivity losses related to climate-change-related workplace disruption in the United States could rise to above $2 trillion by 2030.
February 13, 2020 | More
The psychological, economic, and social costs of air pollution
Air pollution’s toll on human health is well documented. It is the leading cause of mortality in India, contributing to the death of more than 1.6 million people annually. It is responsible for 1.1 million premature deaths each year in China. And in the U.S., about 111 million Americans — 35% of the population — live in counties with unhealthy air, which makes them more susceptible to lung cancer, heart attacks, and strokes.
February 13, 2020 | More
Outlining the case for economic growth as sustainability cure, not curse
For the first 200 or so years of the industrial era, technology and capitalism were sometimes seen as forces that helped economies flourish, to the detriment of the environment. Eventually, some feared, there wouldn’t be enough resources left to sustain humanity.
But Andrew McAfee sees a plot twist: Those same forces are now helping humans reduce their footprint on the planet. McAfee, a research scientist at MIT Sloan, lays out his optimistic case for how the “voracious appetite” of the industrial era led to technology that improves sustainability in his new book, “More from Less: The Surprising Story of How We Learned to Prosper Using Fewer Resources — and What Happens Next.”
January 21, 2020 | More
System trains driverless cars in simulation before they hit the road
A simulation system invented at MIT to train driverless cars creates a photorealistic world with infinite steering possibilities, helping the cars learn to navigate a host of worse-case scenarios before cruising down real streets.
Control systems, or “controllers,” for autonomous vehicles largely rely on real-world datasets of driving trajectories from human drivers. From these data, they learn how to emulate safe steering controls in a variety of situations. But real-world data from hazardous “edge cases,” such as nearly crashing or being forced off the road or into other lanes, are — fortunately — rare.
Some computer programs, called “simulation engines,” aim to imitate these situations by rendering detailed virtual roads to help train the controllers to recover. But the learned control from simulation has never been shown to transfer to reality on a full-scale vehicle.
The MIT researchers tackle the problem with their photorealistic simulator, called Virtual Image Synthesis and Transformation for Autonomy (VISTA). It uses only a small dataset, captured by humans driving on a road, to synthesize a practically infinite number of new viewpoints from trajectories that the vehicle could take in the real world. The controller is rewarded for the distance it travels without crashing, so it must learn by itself how to reach a destination safely. In doing so, the vehicle learns to safely navigate any situation it encounters, including regaining control after swerving between lanes or recovering from near-crashes.
In tests, a controller trained within the VISTA simulator safely was able to be safely deployed onto a full-scale driverless car and to navigate through previously unseen streets. In positioning the car at off-road orientations that mimicked various near-crash situations, the controller was also able to successfully recover the car back into a safe driving trajectory within a few seconds. A paper describing the system has been published in IEEE Robotics and Automation Letters and will be presented at the upcoming ICRA conference in May.
“It’s tough to collect data in these edge cases that humans don’t experience on the road,” says first author Alexander Amini, a PhD student in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “In our simulation, however, control systems can experience those situations, learn for themselves to recover from them, and remain robust when deployed onto vehicles in the real world.”
The work was done in collaboration with the Toyota Research Institute. Joining Amini on the paper are Igor Gilitschenski, a postdoc in CSAIL; Jacob Phillips, Julia Moseyko, and Rohan Banerjee, all undergraduates in CSAIL and the Department of Electrical Engineering and Computer Science; Sertac Karaman, an associate professor of aeronautics and astronautics; and Daniela Rus, director of CSAIL and the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science.
Historically, building simulation engines for training and testing autonomous vehicles has been largely a manual task. Companies and universities often employ teams of artists and engineers to sketch virtual environments, with accurate road markings, lanes, and even detailed leaves on trees. Some engines may also incorporate the physics of a car’s interaction with its environment, based on complex mathematical models.
But since there are so many different things to consider in complex real-world environments, it’s practically impossible to incorporate everything into the simulator. For that reason, there’s usually a mismatch between what controllers learn in simulation and how they operate in the real world.
Instead, the MIT researchers created what they call a “data-driven” simulation engine that synthesizes, from real data, new trajectories consistent with road appearance, as well as the distance and motion of all objects in the scene.
They first collect video data from a human driving down a few roads and feed that into the engine. For each frame, the engine projects every pixel into a type of 3D point cloud. Then, they place a virtual vehicle inside that world. When the vehicle makes a steering command, the engine synthesizes a new trajectory through the point cloud, based on the steering curve and the vehicle’s orientation and velocity.
Then, the engine uses that new trajectory to render a photorealistic scene. To do so, it uses a convolutional neural network — commonly used for image-processing tasks — to estimate a depth map, which contains information relating to the distance of objects from the controller’s viewpoint. It then combines the depth map with a technique that estimates the camera’s orientation within a 3D scene. That all helps pinpoint the vehicle’s location and relative distance from everything within the virtual simulator.
Based on that information, it reorients the original pixels to recreate a 3D representation of the world from the vehicle’s new viewpoint. It also tracks the motion of the pixels to capture the movement of the cars and people, and other moving objects, in the scene. “This is equivalent to providing the vehicle with an infinite number of possible trajectories,” Rus says. “Because when we collect physical data, we get data from the specific trajectory the car will follow. But we can modify that trajectory to cover all possible ways of and environments of driving. That’s really powerful.”
Reinforcement learning from scratch
Traditionally, researchers have been training autonomous vehicles by either following human defined rules of driving or by trying to imitate human drivers. But the researchers make their controller learn entirely from scratch under an “end-to-end” framework, meaning it takes as input only raw sensor data — such as visual observations of the road — and, from that data, predicts steering commands at outputs.
“We basically say, ‘Here’s an environment. You can do whatever you want. Just don’t crash into vehicles, and stay inside the lanes,’” Amini says.
This requires “reinforcement learning” (RL), a trial-and-error machine-learning technique that provides feedback signals whenever the car makes an error. In the researchers’ simulation engine, the controller begins by knowing nothing about how to drive, what a lane marker is, or even other vehicles look like, so it starts executing random steering angles. It gets a feedback signal only when it crashes. At that point, it gets teleported to a new simulated location and has to execute a better set of steering angles to avoid crashing again. Over 10 to 15 hours of training, it uses these sparse feedback signals to learn to travel greater and greater distances without crashing.
After successfully driving 10,000 kilometers in simulation, the authors apply that learned controller onto their full-scale autonomous vehicle in the real world. The researchers say this is the first time a controller trained using end-to-end reinforcement learning in simulation has successful been deployed onto a full-scale autonomous car. “That was surprising to us. Not only has the controller never been on a real car before, but it’s also never even seen the roads before and has no prior knowledge on how humans drive,” Amini says.
Forcing the controller to run through all types of driving scenarios enabled it to regain control from disorienting positions — such as being half off the road or into another lane — and steer back into the correct lane within several seconds. “And other state-of-the-art controllers all tragically failed at that, because they never saw any data like this in training,” Amini says.
Next, the researchers hope to simulate all types of road conditions from a single driving trajectory, such as night and day, and sunny and rainy weather. They also hope to simulate more complex interactions with other vehicles on the road. “What if other cars start moving and jump in front of the vehicle?” Rus says. “Those are complex, real-world interactions we want to start testing.”
March 23, 2020 | More
How to get conductive gels to stick when wet
Polymers that are good conductors of electricity could be useful in biomedical devices, to help with sensing or electrostimulation, for example. But there has been a sticking point preventing their widespread use: their inability to adhere to a surface such as a sensor or microchip, and stay put despite moisture from the body.
Now, researchers at MIT have come up with a way of getting conductive polymer gels to adhere to wet surfaces.
The new adhesive method is described today in the journal Science Advances in a paper by MIT doctoral student Hyunwoo Yuk, former visiting scholar Akihisa Inoue, postdoc Baoyang Lu, and professor of mechanical engineering Xuanhe Zhao.
Most electrodes used for biomedical devices are made of platinum or platinum-iridium alloys, Zhao explains. These are very good electrical conductors that are durable inside the moist environment of the body, and chemically stable so they do not interact with the surrounding tissues. But their stiffness is a major drawback. Because they can’t flex and stretch as the body moves, they can damage delicate tissues.
Conductive polymers such as PEDOT:PSS, by contrast, can very closely match the softness and flexibility of the vulnerable tissues in the body. The tricky part has been getting them to stay attached to the biomedical devices they are connected to. Researchers have been struggling for years to make these polymers durable in the moist and always-moving environments of the body.
“There have been thousands of papers talking about the advantages of these materials,” Yuk says, but the companies that make biomedical devices “just don’t use them,” because they need materials that are exceedingly reliable and stable. A failure of the material could require an invasive surgical procedure to replace it, which carries additional risk for the patient.
Stiff metal electrodes “sometimes harm the tissues, but they work well in terms of reliability and stability over a period of years,” which has not been the case with polymer substitutes until now, he says.
Most efforts to address this problem have involved making significant modifications to the polymer materials to improve their durability and their ability to adhere, but Yuk says that creates problems of its own: Companies have already invested heavily in equipment to manufacture these polymers, and major changes to the formulation would require significant investment in new production equipment. These changes would be for a market that is relatively small in economic terms, though large in potential impact. Other approaches that have been tried are limited to specific materials. Instead, the MIT team focused on making the fewest changes possible, to ensure compatibility with existing production methods, and making the method applicable to a wide variety of materials.
Their method involves an extremely thin adhesive layer between the conductive polymer hydrogel and the substrate material. Though only a few nanometers thick (billionths of a meter), this layer turns out to be effective at making the gels adhere to any of a wide variety of commonly used substrate materials, including glass, polyimide, indium tin oxide, and gold. The adhesive layer penetrates into the polymer itself, producing a tough, durable protective structure that keeps the material in place even when exposed for long periods to a wet environment.
The adhesive layer can be applied to the devices by a variety of standard manufacturing processes, including spin coating, spray coating, and dip coating, making it easy to integrate with existing fabrication platforms. The coating the researchers used in their tests is made of polyurethane, a hydrophilic (water-attracting) material that is readily available and inexpensive, though other similar polymers could also be used. Such materials “become very strong when they form interpenetrating networks,” as they do when coated on the conducting polymer, Yuk explains. This enhanced strength should address the durability problems associated with the uncoated polymer, he says.
The result is a mechanically strong and conductive gel that bonds tightly with the surface it’s attached to. “It’s a very simple process,” Yuk says.
The bonding proves to be highly resistant to bending, twisting, and even folding of the substrate material. The adhesive polymer has been tested in the lab under accelerated aging conditions using ultrasound, but Yuk says that for the biomedical device industry to accept such a new material will require longer, more rigorous testing to confirm the stability of these coated fibers under realistic conditions over long periods of time.
“We’d be very happy to license and put this technology out there to test it further in realistic situations,” he says. The team has begun talking to manufacturers to see “how we can best help them to test this knowledge,” he says.
“I think this is a great piece of work,” says Zhenan Bao, a professor of chemical engineering at Stanford University, who was not associated with this research. “Wet adhesives are already a big challenge. Conductive adhesives that work well in wet conditions are even more rare. They are very much needed for nerve interfaces and recording electrical signals from the heart or brain.”
Bao says this work “is a major advancement in the bioelectronics field.”
The research was supported by the National Science Foundation, the JSR corporation, and Samsung.
March 20, 2020 | More
MIT graduate engineering, business programs ranked highly by U.S. News for 2021
MIT’s graduate program in engineering has again earned a No. 1 spot in U.S. News and Word Report’s annual rankings, a place it has held since 1990, when the magazine first ranked such programs.
The MIT Sloan School of Management also placed highly, occupying the No. 5 spot for the best graduate business programs.
Among individual engineering disciplines, MIT placed first in six areas: aerospace/aeronautical/astronautical engineering (tied with Caltech), chemical engineering, computer engineering, electrical/electronic/communications engineering (tied with Stanford University and the University of California at Berkeley), materials engineering, and mechanical engineering. It placed second in nuclear engineering.
In the rankings of individual MBA specialties, MIT placed first in four areas: business analytics, information systems, production/operations, and project management. It placed second in supply chain/logistics.
U.S. News does not issue annual rankings for all doctoral programs but revisits many every few years. In 2018, MIT ranked in the top five for 24 of the 37 science disciplines evaluated.
The magazine bases its rankings of graduate schools of engineering and business on two types of data: reputational surveys of deans and other academic officials, and statistical indicators that measure the quality of a school’s faculty, research, and students. The magazine’s less-frequent rankings of programs in the sciences, social sciences, and humanities are based solely on reputational surveys.
March 17, 2020 | More
Events postponed or canceled as MIT responds to COVID-19
MIT schools, departments, labs, centers, and offices have acted swiftly to postpone or cancel large events through May 15 in the wake of the Institute’s announcement last week of new policies regarding gatherings likely to attract 150 or more people.
To safeguard against COVID-19, and the spread of the 2019 novel coronavirus, many other MIT events have been modified both on campus and elsewhere, with increased opportunities offered for livestreaming.
The guidelines put forth last week have also now been expanded to include some large classes: The Institute will move classes with more than 150 students online, starting this week.
Impacts on classes and student travel
Following consultation with senior academic leadership and experts within MIT Medical, the Institute has suspended in-person meetings of classes with more than 150 students, effective tomorrow, Tuesday, March 10. The approximately 20 classes impacted by the decision will continue to be offered in virtual form.
“We are being guided by our medical professionals who are in close contact with state and national public health officials,” Ian Waitz, vice chancellor for undergraduate and graduate education, wrote today in a letter to deans and department heads. “They have advised us that while the risk to the community is low and there are no cases on campus as of now, we need to move quickly to help prevent the potential transmission of the disease and to be ready if and when it impacts our campus.”
“Our approach is to be aggressive, but to move forward in stages,” Waitz added, “while keeping in mind that some individual faculty and departments may be moving faster than others, that the level of comfort with remote teaching varies, and that some classes may translate better than others to alternative formats.”
As of now, midterm examinations will proceed as scheduled, but the plan for large courses is to run midterms in several rooms simultaneously so the number of students in each room remains well below 150. The Registrar’s Office is working on room scheduling strategies to best accommodate that approach.
The Institute has also decided to cancel any MIT-sponsored student travel that is related to a class, and all MIT-sponsored student domestic travel of more than 100 miles will have to go through the Institute’s high-risk travel waiver process.
Impacts on undergraduate and graduate admissions
As shared in President L. Rafael Reif’s letter of last Thursday, MIT’s new policy on events will apply to Campus Preview Weekend, ordinarily an on-campus gathering for students admitted to the incoming first-year undergraduate class. In the coming weeks, the Admissions Office will be connecting with admitted students, current students, and campus partners to discuss what to do instead of a conventional CPW. For more information, please see: https://mitadmissions.org/blogs/entry/mits-covid-19-precautions-and-its-impact-on-admissions/
The Admissions Office will not host any programming for K-12 students, including admitted students and their families, between now and May 15, regardless of the size of the event. All scheduled admissions sessions and tours have been canceled between now and May 15, and MIT Admissions is canceling all scheduled admissions officer travel to domestic and international events in that time window.
Additionally, all graduate admissions visit days have been canceled, effective immediately. “Based upon reducing risk, we ask all departments to cancel all remaining graduate open houses and visit days, and to move to virtual formats,” Waitz says. “Many departments have already done this.”
Despite the cancellation of these formal events, the MIT campus currently remains open for visits by prospective students. However, in keeping with suggested best practices for public health, visitors from countries that the U.S. Centers for Disease Control and Prevention (CDC) finds have “widespread sustained (ongoing) transmission” of COVID-19 cannot visit campus until they have successfully completed 14 days of self-quarantine.
Impacts on major campus events
The MIT Excellence Awards and Collier Medal celebration, scheduled for this Thursday, March 12, has been postponed; a rescheduled date will be announced as soon as it is confirmed. The Excellence Awards and Collier Medal recognize the work of service, support, administrative, and sponsored research staff. The Excellence Awards acknowledge the extraordinary efforts made by members of the MIT community toward fulfilling the goals, values, and mission of the Institute. The Collier Medal is awarded to an individual or group exhibiting qualities such as a commitment to community service, kindness, selflessness, and generosity; it honors the memory of MIT Police Officer Sean Collier, who lost his life while protecting the MIT campus. A full list of this year’s honorees is available.
Career Advising and Professional Development is working on plans to change the format of the Spring Career Fair, previously scheduled for April 2, to a virtual career fair for a date to be announced in April. All other large-scale employer engagement events — such as career fairs, mixers, symposiums, and networking events — will also be canceled; adopt a virtual model; be postponed beyond May 15; or adopt other models that meet the new policies involving large events.
MIT is postponing the remaining two Climate Action Symposia, “MIT Climate Initiatives and the Role of Research Universities” and “Summing Up: Why Is the World Waiting?” — previously scheduled for April 2 and April 22, respectively. These symposia will be rescheduled; new dates will be announced on climatesymposia.mit.edu.
Solve at MIT on May 12-14 will be virtual. In addition to a livestream on this page, Solve will continue to bring together its cross-sector community via interactive online workshops and more. Participants can also contribute a solution or a donation to the Health Security and Pandemics Challenge.
Impacts on athletics and intercollegiate athletics events
The Department of Athletics, Physical Education and Recreation (DAPER) is taking steps to safeguard student-athletes, staff, and community members who utilize DAPER facilities for club sports, intramurals, and recreation. Unless otherwise announced, MIT’s intercollegiate athletics events will continue as scheduled. However, visiting teams are asked to bring only student-athletes and essential team personnel to events at MIT.
Additionally, DAPER has requested that only MIT students, faculty, and staff members attend upcoming home athletic events through May 15. All other spectators, including parents, are asked to watch events using DAPER’s video streaming service.
Other impacted events and activities
Discussions are ongoing about many additional events scheduled between now and May 15. The list below will be updated as more information becomes available. Among the affected events and activities announced so far:
- Use of the pillars in Lobby 7 for community discussion is suspended for the rest of the spring semester, to minimize close contact and sharing of writing implements.
- SpaceTech 2020, scheduled for Wednesday, March 11, has been postponed until a later date. The all-day event, part of MIT Space Week, will highlight the future of space exploration by featuring lightning talks from current students; talks and panels from alumni; and an interactive guided tour along the Space Trail to visit Department of Aeronautics and Astronautics (AeroAstro) labs and ongoing research projects. Visit spacetech.mit.edu for the latest information.
- MIT Getfit has canceled both of its midpoint events originally scheduled for Wednesday, March 11. Organizers are working to contact participants with more information.
- The March 13 lecture titled “Fateful Triangle: How China Shaped US-India Relations During the Cold War,” by Tanvi Madan of the Brookings Institution, has been postponed. More information is available at http://southasianpolitics.net/.
- To the Moon to Stay Hackathon, scheduled for Saturday, March 14, has been postponed until a later date. MIT AeroAstro and the MIT Media Lab’s Space Exploration Initiative are partnering to design and build an experiment to go to the moon on board Blue Origin’s inaugural lunar mission. The goal of the hackathon is to bring the MIT community together to think about lunar missions and habitation through a variety of challenges. To receive updates, join their email list or visit tothemoon.mit.edu.
- The Koch Institute is limiting attendance at the SCIENCE with/in/sight: 2020 Visions event on March 17. This event is now for invited guests only.
- All MIT Communications Forum events have been postponed until the fall. This includes Science Under Attack, originally scheduled for March 19, and David Thorburn’s presentation as part of the William Corbett Poetry Series, originally scheduled for April 8.
- The MIT de Florez Award Competition, scheduled for April 15, will be conducted virtually. Additional information will be sent to the Mechanical Engineering community via email.
- The Mechanical Engineering Graduate Student Gala, scheduled for April 19, has been canceled and will be rescheduled for the fall.
- The Mechanical Engineering Student Awards Banquet, scheduled for May 15, has been canceled. Awards will be announced virtually.
- The Office of Engineering Outreach Programs (OEOP) has canceled its SEED Academy program through May 15. This includes the SEED Academy Spring Final Symposium on May 9. OEOP will continue to communicate with SEED Academy students and parents via email and through The Sprout newsletter to offer information on course, project, and engagement options.
- The 2020 Brazil Conference at MIT and Harvard has been canceled. More information can be found at brazilconference.org.
- The March 12 Starr Forum, titled “Russia’s Putin: From Silent Coup to Legal Dictatorship,” has been changed to a live webcast.
- The March 13 Myron Weiner Seminar on International Migration, titled “Future Aspirations Among Refugee Youth in Turkey Between Integration & Mobility,” has been canceled.
- The MIT Sloan School of Management is canceling all international study tours and treks. Student conferences are either being cancelled or modified: The March 7 Robo-AI Exchange Conference, the March 13 New Space Age Conference, and the April 2 Golub Center for Finance and Policy discussion on equity market structure with the SEC are canceled. The March 13 ETA Summit and the April 17 Ops Sim Competition are proceeding, with virtualization. The March 16 Entrepreneurship and Innovation Alumni gathering in San Franciso is also canceled.
This article will be updated as more information on impacted events becomes available.
March 9, 2020 | More
QS World University Rankings rates MIT No. 1 in 12 subjects for 2020
MIT has been honored with 12 No. 1 subject rankings in the QS World University Rankings for 2020.
The Institute received a No. 1 ranking in the following QS subject areas: Architecture/Built Environment; Chemistry; Computer Science and Information Systems; Chemical Engineering; Civil and Structural Engineering; Electrical and Electronic Engineering; Mechanical, Aeronautical and Manufacturing Engineering; Linguistics; Materials Science; Mathematics; Physics and Astronomy; and Statistics and Operational Research.
MIT also placed second in five subject areas: Accounting and Finance; Biological Sciences; Earth and Marine Sciences; Economics and Econometrics; and Environmental Sciences.
Quacquarelli Symonds Limited subject rankings, published annually, are designed to help prospective students find the leading schools in their field of interest. Rankings are based on research quality and accomplishments, academic reputation, and graduate employment.
MIT has been ranked as the No. 1 university in the world by QS World University Rankings for eight straight years.
March 3, 2020 | More
Engineers design bionic “heart” for testing prosthetic valves, other cardiac devices
As the geriatric population is expected to balloon in the coming decade, so too will rates of heart disease in the United States. The demand for prosthetic heart valves and other cardiac devices — a market that is valued at more than $5 billion dollars today — is predicted to rise by almost 13 percent in the next six years.
Prosthetic valves are designed to mimic a real, healthy heart valve in helping to circulate blood through the body. However, many of them have issues such as leakage around the valve, and engineers working to improve these designs must test them repeatedly, first in simple benchtop simulators, then in animal subjects, before reaching human trials — an arduous and expensive process.
Now engineers at MIT and elsewhere have developed a bionic “heart” that offers a more realistic model for testing out artificial valves and other cardiac devices.
The device is a real biological heart whose tough muscle tissue has been replaced with a soft robotic matrix of artificial heart muscles, resembling bubble wrap. The orientation of the artificial muscles mimics the pattern of the heart’s natural muscle fibers, in such a way that when the researchers remotely inflate the bubbles, they act together to squeeze and twist the inner heart, similar to the way a real, whole heart beats and pumps blood.
With this new design, which they call a “biorobotic hybrid heart,” the researchers envision that device designers and engineers could iterate and fine-tune designs more quickly by testing on the biohybrid heart, significantly reducing the cost of cardiac device development.
“Regulatory testing of cardiac devices requires many fatigue tests and animal tests,” says Ellen Roche, assistant professor of mechanical engineering at MIT. “[The new device] could realistically represent what happens in a real heart, to reduce the amount of animal testing or iterate the design more quickly.”
Roche and her colleagues have published their results today in the journal Science Robotics. Her co-authors are lead author and MIT graduate student Clara Park, along with Yiling Fan, Gregor Hager, Hyunwoo Yuk, Manisha Singh, Allison Rojas, and Xuanhe Zhao at MIT, along with collaborators from Nanyang Technology University, the Royal College of Surgeons in Dublin, Boston’s Children’s Hospital, Harvard Medical School, and Massachusetts General Hospital.
The structure of the biorobotic hybrid heart under magnetic resonance imaging. Credit: Christopher T. Nguyen
“Mechanics of the heart”
Before coming to MIT, Roche worked briefly in the biomedical industry, helping to test cardiac devices on artificial heart models in the lab.
“At the time I didn’t feel any of these benchtop setups were representative of both the anatomy and the physiological biomechanics of the heart,” Roche recalls. “There was an unmet need in terms of device testing.”
In separate research as part of her doctoral work at Harvard University, she developed a soft, robotic, implantable sleeve, designed to wrap around a whole, live heart, to help it pump blood in patients suffering from heart failure.
At MIT, she and Park wondered if they could combine the two research avenues, to develop a hybrid heart: a heart that is made partly of chemically preserved, explanted heart tissue and partly of soft artificial actuators that help the heart pump blood. Such a model, they proposed, should be a more realistic and durable environment in which to test cardiac devices, compared with models that are either entirely artificial but do not capture the heart’s complex anatomy, or are made from a real explanted heart, requiring highly controlled conditions to keep the tissue alive.
The team briefly considered wrapping a whole, explanted heart in a soft robotic sleeve, similar to Roche’s previous work, but realized the heart’s outer muscle tissue, the myocardium, quickly stiffened when removed from the body. Any robotic contraction by the sleeve would fail to translate sufficiently to the heart within.
Instead, the team looked for ways to design a soft robotic matrix to replace the heart’s natural muscle tissue, in both material and function. They decided to try out their idea first on the heart’s left ventricle, one of four chambers in the heart, which pumps blood to the rest of the body, while the right ventricle uses less force to pump blood to the lungs.
“The left ventricle is the harder one to recreate given its higher operating pressures, and we like to start with the hard challenges,” Roche says.
The heart, unfurled
The heart normally pumps blood by squeezing and twisting, a complex combination of motions that is a result of the alignment of muscle fibers along the outer myocardium that covers each of the heart’s ventricles. The team planned to fabricate a matrix of artificial muscles resembling inflatable bubbles, aligned in the orientations of the natural cardiac muscle. But copying these patterns by studying a ventricle’s three-dimensional geometry proved extremely challenging.
They eventually came across the helical ventricular myocardial band theory, the idea that cardiac muscle is essentially a large helical band that wraps around each of the heart’s ventricles. This theory is still a subject of debate by some researchers, but Roche and her colleagues took it as inspiration for their design. Instead of trying to copy the left ventricle’s muscle fiber orientation from a 3D perspective, the team decided to remove the ventricle’s outer muscle tissue and unwrap it to form a long, flat band — a geometry that should be far easier to recreate. In this case, they used the cardiac tissue from an explanted pig heart.
In collaboration with co-lead author Chris Nguyen at MGH, the researchers used diffusion tensor imaging, an advanced technique that typically tracks how water flows through white matter in the brain, to map the microscopic fiber orientations of a left ventricle’s unfurled, two-dimensional muscle band. They then fabricated a matrix of artificial muscle fibers made from thin air tubes, each connected to a series of inflatable pockets, or bubbles, the orientation of which they patterned after the imaged muscle fibers.
Motion of the biorobotic hybrid heart mimics the pumping motion of the heart under echocardiography. Credit: Mossab Saeed
The soft matrix consists of two layers of silicone, with a water-soluble layer between them to prevent the layers from sticking, as well as two layers of laser-cut paper, which ensures that the bubbles inflate in a specific orientation.
The researchers also developed a new type of bioadhesive to glue the bubble wrap to the ventricle’s real, intracardiac tissue. While adhesives exist for bonding biological tissues to each other, and and for materials like silicone to each other, the team realized few soft adhesives do an adequate job of gluing together biological tissue with synthetic materials, silicone in particular.
So Roche collaborated with Zhao, associate professor of mechanical engineering at MIT, who specializes in developing hydrogel-based adhesives. The new adhesive, named TissueSil, was made by functionalizing silicone in a chemical cross-linking process, to bond with components in heart tissue. The result was a viscous liquid that the researchers brushed onto the soft robotic matrix. They also brushed the glue onto a new explanted pig heart that had its left ventricle removed but its endocardial structures preserved. When they wrapped the artificial muscle matrix around this tissue, the two bonded tightly.
Finally, the researchers placed the entire hybrid heart in a mold that they had previously cast of the original, whole heart, and filled the mold with silicone to encase the hybrid heart in a uniform covering — a step that produced a form similar to a real heart and ensured that the robotic bubble wrap fit snugly around the real ventricle.
“That way, you don’t lose transmission of motion from the synthetic muscle to the biological tissue,” Roche says.
When the researchers pumped air into the bubble wrap at frequencies resembling a naturally beating heart, and imaged the bionic heart’s response, it contracted in a manner similar to the way a real heart moves to pump blood through the body.
Ultimately, the researchers hope to use the bionic heart as a realistic environment to help designers test cardiac devices such as prosthetic heart valves.
“Imagine that a patient before cardiac device implantation could have their heart scanned, and then clinicians could tune the device to perform optimally in the patient well before the surgery,” says Nyugen. “Also, with further tissue engineering, we could potentially see the biorobotic hybrid heart be used as an artificial heart — a very needed potential solution given the global heart failure epidemic where millions of people are at the mercy of a competitive heart transplant list.”
This research was supported in part by the National Science Foundation.
January 29, 2020 | More
Using artificial intelligence to enrich digital maps
A model invented by researchers at MIT and Qatar Computing Research Institute (QCRI) that uses satellite imagery to tag road features in digital maps could help improve GPS navigation.
Showing drivers more details about their routes can often help them navigate in unfamiliar locations. Lane counts, for instance, can enable a GPS system to warn drivers of diverging or merging lanes. Incorporating information about parking spots can help drivers plan ahead, while mapping bicycle lanes can help cyclists negotiate busy city streets. Providing updated information on road conditions can also improve planning for disaster relief.
But creating detailed maps is an expensive, time-consuming process done mostly by big companies, such as Google, which send
s vehicles around with cameras strapped to their hoods to capture video and images of an area’s roads. Combining that with other data can create accurate, up-to-date maps. Because this process is expensive, however, some parts of the world are ignored.
A solution is to unleash machine-learning models on satellite images — which are easier to obtain and updated fairly regularly — to automatically tag road features. But roads can be occluded by, say, trees and buildings, making it a challenging task. In a paper being presented at the Association for the Advancement of Artificial Intelligence conference, the MIT and QCRI researchers describe “RoadTagger,” which uses a combination of neural network architectures to automatically predict the number of lanes and road types (residential or highway) behind obstructions.
In testing RoadTagger on occluded roads from digital maps of 20 U.S. cities, the model counted lane numbers with 77 percent accuracy and inferred road types with 93 percent accuracy. The researchers are also planning to enable RoadTagger to predict other features, such as parking spots and bike lanes.
“Most updated digital maps are from places that big companies care the most about. If you’re in places they don’t care about much, you’re at a disadvantage with respect to the quality of map,” says co-author Sam Madden, a professor in the Department of Electrical Engineering and Computer Science (EECS) and a researcher in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “Our goal is to automate the process of generating high-quality digital maps, so they can be available in any country.”
The paper’s co-authors are CSAIL graduate students Songtao He, Favyen Bastani, and Edward Park; EECS undergraduate student Satvat Jagwani; CSAIL professors Mohammad Alizadeh and Hari Balakrishnan; and QCRI researchers Sanjay Chawla, Sofiane Abbar, and Mohammad Amin Sadeghi.
Combining CNN and GNN
Qatar, where QCRI is based, is “not a priority for the large companies building digital maps,” Madden says. Yet, it’s constantly building new roads and improving old ones, especially in preparation for hosting the 2022 FIFA World Cup.
“While visiting Qatar, we’ve had experiences where our Uber driver can’t figure out how to get where he’s going, because the map is so off,” Madden says. “If navigation apps don’t have the right information, for things such as lane merging, this could be frustrating or worse.”
RoadTagger relies on a novel combination of a convolutional neural network (CNN) — commonly used for images-processing tasks — and a graph neural network (GNN). GNNs model relationships between connected nodes in a graph and have become popular for analyzing things like social networks and molecular dynamics. The model is “end-to-end,” meaning it’s fed only raw data and automatically produces output, without human intervention.
The CNN takes as input raw satellite images of target roads. The GNN breaks the road into roughly 20-meter segments, or “tiles.” Each tile is a separate graph node, connected by lines along the road. For each node, the CNN extracts road features and shares that information with its immediate neighbors. Road information propagates along the whole graph, with each node receiving some information about road attributes in every other node. If a certain tile is occluded in an image, RoadTagger uses information from all tiles along the road to predict what’s behind the occlusion.
This combined architecture represents a more human-like intuition, the researchers say. Say part of a four-lane road is occluded by trees, so certain tiles show only two lanes. Humans can easily surmise that a couple lanes are hidden behind the trees. Traditional machine-learning models — say, just a CNN — extract features only of individual tiles and most likely predict the occluded tile is a two-lane road.
“Humans can use information from adjacent tiles to guess the number of lanes in the occluded tiles, but networks can’t do that,” He says. “Our approach tries to mimic the natural behavior of humans, where we capture local information from the CNN and global information from the GNN to make better predictions.”
To train and test RoadTagger, the researchers used a real-world map dataset, called OpenStreetMap, which lets users edit and curate digital maps around the globe. From that dataset, they collected confirmed road attributes from 688 square kilometers of maps of 20 U.S. cities — including Boston, Chicago, Washington, and Seattle. Then, they gathered the corresponding satellite images from a Google Maps dataset.
In training, RoadTagger learns weights — which assign varying degrees of importance to features and node connections — of the CNN and GNN. The CNN extracts features from pixel patterns of tiles and the GNN propagates the learned features along the graph. From randomly selected subgraphs of the road, the system learns to predict the road features at each tile. In doing so, it automatically learns which image features are useful and how to propagate those features along the graph. For instance, if a target tile has unclear lane markings, but its neighbor tile has four lanes with clear lane markings and shares the same road width, then the target tile is likely to also have four lanes. In this case, the model automatically learns that the road width is a useful image feature, so if two adjacent tiles share the same road width, they’re likely to have the same lane count.
Given a road not seen in training from OpenStreetMap, the model breaks the road into tiles and uses its learned weights to make predictions. Tasked with predicting a number of lanes in an occluded tile, the model notes that neighboring tiles have matching pixel patterns and, therefore, a high likelihood to share information. So, if those tiles have four lanes, the occluded tile must also have four.
In another result, RoadTagger accurately predicted lane numbers in a dataset of synthesized, highly challenging road disruptions. As one example, an overpass with two lanes covered a few tiles of a target road with four lanes. The model detected mismatched pixel patterns of the overpass, so it ignored the two lanes over the covered tiles, accurately predicting four lanes were underneath.
The researchers hope to use RoadTagger to help humans rapidly validate and approve continuous modifications to infrastructure in datasets such as OpenStreetMap, where many maps don’t contain lane counts or other details. A specific area of interest is Thailand, Bastani says, where roads are constantly changing, but there are few if any updates in the dataset.
“Roads that were once labeled as dirt roads have been paved over so are better to drive on, and some intersections have been completely built over. There are changes every year, but digital maps are out of date,” he says. “We want to constantly update such road attributes based on the most recent imagery.”
January 23, 2020 | More
A new approach to making airplane parts, minus the massive infrastructure
A modern airplane’s fuselage is made from multiple sheets of different composite materials, like so many layers in a phyllo-dough pastry. Once these layers are stacked and molded into the shape of a fuselage, the structures are wheeled into warehouse-sized ovens and autoclaves, where the layers fuse together to form a resilient, aerodynamic shell.
Now MIT engineers have developed a method to produce aerospace-grade composites without the enormous ovens and pressure vessels. The technique may help to speed up the manufacturing of airplanes and other large, high-performance composite structures, such as blades for wind turbines.
The researchers detail their new method in a paper published today in the journal Advanced Materials Interfaces.
“If you’re making a primary structure like a fuselage or wing, you need to build a pressure vessel, or autoclave, the size of a two- or three-story building, which itself requires time and money to pressurize,” says Brian Wardle, professor of aeronautics and astronautics at MIT. “These things are massive pieces of infrastructure. Now we can make primary structure materials without autoclave pressure, so we can get rid of all that infrastructure.”
Wardle’s co-authors on the paper are lead author and MIT postdoc Jeonyoon Lee, and Seth Kessler of Metis Design Corporation, an aerospace structural health monitoring company based in Boston.
Out of the oven, into a blanket
In 2015, Lee led the team, along with another member of Wardle’s lab, in creating a method to make aerospace-grade composites without requiring an oven to fuse the materials together. Instead of placing layers of material inside an oven to cure, the researchers essentially wrapped them in an ultrathin film of carbon nanotubes (CNTs). When they applied an electric current to the film, the CNTs, like a nanoscale electric blanket, quickly generated heat, causing the materials within to cure and fuse together.
With this out-of-oven, or OoO, technique, the team was able to produce composites as strong as the materials made in conventional airplane manufacturing ovens, using only 1 percent of the energy.
The researchers next looked for ways to make high-performance composites without the use of large, high-pressure autoclaves — building-sized vessels that generate high enough pressures to press materials together, squeezing out any voids, or air pockets, at their interface.
“There’s microscopic surface roughness on each ply of a material, and when you put two plys together, air gets trapped between the rough areas, which is the primary source of voids and weakness in a composite,” Wardle says. “An autoclave can push those voids to the edges and get rid of them.”
Researchers including Wardle’s group have explored “out-of-autoclave,” or OoA, techniques to manufacture composites without using the huge machines. But most of these techniques have produced composites where nearly 1 percent of the material contains voids, which can compromise a material’s strength and lifetime. In comparison, aerospace-grade composites made in autoclaves are of such high quality that any voids they contain are neglible and not easily measured.
“The problem with these OoA approaches is also that the materials have been specially formulated, and none are qualified for primary structures such as wings and fuselages,” Wardle says. “They’re making some inroads in secondary structures, such as flaps and doors, but they still get voids.”
Part of Wardle’s work focuses on developing nanoporous networks — ultrathin films made from aligned, microscopic material such as carbon nanotubes, that can be engineered with exceptional properties, including color, strength, and electrical capacity. The researchers wondered whether these nanoporous films could be used in place of giant autoclaves to squeeze out voids between two material layers, as unlikely as that may seem.
A thin film of carbon nanotubes is somewhat like a dense forest of trees, and the spaces between the trees can function like thin nanoscale tubes, or capillaries. A capillary such as a straw can generate pressure based on its geometry and its surface energy, or the material’s ability to attract liquids or other materials.
The researchers proposed that if a thin film of carbon nanotubes were sandwiched between two materials, then, as the materials were heated and softened, the capillaries between the carbon nanotubes should have a surface energy and geometry such that they would draw the materials in toward each other, rather than leaving a void between them. Lee calculated that the capillary pressure should be larger than the pressure applied by the autoclaves.
The researchers tested their idea in the lab by growing films of vertically aligned carbon nanotubes using a technique they previously developed, then laying the films between layers of materials that are typically used in the autoclave-based manufacturing of primary aircraft structures. They wrapped the layers in a second film of carbon nanotubes, which they applied an electric current to to heat it up. They observed that as the materials heated and softened in response, they were pulled into the capillaries of the intermediate CNT film.
The resulting composite lacked voids, similar to aerospace-grade composites that are produced in an autoclave. The researchers subjected the composites to strength tests, attempting to push the layers apart, the idea being that voids, if present, would allow the layers to separate more easily.
“In these tests, we found that our out-of-autoclave composite was just as strong as the gold-standard autoclave process composite used for primary aerospace structures,” Wardle says.
The team will next look for ways to scale up the pressure-generating CNT film. In their experiments, they worked with samples measuring several centimeters wide — large enough to demonstrate that nanoporous networks can pressurize materials and prevent voids from forming. To make this process viable for manufacturing entire wings and fuselages, researchers will have to find ways to manufacture CNT and other nanoporous films at a much larger scale.
“There are ways to make really large blankets of this stuff, and there’s continuous production of sheets, yarns, and rolls of material that can be incorporated in the process,” Wardle says.
He plans also to explore different formulations of nanoporous films, engineering capillaries of varying surface energies and geometries, to be able to pressurize and bond other high-performance materials.
“Now we have this new material solution that can provide on-demand pressure where you need it,” Wardle says. “Beyond airplanes, most of the composite production in the world is composite pipes, for water, gas, oil, all the things that go in and out of our lives. This could make making all those things, without the oven and autoclave infrastructure.”
This research was supported, in part, by Airbus, ANSYS, Embraer, Lockheed Martin, Saab AB, Saertex, and Teijin Carbon America through MIT’s Nano-Engineered Composite aerospace Structures (NECST) Consortium.
January 13, 2020 | More
Making real a biotechnology dream: nitrogen-fixing cereal crops
As food demand rises due to growing and changing populations around the world, increasing crop production has been a vital target for agriculture and food systems researchers who are working to ensure there is enough food to meet global need in the coming years. One MIT research group mobilizing around this challenge is the Voigt lab in the Department of Biological Engineering, led by Christopher Voigt, the Daniel I.C. Wang Professor of Advanced Biotechnology at MIT.
For the past four years, the Abdul Latif Jameel Water and Food Systems Lab (J-WAFS) has funded Voigt with two J-WAFS Seed Grants. With this support, Voigt and his team are working on a significant and longstanding research challenge: transform cereal crops so they are able to fix their own nitrogen.
Chemical fertilizer: how it helps and hurts
Nitrogen is a key nutrient that enables plants to grow. Plants like legumes are able to provide their own through a symbiotic relationship with bacteria that are capable of fixing nitrogen from the air and putting it into the soil, which is then drawn up by the plants through their roots. Other types of crops — including major food crops such as corn, wheat, and rice — typically rely on added fertilizers for nitrogen, including manure, compost, and chemical fertilizers. Without these, the plants that grow are smaller and produce less grain.
Over 3.5 billion people today depend on chemical fertilizers for their food. Eighty percent of chemical nitrogen fertilizers today are made using the Haber-Borsch process, which involves transforming nitrile gas into ammonia. While nitrogen fertilizer has boosted agriculture production in the last century, this has come with some significant costs. First, the Haber-Borsch process itself is very energy- and fossil fuel-intensive, making it unsustainable in the face of a rapidly changing climate. Second, using too much chemical fertilizer results in nitrogen pollution. Fertilizer runoff pollutes rivers and oceans, resulting in algae blooms that suffocate marine life. Cleaning up this pollution and paying for the public health and environmental damage costs the United States $157 billion annually. Third, when it comes to chemical fertilizers, there are problems with equity and access. These fertilizers are made in the northern hemisphere by major industrialized nations, where postash, a main ingredient, is abundant. However, transportation costs are high, especially to countries in the southern hemisphere. So, for farmers in poorer regions, this barrier results in lower crop yield.
These environmental and societal challenges pose large problems, yet farmers still need to apply nitrogen to maintain the necessary agriculture productivity to meet the world’s food needs, especially as population and climate change stress the world’s food supplies. So, fertilizers are and will continue to be a critical tool.
But, might there be another way?
The bacterial compatability of chloroplasts and mitochondria
This is the question that drives researchers in the Voigt lab, as they work to develop nitrogen-fixing cereal grains. The strategy they have developed is to target the specific genes in the nitrogen-fixing bacteria that operate symbiotically with legumes, called the nif genes. These genes cause the expression of the protein structures (nitrogenase clusters) that fix nitrogen from the air. If these genes were able to be successfully transferred and expressed in cereal crops, chemical fertilizers would no longer be needed to add needed nitrogen, as these crops would be able to obtain nitrogen themselves.
This genetic engineering work has long been regarded as a major technical challenge, however. The nif pathway is very large and involves many different genes. Transferring any large gene cluster is itself a difficult task, but there is added complexity in this particular pathway. The nif genes in microbes are controlled by a precise system of interconnected genetic parts. In order to successfully transfer the pathway’s nitrogen-fixing capabilities, researchers not only have to transfer the genes themselves, but also replicate the cellular components responsible for controlling the pathway.
This leads into another challenge. The microbes responsible for nitrogen fixation in legumes are bacteria (prokaryotes), and, as explained by Eszter Majer, a postdoc in the Voigt lab who has been working on the project for the past two years, “the gene expression is completely different in plants, which are eukaryotes.” For example, prokaryotes organize their genes into operons, a genetic organization system that does not exist in eukaryotes such as the tobacco leaves the Voigt is using in its experiments. Reengineering the nif pathway in a eukaryote is tantamount to a complete system overhaul.
The Voigt lab has found a workaround: Rather than target the entire plant cell, they are targetting organelles within the cell — specifically, the chloroplasts and the mitochondria. Mitochondria and chloroplasts both have ancient bacterial origins and once lived independently outside of eukaryotic cells as prokaryotes. Millions of years ago, they were incorporated into the eukaryotic system as organelles. They are unique in that they have their own genetic data and have also maintained many similarities to modern-day prokaryotes. As a result, they are excellent candidates for nitrogenase transfer. Majer explains, “It’s much easier to transfer from a prokaryote to a prokaryote-like system than reengineer the whole pathway and try to transfer to a eukaryote.”
Beyond gene structure, these organelles have additional attributes that make them suitable environments for nitrogenase clusters to function. Nitrogenase requires a lot of energy to function and both chloroplasts and mitochondria already produce high amounts energy — in the form of ATP — for the cell. Nitrogenase is also very sensitive to oxygen and will not function if there is too much of it in its environment. However, chloroplasts at night and mitochondria in plants have low-oxygen levels, making them an ideal location for the nitrogenase protein to operate.
An international team of experts
While the team found devised an approach for transforming eukaryotic cells, their project still involved highly technical biological engineering challenges. Thanks to the J-WAFS grants, the Voigt lab has been able to collaborate with two specialists at overseas universities to obtain critical expertise..
One was Luis Rubio, an associate professor focusing on the biochemistry of nitrogen fixation at the Polytechnic University of Madrid, Spain. Rubio is an expert in nitrogenase and nitrogen-inspired chemistry. Transforming mitochondrial DNA is a challenging process, so the team designed a nitrogenase gene delivery system using yeast. Yeast are easy eukaryotic organisms to engineer and can be used to target the mitochondria. The team inserted the nitrogenase genes into the yeast nuclei, which are then targeted to mitochondria using peptide fusions. This research resulted in the first eukaryotic organism to demonstrate the formation of nitrogenase structural proteins.
The Voigt lab also collaborated with Ralph Bock, a chloroplast expert from the Max Planck Institute of Molecular Plant Physiology in Germany. He and the Voigt team have made great strides toward the goal of nitrogen-fixing cereal crops; the details of their recent accomplishments advancing the field crop engineering and furthering the nitrogen-fixing work will be published in the coming months.
Continuing in pursuit of the dream
The Voigt lab, with the support of J-WAFS and the invaluable international collaboration that has resulted, was able to obtain groundbreaking results, moving us closer to fertilizer independence through nitrogen-fixing cereals. They made headway in targeting nitrogenase to mitochondria and were able to express a complete NifDK tetramer — a key protein in the nitrogenase cluster — in yeast mitochondria. Despite these milestones, more work is yet to be done.
“The Voigt lab is invested in moving this research forward in order to get ever closer to the dream of creating nitrogen-fixing cereal crops,“ says Chris Voigt. With these milestones under their belt, these researchers have made great advances, and will continue to push torward the realization of this transformative vision, one that could revolutionize cereal production globally.
January 10, 2020 | More
Tool predicts how fast code will run on a chip
MIT researchers have invented a machine-learning tool that predicts how fast computer chips will execute code from various applications.
To get code to run as fast as possible, developers and compilers — programs that translate programming language into machine-readable code — typically use performance models that run the code through a simulation of given chip architectures.
Compilers use that information to automatically optimize code, and developers use it to tackle performance bottlenecks on the microprocessors that will run it. But performance models for machine code are handwritten by a relatively small group of experts and are not properly validated. As a consequence, the simulated performance measurements often deviate from real-life results.
In series of conference papers, the researchers describe a novel machine-learning pipeline that automates this process, making it easier, faster, and more accurate. In a paper presented at the International Conference on Machine Learning in June, the researchers presented Ithemal, a neural-network model that trains on labeled data in the form of “basic blocks” — fundamental snippets of computing instructions — to automatically predict how long it takes a given chip to execute previously unseen basic blocks. Results suggest Ithemal performs far more accurately than traditional hand-tuned models.
Then, at the November IEEE International Symposium on Workload Characterization, the researchers presented a benchmark suite of basic blocks from a variety of domains, including machine learning, compilers, cryptography, and graphics that can be used to validate performance models. They pooled more than 300,000 of the profiled blocks into an open-source dataset called BHive. During their evaluations, Ithemal predicted how fast Intel chips would run code even better than a performance model built by Intel itself.
Ultimately, developers and compilers can use the tool to generate code that runs faster and more efficiently on an ever-growing number of diverse and “black box” chip designs. “Modern computer processors are opaque, horrendously complicated, and difficult to understand. It is also incredibly challenging to write computer code that executes as fast as possible for these processors,” says co-author Michael Carbin, an assistant professor in the Department of Electrical Engineering and Computer Science (EECS) and a researcher in the Computer Science and Artificial Intelligence Laboratory (CSAIL). “This tool is a big step forward toward fully modeling the performance of these chips for improved efficiency.”
Most recently, in a paper presented at the NeurIPS conference in December, the team proposed a new technique to automatically generate compiler optimizations. Specifically, they automatically generate an algorithm, called Vemal, that converts certain code into vectors, which can be used for parallel computing. Vemal outperforms hand-crafted vectorization algorithms used in the LLVM compiler — a popular compiler used in the industry.
Learning from data
Designing performance models by hand can be “a black art,” Carbin says. Intel provides extensive documentation of more than 3,000 pages describing its chips’ architectures. But there currently exists only a small group of experts who will build performance models that simulate the execution of code on those architectures.
“Intel’s documents are neither error-free nor complete, and Intel will omit certain things, because it’s proprietary,” Mendis says. “However, when you use data, you don’t need to know the documentation. If there’s something hidden you can learn it directly from the data.”
To do so, the researchers clocked the average number of cycles a given microprocessor takes to compute basic block instructions — basically, the sequence of boot-up, execute, and shut down — without human intervention. Automating the process enables rapid profiling of hundreds of thousands or millions of blocks.
In training, the Ithemal model analyzes millions of automatically profiled basic blocks to learn exactly how different chip architectures will execute computation. Importantly, Ithemal takes raw text as input and does not require manually adding features to the input data. In testing, Ithemal can be fed previously unseen basic blocks and a given chip, and will generate a single number indicating how fast the chip will execute that code.
The researchers found Ithemal cut error rates in accuracy — meaning the difference between the predicted speed versus real-world speed — by 50 percent over traditional hand-crafted models. Further, in their next paper, they showed that Ithemal’s error rate was 10 percent, while the Intel performance-prediction model’s error rate was 20 percent on a variety of basic blocks across multiple different domains.
The tool now makes it easier to quickly learn performance speeds for any new chip architectures, Mendis says. For instance, domain-specific architectures, such as Google’s new Tensor Processing Unit used specifically for neural networks, are now being built but aren’t widely understood. “If you want to train a model on some new architecture, you just collect more data from that architecture, run it through our profiler, use that information to train Ithemal, and now you have a model that predicts performance,” Mendis says.
Next, the researchers are studying methods to make models interpretable. Much of machine learning is a black box, so it’s not really clear why a particular model made its predictions. “Our model is saying it takes a processor, say, 10 cycles to execute a basic block. Now, we’re trying to figure out why,” Carbin says. “That’s a fine level of granularity that would be amazing for these types of tools.”
They also hope to use Ithemal to enhance the performance of Vemal even further and achieve better performance automatically.
January 6, 2020 | More