MIT Research News' Journal
[Most Recent Entries]
[Calendar View]
Wednesday, February 17th, 2016
| Time |
Event |
| 12:00a |
Enabling human-robot rescue teams Autonomous robots performing a joint task send each other continual updates: “I’ve passed through a door and am turning 90 degrees right.” “After advancing 2 feet I’ve encountered a wall. I’m turning 90 degrees right.” “After advancing 4 feet I’ve encountered a wall.” And so on.
Computers, of course, have no trouble filing this information away until they need it. But such a barrage of data would drive a human being crazy.
At the annual meeting of the Association for the Advancement of Artificial Intelligence last weekend, researchers at MIT’s Computer Science and Artificial Intelligence Laboratory (CSAIL) presented a new way of modeling robot collaboration that reduces the need for communication by 60 percent. They believe that their model could make it easier to design systems that enable humans and robots to work together — in, for example, emergency-response teams.
“We haven’t implemented it yet in human-robot teams,” says Julie Shah, an associate professor of aeronautics and astronautics and one of the paper’s two authors. “But it’s very exciting, because you can imagine: You’ve just reduced the number of communications by 60 percent, and presumably those other communications weren’t really necessary toward the person achieving their part of the task in that team.”
The work could have also have implications for multirobot collaborations that don’t involve humans. Communication consumes some power, which is always a consideration in battery-powered devices, but in some circumstances, the cost of processing new information could be a much more severe resource drain.
In a multiagent system — the computer science term for any collaboration among autonomous agents, electronic or otherwise — each agent must maintain a model of the current state of the world, as well as a model of what each of the other agents takes to be the state of the world. These days, agents are also expected to factor in the probabilities that their models are accurate. On the basis of those probabilities, they have to decide whether or not to modify their behaviors.
Communication costs
In some scenarios, a robot’s decision to broadcast a new item of information could force its fellows to update their models and churn through all those probabilities again. If the information is inessential, broadcasting it could introduce serious delays, to no purpose. And the MIT researchers’ work suggests that 60 percent of communications in multiagent systems may be inessential.
The state-of-the-art method for modeling multiagent systems is called a decentralized partially observable Markov decision process, or Dec-POMDP. A Dec-POMDP factors in several types of uncertainty; not only does it consider whether an agent’s view of the world is correct and whether its estimate of its fellows’ worldviews is correct, it also considers whether any action it takes will be successful. The robot may plan, for instance, to move forward 20 feet but find that crosswinds blow it off course.
Dec-POMDPs generally assume some prior knowledge about the environment in which the agents will be operating. Because Shah and Vaibhav Unhelkar, a graduate student in aeronautics and astronautics and first author on the new paper, were designing a system with emergency-response applications in mind, they couldn’t make that assumption. Emergency-response teams will usually be entering unfamiliar environments, and the very nature of the emergency could render the best prior information obsolete.
Adding the requirement of mapping the environment on the fly, however, makes the problem of computing a multiagent plan prohibitively time consuming. So Shah and Unhelkar’s system ignores uncertainty about actions’ effectiveness and assumes that whatever an agent attempts to do, it will do.
Balancing act
When an agent acquires a new item of information — that, for instance, a given passage through a building is blocked — it has three choices: it can ignore the information; it can use it but not broadcast it; or it can use it and broadcast it.
Each of these choices has benefits but imposes costs. In Shah and Unhelkar’s model, communication is a cost. But if an agent incorporates new information into its own model of the world and doesn’t broadcast it, it also incurs a cost, as its worldview becomes more difficult for its fellows to estimate correctly. For every new item of information an agent acquires, Shah and Unhelkar’s system performs that cost-benefit analysis, based on the agent’s model of the world, its expectations of its fellows’ actions, and the likelihood of accomplishing the joint goal more efficiently.
The researchers tested their system on more than 300 computer simulations of rescue tasks in unfamiliar environments. A version of their system that permitted extensive communication completed the tasks at a rate between 2 and 10 percent higher than the version that reduced communication by 60 percent.
In the experiments, however, all the agents were electronic. “What I’d be willing to bet, although we have to wait until we do the human-subject experiments, is that the human-robot team will fail miserably if the system is just telling the person all sorts of spurious information all the time,” Shah says. “For human-robot teams, I think that this algorithm is going to make the difference between a team that can function effectively versus a team that just plain can’t.”
In a separate research project, members of Shah’s group have asked teams of human subjects to execute similar virtual rescue missions that computer systems did in the experiments reported in the new paper. Using machine-learning algorithms, the researchers have mined the results for statistics on human communication patterns, which can be incorporated into the new model to more explicitly accommodate human-robot teams.
“It is well-understood that in human teams, when one team member gains new information, broadcasting this new information to all team members is generally not a good solution, especially when the cost of communication is high,” says Tim Miller, an assistant professor of computing and information systems at the University of Melbourne in Australia. “This work has applications outside of multiagent systems, reaching into the critical area of human-agent collaboration, where communication can be costly, but more importantly, human team members are quickly overloaded if presented with too much information.” | | 10:30a |
A stake in innovation Want to encourage innovation? A new study co-authored by an MIT professor finds that little-known state laws called “constituency statutes” have significant effects on the quantity and quality of innovative business actions.
The statutes, which allow companies to prioritize the interests of “stakeholders” — often employees — rather than just shareholders, tend to allow businesses more time to bring innovations to market, rather than forcing those companies to prioritize quarterly financial results at the exclusion of new products and new activities.
“Constituency statues are pretty important,” says Aleksandra Kacperczyk an associate professor at the MIT Sloan School of Management and an author of a new paper detailing the study.
Overall, constituency states, which exist in 34 U.S. states and were largely introduced in the 1980s, raise the rate of patenting among firms by at least 6.4 percent, according to the study.
“When the company is more stakeholder-focused, one very concrete consequence is that workers are being more protected,” explains Kacperczyk, who is the Fred Kayne (1960) Career Development Professor of Entrepreneurship. “And we know from some other [research] that when that happens, then people are more willing to engage in risk-taking, which is very conducive to innovation. … Breakthrough ideas take time and [can] put your career at stake.”
Moreover, the number of citations per patent filed in states with constituency statutes rises by at least 6.3 percent, the study shows.
“There were not only more patents, but they were more original and influential, “ Kacperczyk adds.
The study is published online in the journal Management Science. Along with Kacperczyk, the other author is Caroline Flammer, an assistant professor at the University of Western Ontario.
A trade-off to obtain innovation
To conduct the study, the researchers used the so-called “differences in differences” methodology to analyze the changing rates of patent activity in the 34 states with constituency statutes, versus activity rates in the 16 states lacking them.
The statutes appear to have helped firms especially in the areas of clean energy and consumer goods.
“Increasingly, companies are engaging consumers in innovation,” Kacperczyk notes.
Ohio was the first state to adopt a constituency statute, in 1984, and Texas is the most recent to have done so, in 2006. The study looked at roughly 160,000 examples of firm performance in the U.S., using data from the National Bureau of Economic Research (NBER) Patent Data Project, as well as Standard & Poor’s Compustat database of financial information for companies.
The key mechanism at work, Kacperczyk emphasizes, is the “trade-off that you face between short-term profits and the long-term view, in that innovation takes longer to develop. … There has been consistent evidence that the market in the short term doesn’t recognize [this] value.”
Rather than feeling pressure to, say, cut a research and development group to boost the short-term bottom line, the laws enable company management to keep betting on innovation investment even when it does not maximize shareholder value at every given moment.
Warding off takeovers
In the study, the researchers do address potentially complicating factors that might seem to make the connection between constituency statutes and innovation merely coincidental. For instance: Could it be the case that innovative firms successfully lobbied to have constituency statutes enacted, and that the increase in patenting would have happened anyway?
Actually, no: Constituency statutes were often implemented to ward off potential hostile takeovers of in-state companies, in which certain investors attempt to seize control of firms to maximize short-term shareholder value.
“Hostile takeovers can be detrimental to workers and communities, so they really needed this,” Kacperczyk says. “This is precisely when the interests of shareholders are being pitted against the interests of stakeholders. You need the stakeholder supremacy model to protect the interests of stakeholders.”
Indeed, Kacperczyk concludes, constituency statutes do help a firm’s financials, but over a lengthier period of time than takeover specialists sometimes want.
“It’s better for the bottom line,” Kapcerczyk says. “To the extent that shareholders care about profits, then in the long run it aligns with shareholder interests. It’s a way of thinking about how to create value for both stakeholders and shareholders.” | | 12:59p |
Neuroscientists reverse autism symptoms Autism has diverse genetic causes, most of which are still unknown. About 1 percent of people with autism are missing a gene called Shank3, which is critical for brain development. Without this gene, individuals develop typical autism symptoms including repetitive behavior and avoidance of social interactions.
In a study of mice, MIT researchers have now shown that they can reverse some of those behavioral symptoms by turning the gene back on later in life, allowing the brain to properly rewire itself.
“This suggests that even in the adult brain we have profound plasticity to some degree,” says Guoping Feng, an MIT professor of brain and cognitive sciences. “There is more and more evidence showing that some of the defects are indeed reversible, giving hope that we can develop treatment for autistic patients in the future.”
Feng, who is the James W. and Patricia Poitras Professor of Neuroscience and a member of MIT’s McGovern Institute for Brain Research and the Stanley Center for Psychiatric Research at the Broad Institute, is the senior author of the study, which appears in the Feb. 17 issue of Nature. The paper’s lead authors are former MIT graduate student Yuan Mei and former Broad Institute visiting graduate student Patricia Monteiro, now at the University of Coimbra in Portugal.
Boosting communication
The Shank3 protein is found in synapses — the connections that allow neurons to communicate with each other. As a scaffold protein, Shank3 helps to organize the hundreds of other proteins that are necessary to coordinate a neuron’s response to incoming signals.
Studying rare cases of defective Shank3 can help scientists gain insight into the neurobiological mechanisms of autism. Missing or defective Shank3 leads to synaptic disruptions that can produce autism-like symptoms in mice, including compulsive behavior, avoidance of social interaction, and anxiety, Feng has previously found. He has also shown that some synapses in these mice, especially in a part of the brain called the striatum, have a greatly reduced density of dendritic spines — small buds on neurons’ surfaces that help with the transmission of synaptic signals.
In the new study, Feng and colleagues genetically engineered mice so that their Shank3 gene was turned off during embryonic development but could be turned back on by adding tamoxifen to the mice’s diet.
When the researchers turned on Shank3 in young adult mice (two to four and a half months after birth), they were able to eliminate the mice’s repetitive behavior and their tendency to avoid social interaction. At the cellular level, the team found that the density of dendritic spines dramatically increased in the striatum of treated mice, demonstrating the structural plasticity in the adult brain.
However, the mice’s anxiety and some motor coordination symptoms did not disappear. Feng suspects that these behaviors probably rely on circuits that were irreversibly formed during early development.
When the researchers turned on Shank3 earlier in life, only 20 days after birth, the mice’s anxiety and motor coordination did improve. The researchers are now working on defining the critical periods for the formation of these circuits, which could help them determine the best time to try to intervene.
“Some circuits are more plastic than others,” Feng says. “Once we understand which circuits control each behavior and understand what exactly changed at the structural level, we can study what leads to these permanent defects, and how we can prevent them from happening.”
Gordon Fishell, a professor of neuroscience at New York University School of Medicine, praises the study’s “elegant approach” and says it represents a major advance in understanding the circuitry and cellular physiology that underlie autism. “The combination of behavior, circuits, physiology, and genetics is state-of-the art,” says Fishell, who was not involved in the research. "Moreover, Dr. Feng's demonstration that restoration of Shank3 function reverses autism symptoms in adult mice suggests that gene therapy may ultimately prove an effective therapy for this disease."
Early intervention
For the small population of people with Shank3 mutations, the findings suggest that new genome-editing techniques could in theory be used to repair the defective Shank3 gene and improve these individuals’ symptoms, even later in life. These techniques are not yet ready for use in humans, however.
Feng believes that scientists may also be able to develop more general approaches that would apply to a larger population. For example, if the researchers can identify defective circuits that are specific for certain behavioral abnormalities in some autism patients, and figure out how to modulate those circuits’ activity, that could also help other people who may have defects in the same circuits even though the problem arose from a different genetic mutation.
“That’s why it’s important in the future to identify what subtype of neurons are defective and what genes are expressed in these neurons, so we can use them as a target without affecting the whole brain,” Feng says. | | 11:59p |
Humans settled, set fire to Madagascar’s forests 1,000 years ago There’s no question that our species has had a dramatic impact on the planet’s physical environment, particularly over the last few centuries, with the rise of modern industry, transportation, and infrastructure. But as new research shows, humans have been transforming the landscape, with lasting impacts, since long before the start of the Industrial Era.
Scientists from MIT and the University of Massachusetts at Amherst have found that a widespread and permanent loss of forests in Madagascar that occurred 1,000 years ago was due not to climate change or any natural disaster, but to human settlers who set fire to the forests to make way for grazing cattle.
The researchers came to this conclusion after determining the composition of two stalagmites from a cave in northwestern Madagascar. Stalagmites form from water that percolates from the surface, through the soil, and into a cave. These finely layered pillars can be preserved for thousands of years, and their composition serves as a historical record of the environment above ground.
From their analysis, the team found that around 1,000 years ago, both stalagmites’ calcium carbonate composition shifted suddenly and completely, from carbon isotope ratios typical of trees and shrubs, to those more consistent with grassland, within just 100 years.
Was this landscape transformation triggered by climate change? The team’s results suggest otherwise. Around the same period, they found that oxygen isotope levels remained unchanged in both stalagmites, indicating that rainfall rates — and climate in general — remained relatively stable.
“We went in expecting to just tell a climate change story, and were surprised to see a huge carbon isotope change in both stalagmites,” says David McGee, the Kerr-McGee Career Development Assistant Professor in the Department of Earth, Atmospheric and Planetary Sciences at MIT. “Both the speed at which this shift occurred and the fact that there’s no real climate signal suggest human involvement.”
The team’s results are published this week in the journal Quaternary Science Reviews.
Caving for clues
McGee, who studies the composition of stalagmites as an indicator of past climates, teamed up with lead author Stephen Burns, professor of geosciences at the University of Massachusetts at Amherst; Laurie Godfrey, professor of anthropology also at UMass Amherst; and colleagues at the University of Antananarivo in Madagascar. Godfrey has been studying the extinctions of giant lemurs that occurred in Madagascar over the past 1,000 years. Populations of other large animals declined dramatically around this time, including pygmy hippos and giant tortoises.
The megafaunal extinction was likely accelerated by habitat loss and the widespread destruction of forests at the time. However, it’s been difficult to pin down exactly why the forests shrank, and when. Scientists who have analyzed sediment deposits from ancient lakes in the region and in other parts of Madagascar have observed an increased abundance of charcoal microparticles — a signal of fire. They’ve also noticed a spike in grass pollen levels, indicating a larger extent of grasslands. But dates for these sediments are uncertain.
McGee says stalagmites offer a more precise record of environmental change.
“You’d think stalagmites in a cave are insensitive to what’s going on in the landscape above them,” McGee says. “But because they’re basically fossilized groundwater deposits, precipitated in very regular layers, they’re a fairly sensitive recorder of climate and ecosystem changes.”
In a 2014 expedition to the island, Burns, Godfrey, and their Malagasy colleagues collected samples of stalagmites from Anjohibe Cave, a large cave system in northwestern Madagascar. They sent two meter-long stalagmites to McGee to analyze at MIT.
Seeing a shift
In the lab, McGee and research scientist Benjamin Hardt determined the ages of each stalagmite’s layers by measuring the ratio of uranium to thorium — a common geological dating technique, but difficult in these samples given their relative youth. Burns then measured their carbon and oxygen isotope ratios. All plants take up carbon dioxide for photosynthesis. While carbon dioxide in the air consists of a fixed isotopic ratio of carbon-12 to carbon-13, all plants preferentially take up carbon-12. Among plants, trees and shrubs more strongly exclude carbon-13 compared with grasses.
When the dating and isotope results were put together, McGee and Burns observed a dramatic shift in the ratio of carbon-12 to carbon-13 around 1,000 years ago in both stalagmites.
“What we see in the record is that the change from carbon isotopes that look like forest, to isotopes that look like grassland, happens really rapidly, within a century, and it would be unusual for a forest to naturally completely turn into grassland that quickly,” McGee says.
With additional analysis, Burns and McGee determined there was no corresponding change in oxygen isotopes at the time, eliminating climate change, or any natural drop in precipitation, as a trigger for forest loss.
Godfrey and others have found evidence that humans settled on Madagascar around 3,000 years ago and later adopted a more agrarian lifestyle, introducing cattle to the island before 1,000 years ago. McGee says the results suggest that humans used “slash and burn techniques” around this time to create pastureland for cattle.
“I think this is one more piece of evidence that human impacts on the environment don’t just start with Europeans and the Industrial era,” McGee says.
Going forward, Godfrey says the team plans to sample more caves across Madagascar to determine the timing and extent to which humans transformed the landscape.
“The transition from ephemeral forager to dedicated agro-pastoralist occurred, probably across Madagascar, around 1,000 years ago,” Godfrey says. “We know that a dramatic landscape transformation occurred in the northwest. We know that this transformation was not triggered by climate change. But we don’t yet know whether similar shifts, also unrelated to natural aridification, occurred elsewhere on the island, and if so, when, exactly. We are currently seeking to answer these questions.”
This research was funded, in part, by the National Science Foundation and National Geographic Society. |
|