Power in numbers applies to robots learning grasping skills

  National Public Radio said it well: How can robots learn new tasks? Practice, practice practice.

  Stefanie Tellex, Brown University assistant professor in computer science department, talked to NPR recently about getting robots to do better at real-world tasks—specifically, grasping, why grasping is hard for a robot, and what her team would like to do about it.

  Tellex works with a "Baxter," the industrial robot made by Rethink Robotics with its box-like torso and arms.

  Joe Palca, NPR science correspondent, watched the Baxter try to pick up a battery; he said watching it was a little like watching paint dry. He told her this looked pathetic. She laughed, knowingly.

  Stuff that is really hard for a robot to do is almost effortless for a person to do. A human spends little conscious awareness of exercising perception, planning and control in picking up an object. But the robot from scratch doesn't know from kazoos, batteries and pens.

  It gets information from its cameras. And that information, he said, is just a bunch of numbers.

  Palca said Tellex thinks the way robots will get better and faster in picking up unfamiliar objects is to give them programs which let them learn from experience just as a child would.

  Unsurprisingly, Tellex has had her Baxter working around the clock. It picks up objects. It puts them down. Repeat. Repeat. And so on. But Tellex also has an idea for speeding up the learning curve.

  She hopes to recruit some more Baxter robots elsewhere which are left idle off-hours during robotics research projects to do the same tasks as her Baxter to speed up the process. Pace said it evokes the saying, many hands do light work.

  Simply put, said Will Knight of MIT Technology Review, "hundreds of robots could accelerate the process by sharing knowledge."

Just how fast would this learning curve speed up learning?

  According to Knight, Tellex counted around 300 Baxter robots in research labs around the world and if each of those robots were to use both arms to examine new objects, it would be possible for them to learn to grasp a million objects in 11 days.

  The concept of a robot teaching itself how to do tasks is of interest among researchers beyond Brown. Last month, MIT Technology Review looked at the work of Lerrel Pinto and Abhinav Gupta at Carnegie Mellon University. They worked with a Baxter, giving it deep learning capabilities, placing it in front of a table with different sized objects and left it to learn how to grasp them.

  The robot was left for up to 10 hours a day. If the robot dropped an object on the floor, there were others it could continue with.

  Lab work dedicated to robots gaining better grasping skills may result in robots that will be part of daily life.

  Tellex said, "In twenty years, every home will have a personal robot which can perform tasks such as clearing the dinner table, doing laundry, and preparing dinner. As these machines become more powerful and more autonomous, it is critical to develop methods for enabling people to tell them what to do."

via

Researchers develope software for finding tipping points and critical network structures

  If you wanted to know whether shifts in the African climate during Paleolithic times correlated with the appearance and disappearance of hominin species, how would you find the answer? It's a tricky question because of the massive amounts of noisy, complicated data you would need to analyze.

  Now researchers in Germany have developed a new tool to help grapple with enormous data sets and reveal big picture trends, such as climatic tipping points and their effects on species. The researchers created a software package based on the Python programming language that unifies complex network theory and nonlinear time series analysis - two important data analysis concepts.

A complex network is just that - a social, biological or technological network with patterns of connections that are neither regular nor purely random. Nonlinear time series analyses are often used to look at complex systems, including those that unfold in a chaotic manner. Many natural phenomena, like changing weather patterns, are nonlinear in nature—as are man-made systems, like financial markets.

  

  The researchers named the software that unifies the two concepts pyunicorn. They discuss their findings in this week's CHAOS.

  "Pyunicorn works like a macroscope, [which], if used the right way, allows to distill the essence of information from a network or time series data," said Jonathan Donges, a former Ph.D. student in the group of Jürgen Kurths and co-speaker of a flagship project, called COPAN, at the Potsdam Institute for Climate Impact Research (PIK) in Germany, which aims to develop conceptual models of global socio-environmental dynamics.

The software could be used to identify critical network structures, such as bottlenecks and backbones, for transport processes, as well as revealing tipping points in climatological or physiological time series.

  Accordingly, the package's main application is the analysis of data from observations, experiments and model systems by way of graphs and time series of several quantities in parallel, such as temperature, precipitation and wind for climate, or blood pressure and breathing for physiology. By applying recurrence network analysis, which studies when a system returns to a former state, pyunicorn was able to detect tipping points in time series. This includes the aforementioned paleoclimate records, as well as the early emergence of a severe condition in pregnant women known as preeclampsia.

  Donges's previous work has involved complex networks and nonlinear time series analysis and their applications to real world data analysis. Developing the pyunicorn package involved collaborators at PIK, Humboldt University Berlin, the Stockholm Resilience Centre, Institute for Marine and Atmospheric Research Utrecht, University of Aberdeen and Nishny Novgorod State University, located respectively in Germany, Sweden, The Netherlands, the United Kingdom and Russia.

  "Many of these methods were newly developed by our team and, moreover, there was a lack of coherent software implementations for existing methods," said Donges. "Pyunicorn was developed to close this gap and to provide an integrative software framework for applying and further developing methods for complex networks and nonlinear time series analysis and their combinations."

  As its name might imply, pyunicorn is written in Python, a popular open-source programming language. The package is designed in a modular fashion that makes it easy to use in different settings, ranging from interactive analysis sessions on laptops to large-scale parallel data analysis on supercomputer clusters. As with all Python software, pyunicorn runs on a variety of operating systems, including Linux, Mac OSX, Windows and Android.

The software's versatility fulfills a key aim of the project, which was to make the software publicly available and easy to use for researchers and practitioners in a variety of fields, ranging from complex systems science to climatology, medicine, neuroscience, economics and engineering.

  "Many of the provided methods were not freely available before to the scientific community, and weren't available in the flexible and popular Python programming language," said Jürgen Kurths, who supervised the work.

  Future work for Donges and his colleagues involves speeding up the package's code and ensuring compatibility with the Python 3.x platform. Donges remains optimistic but cautious about the uses of the package.

  "Combining well-known approaches in a new way can yield exciting insights and perspectives in complex systems science," he said. "Software packages such as pyunicorn can be highly useful in catalyzing this process, but need to be applied in a thoughtful and theory-based way. Otherwise, the result might be junk science."

  The pyunicorn package can be freely downloaded at: https://github.com/pik-copan/pyunicorn.

via

Shape-changing LineFORM may belong to interface future

  LineFORM from the Tangible Media Group at MIT Media Lab is the result of its creators asking questions. What if we have a shape-changing material that consists of a Line? Using such material, how will interactions with computers or tools change?

  They designed a shape-changing interface and described it in their research paper, "LineFORM: Actuated curve interfaces for display, interaction and constraints." Ken Nakagaki, Sean Follmer and Hiroshi Ishii make up the team.

  "Lines have several interesting characteristics from the perspective of interaction design: abstractness of data representation; a variety of inherent interactions / affordances; and constraints as boundaries or borderlines," they said.

  "By utilizing such aspects of lines together with the added capability of shape-shifting, we present various applications in different scenarios."

The video of their work demonstrates a variety of shape changes. These include body constraints and data manipulation, to investigate the design space of line-based, shape-changing interfaces. In their concept, the material can also reshape itself into telephone mode. A smart wristband gives the user haptic feedback.

Engadget referred to LineFORM as a "serpentine robot." It involves a linear series of actuators, wrote Aaron Souppouris, which can move independently or together to arrange itself in new shapes.

  

  They said in their paper that "The overall design of the system involves three parts. There is a series of connected servo motors. It has an Arduino Mega microcontroller for motor control and sensing. It has a Mac OS computer running custom applications written in processing that control LineFORM."

  Their paper is an accepted TOCHI (ACM Transactions on Computer-Human Interaction) paper for the ACM Symposium on User Interface Software and Technology (UIST) from November 8 to 11 in Charlotte, NC.

  The authors are exploring this concept because they see a future where devices in the nature of their LineFORM would be paired with flexible displays. They see the concept as "next generation mobile devices." As for the range of functions: Their concept could be used to display complex information, provide affordances on demand for different tasks and constrain user interaction.

  "We have shown that a relatively small number of actuators can be used to achieve an expressive display, and these systems may be easier to prototype than other form factors of high resolution shape display."

  Where do they go from here? They hope their work will be of interest to researchers. "Our hope is that this work will motivate others to further explore the space of actuated curve interfaces, from novel actuators to new interaction techniques."

  On that note and practically speaking, Kelsey Campbell-Dollaghan in Gizmodo said, "they imagine Lineform could replace a lot of the hardware we need for interacting with the world today–the keyboards, the phones, the cables, and so on–acting like a plug-and-play interface that could transform based on how you need to interact at a given moment."

  In the bigger picture, the Tangible Media Group has a clear focus on which technology path they will follow as they continue their innovative research: "From the three approaches in design research: technology-driven, needs-driven, and vision-driven, we focus on the vision-driven approach due to its lifespan."

  They said they know technologies become obsolete in ~1 year, users' needs change quickly and dramatically in ~10 years. "However, we believe that a clear vision can last beyond our lifespan. While we might need to wait decades before atom hackers (like material scientists or self-organizing nano-robot engineers) can invent the necessary enabling technologies for Radical Atoms, we strongly believe the exploration of interaction design should begin from today."
via

Chrome for Android vulnerability discovered by researcher

  Making news this week at the MobilePwn2Own event at the PacSec conference in Tokyo: an exploit of Google's Chrome for Android—in one shot, said PacSec organizer Dragos Ruiu. Researcher Guang Gong showcased the exploit. (PacSec is a computer security event. James O'Malley in TechRadar described it as "a meeting of security experts who show off what they've discovered for the kudos.")

  Ruiu talked to Vulture South, the Asia-Pacific bureau based in Sydney of The Register. He told Vulture South the exploit was demonstrated on a Google Project Fi Nexus 6. The exploit targets the JavaScript v8 engine and was notable, said the report, as "a single clean exploit that does not require multiple chained vulnerabilities to work."

While the exploit was not disclosed in full detail, Ruiu described the results in The Register. As soon as the phone accessed the website, the JavaScript v8 vulnerability in Chrome was used to install an arbitrary application, without user interaction, to demonstrate control of the phone. In theory, it would translate into unauthorized code running on a person's phone.

V8 is Google's open source JavaScript engine. V8 is written in C++ and is used in Google Chrome, the open source browser from Google.

  A Google security engineer on site received the bug. Softpedia stated that "A Google engineer immediately got in contact with Gong after his presentation, and rumors have it that the Chrome team is already getting a fix ready."

(Not responding specifically to this event but relevant, the HackerOne blog recently observed how "data show that programs that respond quickly to new reports, and keep open communication channels during the triage and resolution process, tend to get more reports and more repeat researchers, leading to a virtuous, security-enhancing cycle. In addition, the timely resolution of vulnerabilities reduces the risk of potential exploitation, leading to greater security.")

  Gong is a security researcher at Qihoo 360. "Thankfully," commented 9to5Google, "the exploit was developed by someone whose job it is to find vulnerabilities, and not a hacker with malicious intent."

  Ruiu will fly Gong to the CanSecWest security conference next year, said The Register.

The Android security team recognizes those who help to improve Android security by responsibly reporting vulnerabilities or by committing code with positive impact on Android security.

  In the bigger picture, Fortune senior writer Barb Darrow observed that "Given the high interest level in hacking and growing intensity of security breaches, there is definitely a need for legitimate hackers to test the limits of software."

Gong said it took him three months of work prior to the competition to find the hole, according to Business Insider Australia.

  "Good news here is that since it's through Chrome, we don't need to wait for an OTA to be approved by the manufacturers, and then the carriers," said Android Headlines on Friday.

via

A printable, flexible, lightweight temperature sensor

  A University of Tokyo research group has developed a flexible, lightweight sensor that responds rapidly to tiny thermal changes in the range of human body temperature. This sensor is expected to find healthcare and welfare applications in devices for monitoring body temperature, for example of newborn infants or of patients in intensive care settings.

  Flexible and wearable devices are increasingly being developed for healthcare and other applications where temperature and other sensors are integrated to provide feedback on patient health and wellbeing. Body temperature is a fundamental measurement and many low-cost flexible temperature sensors have been demonstrated, but devices developed to date require external circuitry to amplify the signal to allow accurate temperature measurement.

  

  In their latest research, Professor Takao Someya and Dr. Tomoyuki Yokota's research group at the Graduate School of Engineering have developed a new printable, flexible, lightweight temperature sensor that shows a very high change in electrical resistance of up to 100,000 times over a range of just five degrees centigrade, allowing accurate temperature measurement without additional complicated display circuitry.

  The key to the new sensor is the ability to precisely control the target temperature of the sensors. The sensor is composed of graphite and a semicrystalline acrylate polymer formed of two monomers, molecules that bond together to form a polymer chain. The target temperature range at which the sensor is most precise can be selected simply by altering the proportions of the two monomers. The research group achieved target temperatures between 25 and 50 degrees centigrade, a range which includes average human body temperature, and simultaneously realizing response times of less than 100 milliseconds and a temperature sensitivity of 0.02 degrees centigrade. The device was also stable even under physiological conditions, providing repeated readings up to 1,800 times.

via

Data analytics on driving behavior help users improve safety and lower insurance rates

  Those who consider themselves safe drivers may tailgate, speed, or use cellphones while driving, which significantly increase the probability of an accident, Cordova says. "For most of us, the most dangerous thing you do from day to day is driving," he says.

  To improve driver safety, Censio has developed an app that captures and analyzes data on driving behavior to show drivers where they can improve. In September, Progressive Insurance began piloting the app with customers nationwide, with aims of reducing insurance rates for good drivers.

  In a way, the app acts as a sort of "external brain" for drivers, Cordova says, helping them see the risks associated with certain driving behaviors—especially distracted driving. "The human brain is not good at statistics and probability, so most people aren't thinking how sending this text will affect their probability of getting into an accident," he says. "We calculate these complicated probability distributions and send that back to the app in a very digestible way."

  The National Highway Traffic Safety Administration attributes one accident every 24 seconds in the United States to distracted driving; the National Safety Council estimates that 1.6 million crashes annually are due to cellphone use, with another 1 million due to texting while driving.

  Beyond safety, there's also a monetary incentive: The startup hopes to shift insurers toward user-based insurance programs, where rates are based on how well a person drives. Those policies could lower rates for the 200 million insured drivers across the United States, Censio president Kevin Farrell says, but are being held back by the logistics and costs of introducing hardware into cars. "We believe bringing an app into the market really opens things up," he says.

Breaking bad driving habits

  To capture driving behavior, the app identifies when a person is driving—not, for instance, in the backseat of a cab—and then uses a smartphone's accelerometer, gyroscope (positioning), and GPS to track driving dynamics. Added to that is external data, such as on speed limit, weather, and street information, such as on safe or dangerous intersections.

  Using this data, the app looks for habits such as speeding through intersections and braking hard, which could indicate tailgating or not paying attention to the road. It also observes cellphone use while driving.

  Analyzing all this data, the app then scores the driver, on a score of 1 to 100, and keeps track of the habits a driver has, or may need to improve upon. Scores are also compared with other drivers across the nation. A user, for instance, may brake hard more frequently, or pick up their phone less often, than the national average.

  That data are shared with insurers to help them assess the overall risk of a particular driver.

  Behind the scenes, Cordova says the app earned the business of Progressive, over 10 competitors, because it doesn't drain a phone's battery and is as accurate as a hardware-based solution.

  Beating battery drain, Cordova says, came down to developing optimization algorithms that collect data only when needed, and shut off when not collecting data. Gaining accuracy, he says, was about cleaning up messy data. "When the data comes in from the car, it's extremely noisy," Cordova says. "A lot of the engineering effort went into better signal processing and machine learning to clean up those signals."

Driven by social change

  Leaving an engineering position at CERN in 2011, Cordova came to MIT as a PhD student in electrical engineering and computer science with aims of "making a social difference." (He left after completing his master's degree to pursue Censio.)

  While taking 15.390 (New Ventures), he met Censio co-founder and current vice president of operations Joe Adelmann, a Harvard University student who shared his inclination toward social change.

  Inspired to launch a startup, Cordova and Adelmann designed a system that used a smartphone to alert people who fall asleep at the wheel. Mounting the camera on the windshield, the camera would track the road and the app would calculate various driving data, such as braking frequency and vehicle positioning.

  That system wasn't exactly practical—but at its core was an app that could collect movement and driving data. As a test, several MIT students downloaded the app to their phones, and it accurately predicted if the students were walking, sitting at a desk, or climbing stairs—and when they were about to get in a car and drive.

  But that app drained a phone's battery with a day of use. "As an engineer, you try to make the most awesome product, but you miss something like battery life," Cordova says. "We had to go back to the drawing board."

  After much refining, the end product was a prototype for the Censio app that collected necessary data in the background to let drivers know the risks of their behavior.

  This prototype landed Cordova and Adelmann a finalist spot in MIT's $100K Entrepreneurship Competition in 2012. They further developed the business in office space in the Martin Trust Center for MIT Entrepreneurship and Harvard's Innovation Lab, where they met co-founder Jon McNeill, an entrepreneur with prior experience in the insurance industry, and co-founder Scott Griffith, former CEO of Zipcar.

  Three years later, the startup earned its initial funding round of $10 million and the partnership with Progressive, which is offering the app as a replacement for its Snapshot hardware component—a hub that plugs into a car's onboard diagnostics port—currently being used by more than 3 million participants. The plan is to officially release a commercial version of the app early next year.

  Although the app has proven to have a viable business model, Cordova says the mission still harkens back to his MIT years, with a focus on social change: "The aim is to make drivers around the world better and safer."

via