MIT Research News' Journal
[Most Recent Entries]
[Calendar View]
Wednesday, May 22nd, 2019
| Time |
Event |
| 12:00a |
3Q: The fact finders When publication such as U.S. News and World Report roll out their annual university rankings, typically with MIT among the top schools listed, some may wonder where the data they’re based on actually comes from.
The source of that information is MIT Instituational Research, which collects and compiles data on many facets of the Institute, or, as Director Lydia Snover puts it, on MIT’s “people, money, and space.” The Institutional Research (IR) website is a wonderland of data that tells the story of MIT’s evolution over recent decades. There are surveys of faculty, graduate students, undergraduates — and even undergraduates’ parents. Users can also take a deep dive into the demographics of different subsets of the MIT community and peruse financial figures on research expenditures, tuition, and more.
Public universities have been providing this kind of information for decades to state and federal agencies that fund them. It’s unusual for a private university such as MIT to have such a robust IR operation and to share so much of its data publicly, but Snover has long been a leader in the field of IR at the national, and even international, level. She was recently awarded the John Stecklein Distinguished Member Award from the Association for Institutional Research, for advancing the field of institutional research through extraordinary scholarship, leadership, and service.
MIT News caught up with Snover to talk about IR at MIT, her philosophy about transparency, and why she’s a fan of the Institute’s data warehouse.
Q: What are the main types of data that your office collects, and what are they used for?
A: We bring together data from lots of different operational areas at MIT — including human resources, the registrar, admissions, and facilities, to name just a few — to simplify it in some ways and create metrics that can be used by departments, labs, and centers to help them meet their goals.
We complete all information requests for university rankings, guidebooks, and various consortiums. We also administer surveys for organizations like the Consortium for Financing of Higher Education as well as some of our other peer institutions. The majority of surveys we administer are just for the MIT community, or subsets of it. We administer over 100 surveys a year. We support the accreditation process and assist when asked with grant applications.
We provide reports for department heads in preparation for meetings with the Corporation’s visiting committees. We’ll put together a 10-year profile that includes department-level trends in staffing, retention, enrollment, sponsored research expenditures, how graduate students are being funded, things like that. We can compare those numbers within MIT and for a subset of metrics with other peer institutions.
People like to talk about making data-driven decisions, but we prefer the term “data-informed.” We collect data that help MIT’s senior officers make decisions about what’s best for the Institute.
A lot of the data we collect are available on our website, including our survey data. We have philosophy that if we ask people to fill out a survey, they’re entitled to see the results!
Q: How has your mandate changed in the last 20 years, and what do you see in the office’s future?
A: Institutional Research was established in 1986 and initially we focused primarily on physical planning. Over the next 15 years we began administering surveys, responding on behalf of the Institute to external data requests, and providing briefing materials. In 2000 we moved to the Office of the Provost, and our portfolio has continued to evolve and grow, both in terms of the services we provide to MIT leadership and the greater MIT community, and our involvement in sharing data with other universities. The staff has evolved as well to include analysts, programmers, experts in survey design, data visualization, database design, statistics, and qualitative analysis. MIT IR has an extraordinarily gifted staff.
Nationally, large institutional research offices were needed mostly by public institutions to respond to state legislatures. Private universities and colleges have slowly built up their capacity, in large part to provide internal analysis. In 1988, MIT joined the Association of American Universities Data Exchange (AAUDE), a consortium which facilitates data sharing with other AAU universities on things like the composition of faculty at the department level. The number of private AAU universities participating in the AAU Data Exchange has gone from a handful in 1990 to all 27 since I and others began encouraging our colleagues to become active.
Before MIT became involved, members were mailing each other this information on paper — you’d have file cabinets of paper! — so MIT first volunteered to provide an FTP server to facilitate electronic exchange of data. Now the AAU Data Exchange has a data warehouse, which has made the whole system very efficient.
One area were we focus a lot of attention, through our surveys and other data collections, is on what happens to our graduates: What percentage are going into industry? What are the companies that are hiring them? It used to be that all universities cared about was how many students go to graduate school, but MIT sends a lot of graduates to industry.
One new project is working with Professors Susan Hockfield, Sangeeta Bhatia, Nancy Hopkins, Fiona Murray of the MIT Innovation Initiative and the Boston Biotech working group on some interesting issues in gender representation in biotechnology, looking at company leadership, issuance of patents, and other areas.
Q: MIT’s IR office is relatively big for a private university. Why is that?
A: The scope of work for MIT’s Institutional Research Office is unusual because we’re involved in many projects that are important to MIT but not typical for institutional research. For example, at MIT we work with MITx data and sponsored research trends.
We’re very lucky and unusual because at MIT we have centralized data systems but local decision making. The fact that we have only one registrar, for example, and centralized accounting systems makes it much easier for my office to pull data together and analyze it.
I can’t emphasize enough how important the MIT data warehouse is — to everyone at MIT, not just to us. If you’re an analyst in an office like ours, you’d have to learn query languages for all the different databases. You would also spend a large proportion of your time compiling and cleaning data. But IS&T set up this system so that data could feed into one central warehouse, and you don’t need special programming skills to pull information out of it. The MIT data warehouse has been the envy of most of our peers.
MIT is the best place in the world to do institutional research because we have faculty who aren’t afraid of the data, even if they show there’s room for improvement. There’s an engineering mentality that permeates MIT. If we find we’re different from our peers in a way that we need to fix, then we identify that and fix it. You never think you’re the best because there’s always something to improve on. | | 12:15p |
Paving sustainably Although the nearly 21 million miles of paved roads around the globe appear static, their environmental footprints are anything but set.
When studying all stages of a road’s life using a technique called pavement life-cycle assessment, it becomes clear that a pavement’s environmental impact doesn't end with construction. In fact, there are significant emissions associated with a pavement during its operational life, also known as its use phase.
Several factors, like the pavement quality’s impact on fuel efficiency, lighting, and its ability to absorb carbon dioxide through carbonation all contribute to this footprint. What’s more, these factors can vary depending on the pavement’s context, which includes the climate and the amount of traffic. This can make a pavement’s use phase impacts difficult to calculate.
In a paper published in the Journal of Cleaner Production, researchers at the MIT Concrete Sustainability Hub (CSHub) examine the use phase of pavements and calculate the influence of context on their environmental footprint. Their work finds that the use phase is highly context-dependent.
Where the rubber meets the road
Although the use phase can have a sizable environmental footprint, decisions made before a pavement is even constructed can influence the size of that footprint.
“It turns out that the design and maintenance of pavements indirectly impact the environment,” explains Jeremy Gregory, CSHub executive director and an author of the recent paper. “Some of these impacts include the way that pavements impact climate through their reflectivity, through the absorption of carbon dioxide over time through the paving materials, and by how they affect the fuel consumption of the vehicles that drive on them.”
This latter effect, called pavement-vehicle interaction (PVI), causes excess fuel consumption and is one of the greatest contributors to use-phase pavement emissions.
As its name suggests, PVI refers to the interaction between a vehicle’s tires and the road it drives upon. It is a multifaceted phenomenon.
The first, and most apparent, aspect of PVI is roughness, which refers to irregularities in the surface of the pavement. In addition to affecting ride comfort, roughness can have a significant effect on fuel consumption.
“The rougher a pavement is, the more energy dissipation there is in the shock absorber system of a vehicle,” explains Gregory. “A vehicle must then consume more fuel to overcome this additional energy dissipation. We refer to this as excess fuel consumption.”
Along with roughness, the second aspect of PVI is deflection. “Deflection has to do with very heavy vehicles, primarily trucks,” notes Gregory. “The weight of a truck makes a small indentation in the pavement so that the vehicle is always driving up a very shallow hill. Like roughness, deflection also causes excess fuel consumption.”
Since excess fuel consumption only decreases fuel economy by a few percentage points, it isn’t that noticeable to the average driver. But when factoring in the often thousands of vehicles that drive across a stretch of pavement every single day, these few percentage points add up. In the case of California, excess fuel consumption on highways totaled 1 billion gallons over five years.
Considering context
While roughness and deflection contribute significantly to use-phase environmental impacts, another factor is also in play — a pavement’s context.
“When we look at the overall life cycle assessment of pavements, we find that the results are very context dependent,” says Gregory. “The context includes the climate the pavement exists in, the amount of traffic for that pavement, the type of pavement design, and also the maintenance and rehabilitation schedule that’s planned for that pavement in the future. All of those factors will combine to determine the environmental impact of a pavement.”
The authors of the paper selected nine different scenarios to study the impacts of these context-specific conditions. They analyzed pavements in four U.S. states with different climates — Missouri, Arizona, Colorado, and Florida. Within each climate zone, they then looked at roads with different traffic levels.
After studying the data, they found that traffic was the most significant factor affecting pavement environmental impacts.
“It turns out that for pavements with really high traffic loads, a much bigger fraction of their overall environmental impact is associated with the use phase and the excess fuel consumption of vehicles,” explains Gregory.
For example, interstates, which have the most traffic, also had the greatest use-phase impacts — as much as 78 percent of total life cycle impacts.
“On the other hand, for pavements that have much fewer vehicles that travel on them, most of the environmental impact is associated with the materials and construction,” reports Gregory. These kinds of less-trafficked roads, like state and rural highways, displayed lower use-phase impacts of 38 percent and 37 percent, respectively.
In addition to traffic, the design and maintenance of a pavement also influence its environmental footprint.
For example, since interstates see a lot of passenger vehicle traffic, the roughness of their pavements is their primary source of excess fuel consumption. If not regularly maintained, an interstate’s roughness might increase, leading to greater excess fuel consumption.
Since truck traffic is higher on rural and state highways than on interstates, the deflection of those pavements may have a greater impact on excess fuel consumption than roughness. To mitigate the effects of deflection, the pavement must be designed to be stiff enough to withstand a truck’s weight.
Climate also affects the environmental footprint of a pavement’s use phase. In colder climates, some pavements can deteriorate more quickly due to freeze-thaw damage, and therefore can have higher roughness. This increases the excess fuel consumption of vehicles on these pavements in cold climates.
In warmer climates, pavements made with petroleum-based materials deform more easily, which increases their susceptibility to deflection. In turn, trucks driving on these pavements in warm climates have greater excess fuel consumption.
Ultimately, this recent paper shows just how many contextual factors must be considered during a pavement’s use phase in order to make it as sustainable as possible. “It’s important to not assume any environmental impact for any given context,” explains Gregory. “You really have to run the numbers.”
The MIT Concrete Sustainability Hub is a team of researchers from several departments across MIT working on concrete and infrastructure science, engineering, and economics. Its research is supported by the Portland Cement Association and the Ready Mixed Concrete Research and Education Foundation. | | 11:59p |
Bringing human-like reasoning to driverless car navigation With aims of bringing more human-like reasoning to autonomous vehicles, MIT researchers have created a system that uses only simple maps and visual data to enable driverless cars to navigate routes in new, complex environments.
Human drivers are exceptionally good at navigating roads they haven’t driven on before, using observation and simple tools. We simply match what we see around us to what we see on our GPS devices to determine where we are and where we need to go. Driverless cars, however, struggle with this basic reasoning. In every new area, the cars must first map and analyze all the new roads, which is very time consuming. The systems also rely on complex maps — usually generated by 3-D scans — which are computationally intensive to generate and process on the fly.
In a paper being presented at this week’s International Conference on Robotics and Automation, MIT researchers describe an autonomous control system that “learns” the steering patterns of human drivers as they navigate roads in a small area, using only data from video camera feeds and a simple GPS-like map. Then, the trained system can control a driverless car along a planned route in a brand-new area, by imitating the human driver.
Similarly to human drivers, the system also detects any mismatches between its map and features of the road. This helps the system determine if its position, sensors, or mapping are incorrect, in order to correct the car’s course.
To train the system initially, a human operator controlled an automated Toyota Prius — equipped with several cameras and a basic GPS navigation system — to collect data from local suburban streets including various road structures and obstacles. When deployed autonomously, the system successfully navigated the car along a preplanned path in a different forested area, designated for autonomous vehicle tests.
“With our system, you don’t need to train on every road beforehand,” says first author Alexander Amini, an MIT graduate student. “You can download a new map for the car to navigate through roads it has never seen before.”
“Our objective is to achieve autonomous navigation that is robust for driving in new environments,” adds co-author Daniela Rus, director of the Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science. “For example, if we train an autonomous vehicle to drive in an urban setting such as the streets of Cambridge, the system should also be able to drive smoothly in the woods, even if that is an environment it has never seen before.”
Joining Rus and Amini on the paper are Guy Rosman, a researcher at the Toyota Research Institute, and Sertac Karaman, an associate professor of aeronautics and astronautics at MIT.
Point-to-point navigation
Traditional navigation systems process data from sensors through multiple modules customized for tasks such as localization, mapping, object detection, motion planning, and steering control. For years, Rus’s group has been developing “end-to-end” navigation systems, which process inputted sensory data and output steering commands, without a need for any specialized modules.
Until now, however, these models were strictly designed to safely follow the road, without any real destination in mind. In the new paper, the researchers advanced their end-to-end system to drive from goal to destination, in a previously unseen environment. To do so, the researchers trained their system to predict a full probability distribution over all possible steering commands at any given instant while driving.
The system uses a machine learning model called a convolutional neural network (CNN), commonly used for image recognition. During training, the system watches and learns how to steer from a human driver. The CNN correlates steering wheel rotations to road curvatures it observes through cameras and an inputted map. Eventually, it learns the most likely steering command for various driving situations, such as straight roads, four-way or T-shaped intersections, forks, and rotaries.
“Initially, at a T-shaped intersection, there are many different directions the car could turn,” Rus says. “The model starts by thinking about all those directions, but as it sees more and more data about what people do, it will see that some people turn left and some turn right, but nobody goes straight. Straight ahead is ruled out as a possible direction, and the model learns that, at T-shaped intersections, it can only move left or right.”
What does the map say?
In testing, the researchers input the system with a map with a randomly chosen route. When driving, the system extracts visual features from the camera, which enables it to predict road structures. For instance, it identifies a distant stop sign or line breaks on the side of the road as signs of an upcoming intersection. At each moment, it uses its predicted probability distribution of steering commands to choose the most likely one to follow its route.
Importantly, the researchers say, the system uses maps that are easy to store and process. Autonomous control systems typically use LIDAR scans to create massive, complex maps that take roughly 4,000 gigabytes (4 terabytes) of data to store just the city of San Francisco. For every new destination, the car must create new maps, which amounts to tons of data processing. Maps used by the researchers’ system, however, captures the entire world using just 40 gigabytes of data.
During autonomous driving, the system also continuously matches its visual data to the map data and notes any mismatches. Doing so helps the autonomous vehicle better determine where it is located on the road. And it ensures the car stays on the safest path if it’s being fed contradictory input information: If, say, the car is cruising on a straight road with no turns, and the GPS indicates the car must turn right, the car will know to keep driving straight or to stop.
“In the real world, sensors do fail,” Amini says. “We want to make sure that the system is robust to different failures of different sensors by building a system that can accept these noisy inputs and still navigate and localize itself correctly on the road.” |
|