MIT Research News' Journal
[Most Recent Entries]
[Calendar View]
Monday, February 12th, 2018
| Time |
Event |
| 12:00a |
Programming drones to fly in the face of uncertainty Companies like Amazon have big ideas for drones that can deliver packages right to your door. But even putting aside the policy issues, programming drones to fly through cluttered spaces like cities is difficult. Being able to avoid obstacles while traveling at high speeds is computationally complex, especially for small drones that are limited in how much they can carry onboard for real-time processing.
Many existing approaches rely on intricate maps that aim to tell drones exactly where they are relative to obstacles, which isn’t particularly practical in real-world settings with unpredictable objects. If their estimated location is off by even just a small margin, they can easily crash.
With that in mind, a team from MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) has developed NanoMap, a system that allows drones to consistently fly 20 miles per hour through dense environments such as forests and warehouses.
One of NanoMap’s key insights is a surprisingly simple one: The system considers the drone’s position in the world over time to be uncertain, and actually models and accounts for that uncertainty.
“Overly confident maps won’t help you if you want drones that can operate at higher speeds in human environments,” says graduate student Pete Florence, lead author on a new related paper. “An approach that is better aware of uncertainty gets us a much higher level of reliability in terms of being able to fly in close quarters and avoid obstacles.”
Specifically, NanoMap uses a depth-sensing system to stitch together a series of measurements about the drone’s immediate surroundings. This allows it to not only make motion plans for its current field of view, but also anticipate how it should move around in the hidden fields of view that it has already seen.
“It’s kind of like saving all of the images you’ve seen of the world as a big tape in your head,” says Florence. “For the drone to plan motions, it essentially goes back into time to think individually of all the different places that it was in.”
The team’s tests demonstrate the impact of uncertainty. For example, if NanoMap wasn’t modeling uncertainty and the drone drifted just 5 percent away from where it was expected to be, the drone would crash more than once every four flights. Meanwhile, when it accounted for uncertainty, the crash rate reduced to 2 percent.
The paper was co-written by Florence and MIT Professor Russ Tedrake alongside research software engineers John Carter and Jake Ware. It was recently accepted to the IEEE International Conference on Robotics and Automation, which takes place in May in Brisbane, Australia.
For years computer scientists have worked on algorithms that allow drones to know where they are, what’s around them, and how to get from one point to another. Common approaches such as simultaneous localization and mapping (SLAM) take raw data of the world and convert them into mapped representations.
But the output of SLAM methods aren’t typically used to plan motions. That's where researchers often use methods like “occupancy grids,” in which many measurements are incorporated into one specific representation of the 3-D world.
The problem is that such data can be both unreliable and hard to gather quickly. At high speeds, computer-vision algorithms can’t make much of their surroundings, forcing drones to rely on inexact data from the inertial measurement unit (IMU) sensor, which measures things like the drone’s acceleration and rate of rotation.
The way NanoMap handles this is that it essentially doesn’t sweat the minor details. It operates under the assumption that, to avoid an obstacle, you don’t have to take 100 different measurements and find the average to figure out its exact location in space; instead, you can simply gather enough information to know that the object is in a general area.
“The key difference to previous work is that the researchers created a map consisting of a set of images with their position uncertainty rather than just a set of images and their positions and orientation,” says Sebastian Scherer, a systems scientist at Carnegie Mellon University’s Robotics Institute. “Keeping track of the uncertainty has the advantage of allowing the use of previous images even if the robot doesn’t know exactly where it is and allows in improved planning.”
Florence describes NanoMap as the first system that enables drone flight with 3-D data that is aware of “pose uncertainty,” meaning that the drone takes into consideration that it doesn't perfectly know its position and orientation as it moves through the world. Future iterations might also incorporate other pieces of information, such as the uncertainty in the drone’s individual depth-sensing measurements.
NanoMap is particularly effective for smaller drones moving through smaller spaces, and works well in tandem with a second system that is focused on more long-horizon planning. (The researchers tested NanoMap last year in a program tied to the Defense Advanced Research Projects Agency, or DARPA.)
The team says that the system could be used in fields ranging from search and rescue and defense to package delivery and entertainment. It can also be applied to self-driving cars and other forms of autonomous navigation.
“The researchers demonstrated impressive results avoiding obstacles and this work enables robots to quickly check for collisions,” says Scherer. “Fast flight among obstacles is a key capability that will allow better filming of action sequences, more efficient information gathering and other advances in the future.”
This work was supported in part by DARPA’s Fast Lightweight Autonomy program. | | 4:05p |
MIT neuroscientists give "invisible" cells a new look Neurons are the star of the show in brain science, but MIT researchers believe they don’t work alone to process information.
In new research funded by a $1.9 million grant from the National Institutes of Health, a team at MIT’s Picower Institute for Learning and Memory is working to uncover the likely crucial role of a supporting cast member with a stellar-sounding name: the astrocyte. The work could ultimately provide insight into many brain disorders.
Astrocytes are at least as abundant in the brain as neurons, but because they don’t spike with electrical impulses like neurons do, they’ve essentially been “invisible” in studies of how brain circuits process information, says Mriganka Sur, the Newton Professor of Neuroscience in the Department of Brain and Cognitive Sciences and director of the Simons Center for the Social Brain at MIT. Astrocytes have instead been appreciated mostly for shuttling various molecules and ions around to keep the brain’s biochemistry balanced and functioning.
While they don’t spike, astrocytes do signal their activity with increases of calcium. A decade ago in Science, Sur and colleagues used that insight to discover that astrocyte activity in the visual cortex, the part of the brain that processes vision, matched in lock-step with the activity of neurons in response to visual stimuli. That suggested that astrocytes make a vital contribution to vision processing. In the new study, Sur’s lab will investigate exactly what astrocytes are doing, for instance, to regulate the formation of neural connections called synapses and how the calcium activity arises and what difference that activity makes. They’ll look not only during the course of normal vision, but also during the critical period early in life when vision is first developing.
Using sophisticated and precise imaging tools, Sur’s team will monitor astrocyte and neuron activity in the visual cortex as mice see different stimuli. They’ll also use genetic and pharmaceutical tools to manipulate astrocyte activity. A key mechanism that’s likely involved, Sur says, is the way astrocytes deploy a molecule called GLT1 to regulate the level and time course of the neurotransmitter glutamate. Glutamate is vital because it mediates communication between neurons across synapses. By systematically manipulating the GLT1 activity of astrocytes in the visual cortex and measuring the effects, Sur says, the team will be able to determine how astrocytes contribute to the performance and formation of neural circuits.
“Just as neurons have their spiking code, we think there is an astrocyte calcium code that reflects and works in partnership with neurons,” Sur says. “That’s totally underappreciated but very important.”
The results will matter for more than just vision, Sur says. The visual cortex is a perfect model system in which to work, he says, but astrocytes are also believed to be important, if poorly understood, in disorders as wide-ranging as Alzheimer’s disease and developmental disorders such as schizophrenia and autism.
“Astrocytes are emerging as a major player because disorders of brain development have genetic origins,” Sur says. “Genes expressed in astrocytes are emerging as very important risk factors for autism and schizophrenia.”
The new grant from the National Eye Institute (grant number R01EY028219) lasts for four years.
| | 4:30p |
In fieldwork program, students take the lead A group of MIT students said “Aloha, Hawaii!” during the latest Independent Activities Period, but it wasn’t for a month of vacation. The students were tasked with conducting research and collecting data samples, which will help them further understand the environmental conditions of soil and air quality on the Island of Hawaii (a.k.a. “the Big Island”).
The research was part of the Traveling Research Environmental eXperiences (TREX) program hosted by the Department of Civil and Environmental Engineering (CEE), which offers a unique fieldwork opportunity for students.
“It is very important to us in CEE that students get hands-on research experience and tangible skills that they can take with them in their careers, especially in the field where a lot of the action is for our discipline,” says Markus Buehler, the McAfee Professor of Engineering and department head of CEE. “TREX continues to do both of these things, while producing impressive environmental research.”
TREX brings students out of the classroom to experience firsthand the benefits and challenges of fieldwork. Each year, the projects evolve and adapt to changing research interests and tools, as well as to different environmental issues in Hawaii.
“Returning to Hawaii every year has allowed us to cultivate ongoing relationships with scientists and land owners who live and work on the island,” says Ben Kocar, assistant professor of CEE and lead instructor of TREX. “As a result, our projects continuously improve, and our findings become increasingly thorough and impactful. Giving the students control over the projects also means that each year is a little bit different, because the students have their own unique backgrounds and talents.”
The students, advised by Kocar, Associate Professor Jesse Kroll, and teaching assistants Josh Moss and Ben Crawford, are the driving force behind the research. While Kocar, Kroll, Moss, and Crawford oversaw the projects and provided guidance, the fieldwork execution was the students’ responsibility. Kocar specializes in soil science, and Kroll is an expert in atmospheric chemistry, so students pursued projects on the chemical composition of soil and on monitoring particulate matter in the air.
One project worked on building and managing a network of air quality sensors across the island to monitor the levels of air pollution from volcanoes, while another project used a combination of imagery from unpiloted aerial vehicles (UAVs) and soil samples to monitor plant health. In early January, the students familiarized themselves with the basics of soil science, air quality research, and topography to prepare for the projects in Hawaii.
In addition to the scholarship behind the research, the students built air quality sensors and prepared their UAVs for the fieldwork. With the necessary skills and background information on the projects, the students were in charge of managing the projects and for the data collection.
“We were told our two main projects and given an outline of what should be completed, but all of the detailed decisions were made among our group,” explains Meghan Reisenauer, a junior in civil and environmental engineering. “We needed to decide how and where to place our air particle sensors, how and where to sample the corn and surrounding soil, as well as decide how to approach the data analysis once we had collected it.”
For the air quality project, the students also built mounts for the air quality sensors and ensured that they were angled to the correct degree, so that the solar panels would have the appropriate amount of sun. The group then installed them around the island, expanding the preexisting network created by previous TREX students. To do this, the students contacted local residents and business owners and asked to mount air quality sensors on their property.
“I learned just how important networking and simple interaction with strangers is to the success of any project. I would be lying if the concept of asking someone for permission to use their property left me without the tiniest bit of apprehension heading into our voyage,” wrote Josh Wilson, a junior in civil and environmental engineering, in a blog post about the day. What he found instead was that “everyone was more than willing to help and, moreover, interested and excited about the research we were doing.”
For the soil analysis project, teams collected soil samples from a local farm owned by Richard Ha. They compared those results with images taken with UAVs, and, scavenging a forward looking infrared (FLIR) thermal camera from a broken UAV, added an extra dimension to their data collection by capturing aerial soil and crop temperature data to determine whether heat stress was limiting crop growth. The students had to figure out how to integrate and support the FLIR camera on the UAV and how to build a sturdy, lightweight platform for the it on the UAV. Using materials from a hardware store, the students built and tested an effective platform to make their data collection possible.
A key aspect of the fieldwork was troubleshooting and changing direction at the last minute. A few days of heavy rain impeded the data collection, soaking the soil samples, making the farm inaccessible for work and impeding the UAV's ability to fly. When the government unexpectedly shut down, the students were forced to relocate from the Kilauea Military Camp in Hawaii Volcanoes National Park to nearby, privately owned housing. The move disrupted the research schedule, adding additional pressure to complete the work on time. Combined with the new housing situation, the students were challenged with maintaining an efficient workspace and performing initial analysis on the samples.
“You have this general idea of what you need to do, but you also need to adjust for weather conditions and limitations,” explains junior David Wu. “But when you’re out in the field, anything can happen, so you have to be ready for it.”
The data collection and preliminary fieldwork research from Hawaii is brought back to campus and analyzed as part of 1.092 (Traveling Research Environmental eXperience: Fieldwork Analysis and Communication) in the spring semester, giving students a chance to perform further analysis in a more controlled setting.
Despite initial setbacks and surprising results in their data collection, the students collected and analyzed a wealth of data, and presented their preliminary findings to different audiences. One night, the group met with local MIT alumni and shared their findings and research methods.
“There were alums from all different majors, who each asked very in-depth questions about their field of study in relation to our projects with the particle sensors and drone usage,” Reisenauer wrote in a blog post. “Although we didn’t have all the answers, we did our best explaining our work and fielding questions from the room.”
At the end of the program, the group also presented their research and preliminary results to local citizens of the Big Island.
“When we presented to the public, their response and questions were much broader [than the alumni questions]; they asked mainly about what we thought the long-term effects could be, or our prediction of the material in other air particles they had dealt with,” Reisenauer recalls. “They certainly seemed appreciative of our research into an issue that affects them almost every day, so it was satisfying to show our hard work into the topic!”
In addition to collecting and analyzing data, the students also partook in local activities such as hiking around the Kilauea Iki Crater and studying the plants and ecosystem that has developed at the site; snorkeling and learning about local fish species and coral; and visiting a local coffee farm, using the location as a test site for the UAV.
“Besides the beaches that people think of, there’s a lot of cool history and a lot of ecosystems in Hawaii,” Wu says. “The Big Island has almost every type of ecosystem, from deserts and rainforests, to mountains 14,000 feet tall. Since it’s so isolated, it was a good place to do research. It was like a research playground.” | | 11:59p |
Energy-efficient encryption for the internet of things Most sensitive web transactions are protected by public-key cryptography, a type of encryption that lets computers share information securely without first agreeing on a secret encryption key.
Public-key encryption protocols are complicated, and in computer networks, they’re executed by software. But that won’t work in the internet of things, an envisioned network that would connect many different sensors — embedded in vehicles, appliances, civil structures, manufacturing equipment, and even livestock tags — to online servers. Embedded sensors that need to maximize battery life can’t afford the energy and memory space that software execution of encryption protocols would require.
MIT researchers have built a new chip, hardwired to perform public-key encryption, that consumes only 1/400 as much power as software execution of the same protocols would. It also uses about 1/10 as much memory and executes 500 times faster. The researchers describe the chip in a paper they’re presenting this week at the International Solid-State Circuits Conference.
Like most modern public-key encryption systems, the researchers’ chip uses a technique called elliptic-curve encryption. As its name suggests, elliptic-curve encryption relies on a type of mathematical function called an elliptic curve. In the past, researchers — including the same MIT group that developed the new chip — have built chips hardwired to handle specific elliptic curves or families of curves. What sets the new chip apart is that it is designed to handle any elliptic curve.
“Cryptographers are coming up with curves with different properties, and they use different primes,” says Utsav Banerjee, an MIT graduate student in electrical engineering and computer science and first author on the paper. “There is a lot of debate regarding which curve is secure and which curve to use, and there are multiple governments with different standards coming up that talk about different curves. With this chip, we can support all of them, and hopefully, when new curves come along in the future, we can support them as well.”
Joining Banerjee on the paper are his thesis advisor, Anantha Chandrakasan, dean of MIT’s School of Engineering and the Vannevar Bush Professor of Electrical Engineering and Computer Science; Arvind, the Johnson Professor in Computer Science Engineering; and Andrew Wright and Chiraag Juvekar, both graduate students in electrical engineering and computer science.
Modular reasoning
To create their general-purpose elliptic-curve chip, the researchers decomposed the cryptographic computation into its constituent parts. Elliptic-curve cryptography relies on modular arithmetic, meaning that the values of the numbers that figure into the computation are assigned a limit. If the result of some calculation exceeds that limit, it’s divided by the limit, and only the remainder is preserved. The secrecy of the limit helps ensure cryptographic security.
One of the computations to which the MIT chip devotes a special-purpose circuit is thus modular multiplication. But because elliptic-curve cryptography deals with large numbers, the chip’s modular multiplier is massive. Typically, a modular multiplier might be able to handle numbers with 16 or maybe 32 binary digits, or bits. For larger computations, the results of discrete 16- or 32-bit multiplications would be integrated by additional logic circuits.
The MIT chip’s modular multiplier can handle 256-bit numbers, however. Eliminating the extra circuitry for integrating smaller computations both reduces the chip’s energy consumption and increases its speed.
Another key operation in elliptic-curve cryptography is called inversion. Inversion is the calculation of a number that, when multiplied by a given number, will yield a modular product of 1. In previous chips dedicated to elliptic-curve cryptography, inversions were performed by the same circuits that did the modular multiplications, saving chip space. But the MIT researchers instead equipped their chip with a special-purpose inverter circuit. This increases the chip’s surface area by 10 percent, but it cuts the power consumption in half.
The most common encryption protocol to use elliptic-curve cryptography is called the datagram transport layer security protocol, which governs not only the elliptic-curve computations themselves but also the formatting, transmission, and handling of the encrypted data. In fact, the entire protocol is hardwired into the MIT researchers’ chip, which dramatically reduces the amount of memory required for its execution.
The chip also features a general-purpose processor that can be used in conjunction with the dedicated circuitry to execute other elliptic-curve-based security protocols. But it can be powered down when not in use, so it doesn’t compromise the chip’s energy efficiency.
“They move a certain amount of functionality that used to be in software into hardware,” says Xiaolin Lu, director of the internet of things (IOT) lab at Texas Instruments. “That has advantages that include power and cost. But from an industrial IOT perspective, it’s also a more user-friendly implementation. For whoever writes the software, it’s much simpler.” |
|