Researchers Develop Soft Robot That Shifts from Land to Sea with Ease

Mobile robotics in multiple domains.

ACM TECHNEWS

Researchers Develop Soft Robot That Shifts from Land to Sea with Ease

By Carnegie Mellon University, March 21, 2023

Actuators allow the reconfigurable robot to curl its body to swiftly roll away.

Credit: Morphing Matter Lab

Soft robots developed by researchers at Carnegie Mellon University can transition from walking to swimming or crawling to rolling, shifts found in most animals.

The researchers created a bistable actuator using three-dimensionally printed soft rubber with alloy springs that contract in response to electrical currents, allowing the actuator to bend. The robot remains in the new shape until it reverts to its previous configuration in response to another electrical charge. Only a hundred milliseconds of electrical charge is needed to change shape.

The researchers created robots that can walk and swim, crawl and jump, and crawl and roll.

"Our bistable actuator is simple, stable and durable, and lays the foundation for future work on dynamic, reconfigurable soft robotics," says Dinesh K. Patel, a post-doctoral fellow in CMU's Morphing Matter Lab.

From Carnegie Mellon University

View Full Article  

Wi Fi Routers Detect People in a Room

Intriguing capability, unexpected. 

Wi-Fi Routers Can Detect Human Locations, Poses Within a Room

Tom's Hardware, Mark Tyson, January 18, 2023

Carnegie Mellon University scientists have been testing a system that uses Wi-Fi signals to detect the positions and poses of people in a room. The researchers positioned TP-Link Archer A7 AC1750 Wi-Fi routers at either end of the room, while algorithms generated wireframe models of people in the room by analyzing the signal interference the people caused. The researchers based the perception system on Wi-Fi signal channel-state-information, or the ratio between transmitted and received signal waves. A computer vision-capable neural network architecture processes this data to execute dense pose estimation; the researchers deconstructed the human form into 24 segments to accelerate wireframe representation. They claim the wireframes' position and pose estimates are as good as those generated by certain "image-based approaches."  ... '

Low Cost, Omni Purpose Robotics

Robust Robotics Solutions 

Low-Cost Robot Ready for Any Obstacle

Carnegie Mellon University News

Aaron Aupperle, November 16, 2022

Scientists at Carnegie Mellon University (CMU) and the University of California, Berkeley, have enabled a low-cost and relatively small legged robot to adapt to obstacles. The robot uses its vision and an onboard computer to quickly adjust to new situations and master difficult terrain. The researchers trained it using 4,000 robot clones as they walked and climbed in a simulator, giving the machine six years of experience in one day. The simulator also retained motor skills acquired in training in a neural network that the team copied to the actual robot. "This system uses vision and feedback from the body directly as input to output commands to the robot's motors," explained CMU's Ananye Agarwal. "This technique allows the system to be very robust in the real world."  ... ' 

Low Cost Legged Robotics

 Low cost,  especially useful for testing out proposed uses,  sounds good. 

Low-Cost Robot Ready for Any Obstacle   

By Carnegie Mellon University News, November 18, 2022

A robotic system designed by researchers at Carnegie Mellon University's School of Computer Science and the University of California, Berkeley, enables small, low-cost legged robots to maneuver in challenging environments.

Scientists at Carnegie Mellon University (CMU) and the University of California, Berkeley, have enabled a low-cost and relatively small legged robot to adapt to obstacles.

The robot uses its vision and an onboard computer to quickly adjust to new situations and master difficult terrain.  The researchers trained it using 4,000 robot clones as they walked and climbed in a simulator, giving the machine six years of experience in one day.  The simulator also retained motor skills acquired in training in a neural network that the team copied to the actual robot.

"This system uses vision and feedback from the body directly as input to output commands to the robot's motors," explained CMU's Ananye Agarwal. "This technique allows the system to be very robust in the real world."

From Carnegie Mellon University News

View Full Article    

Network Pruning Can Skew Deep Learning Models

Making Better Models

Network Pruning Can Skew Deep Learning Models

North Carolina State University News

Matt Shipman, November 2, 2022

Computer science researchers at North Carolina State (NC State), Syracuse, and Carnegie Mellon universities have shown that neural network pruning can undermine the performance of deep learning models at identifying certain groups. The researchers cited disparities in gradient norms across groups, and in Hessian norms linked to inaccuracies of a group's data, as factors impacting performance. This implies network pruning can compound existing accuracy deficiencies. NC State's Jung-Eun Kim said the team has demonstrated a remedial mathematical method "to equalize the groups that the deep learning model is using to categorize data samples." Tests of the mitigation technique found basically restored a deep learning model to pre-pruning levels of accuracy.  ... '

Full Article

Machine Learning Model Predicts MS Health Conditions

A very good application direction

Machine Learning Model Predicts Health Conditions of People with MS

By Carnegie Mellon University

Researchers at Carnegie Mellon University, the University of Pittsburgh, and the University of Washington developed a machine learning model that can predict the health of chronic neurological disorder patients during stay-at-home periods.

The researchers amassed sensor data from smartphones and fitness trackers of multiple sclerosis (MS) patients before and during the early COVID-19 surge. They fed the data into the model to anticipate depression, fatigue, poor sleep quality, and exacerbated MS symptoms. Their work is published in the Journal of Medical Internet Research Mental Health.

"We were able to capture the change in people's behaviors and accurately predict clinical outcomes when they are forced to stay at home for prolonged periods," says Mayank Goel, head of the Smart Sensing for Humans Lab at CMU. "Now that we have a working model, we could evaluate who is at risk for worsening mental health or physical health, inform clinical triage decisions, or shape future public health policies." ... 

The researchers' model can predict how stay-at-home orders affect the mental health of people with chronic neurological disorders ... 

From Carnegie Mellon University   View Full Article  

Watching Humans to Learn

Humans can be taught,  carefully.   dealt with lots of these.  

 Robots Learn Household Tasks by Watching Humans

Carnegie Mellon University School of Computer Science

Aaron Aupperle, July 20, 2022

Carnegie Mellon University's Shikhar Bahl, Deepak Pathak, and Abhinav Gupta developed the In-the-Wild Human Imitating Robot Learning (WHIRL) algorithm to teach robots to perform tasks by observing people. WHIRL enables robots to gain knowledge from human-interaction videos and apply that data to new tasks, making them well-suited to learning household chores. The researchers outfitted a robot with a camera and the algorithm, and it learned to complete more than 20 tasks in natural environments. In each case, the robot watched a human execute the task once, then practiced and learned to complete the task by itself. "Instead of waiting for robots to be programmed or trained to successfully complete different tasks before deploying them into people's homes, this technology allows us to deploy the robots and have them learn how to complete tasks, all the while adapting to their environments and improving solely by watching," Pathak explained.  ... ' 

Algorithm Optimally Divvies up Tasks for Human-Robot Teams

Good look at the cooperative issue. 

 ACM NEWS

Algorithm Optimally Divvies up Tasks for Human-Robot Teams  By SciTechDaily, May 26, 2022

As robots increasingly join people working on the factory floor, in warehouses, and elsewhere on the job, determining who will do which tasks increases in complexity and importance. People are better suited for some jobs, robots for others. And in some cases, it is advantageous to spend time teaching a robot to do a task now and reap the rewards later.

Researchers at Carnegie Mellon University's Robotics Institute (RI) have developed an algorithmic planner that helps delegate tasks to humans and robots. The planner, "Act, Delegate or Learn" (ADL), considers a list of duties and decides how best to assign them. The researchers asked three questions: When should a robot act to complete a task? When should a task be delegated to a human? And when should a robot learn a new task?

“There are costs associated with the decisions made, such as the time it takes a human to complete a task or teach a robot to complete a task and the cost of a robot failing at a task,” said Shivam Vats, the lead researcher and a Ph.D. student in the RI. “Given all those costs, our system will give you the optimal division of labor.”

From SciTechDaily

View Full Article  

Making Pizza etc. with Robots

Here a long time amateur cook, very interested in how this could all be well automated.  Starting to see it in fast food. Like also the connection to things like clothing.  

Solving the challenges of robotic pizza-making

A new technique could enable a robot to manipulate squishy objects like pizza dough or soft materials like clothing.

Adam Zewe | MIT News Office   Publication Date: March 31, 2022

Imagine a pizza maker working with a ball of dough. She might use a spatula to lift the dough onto a cutting board then use a rolling pin to flatten it into a circle. Easy, right? Not if this pizza maker is a robot.

For a robot, working with a deformable object like dough is tricky because the shape of dough can change in many ways, which are difficult to represent with an equation. Plus, creating a new shape out of that dough requires multiple steps and the use of different tools. It is especially difficult for a robot to learn a manipulation task with a long sequence of steps — where there are many possible choices — since learning often occurs through trial and error.

Researchers at MIT, Carnegie Mellon University, and the University of California at San Diego, have come up with a better way. They created a framework for a robotic manipulation system that uses a two-stage learning process, which could enable a robot to perform complex dough-manipulation tasks over a long timeframe. A “teacher” algorithm solves each step the robot must take to complete the task. Then, it trains a “student” machine-learning model that learns abstract ideas about when and how to execute each skill it needs during the task, like using a rolling pin. With this knowledge, the system reasons about how to execute the skills to complete the entire task.

The researchers show that this method, which they call DiffSkill, can perform complex manipulation tasks in simulations, like cutting and spreading dough, or gathering pieces of dough from around a cutting board, while outperforming other machine-learning methods. .... '   

Pamela McCorDuck Dies

 We met with and used some of her  books in training exercises in the early days of AI:

Pamela McCorduck, Historian of AI, Dies at 80

The New York Times, Richard Sandomir, November 4, 2021

Pamela McCorduck, who authored a history of the first two decades of artificial intelligence (AI), has died at 80. She first co-edited an influential book of academic papers on AI at the University of California, Berkeley with computer scientists Edward Feigenbaum (an ACM A.M. Turing Award recipient) and Julian Feldman. As an English teacher at Carnegie Mellon University, McCorduck got to know AI pioneers like Turing Award recipients Herbert Simon and Raj Reddy. Feigenbaum said, "She was dumped into this saturated milieu of the great and greatest in AI at Carnegie Mellon—some of the same people whose papers she'd helped us assemble—and decided to write a history of the field." The book was "Machines Who Think: A Personal Inquiry Into the History and Prospects of Artificial Intelligence." Said Simon, "She was interacting with all the movers and shakers of AI. She was in the middle of it, an eyewitness to history."  .... 

Robots Adapt to Challenging Terrain in Real Time

Terrain-able robotics have been very impressive in recent years.   Here yet more advances by some relatively unexpected parties.  Ultimately will it become rare to have humans repetitively port around themselves and things they can carry?  

Stumble-Proof Robot Adapts to Challenging Terrain in Real Time. By TechCrunch, July 16, 2021

A new robotic locomotion model capable of real-time terrain adaptation has been developed by a multi-institutional research team.

Engineers at Facebook AI, the University of California, Berkeley (UC Berkeley), and Carnegie Mellon University based Rapid Motor Adaptation (RMA) on the ability of humans and animals to quickly and unconsciously adjust their locomotion to different conditions.

The team trained the system in a virtual model of the real world, where the robot's brain learned to maximize forward motion with the least amount of energy, and to avoid falls by responding to incoming data from physical sensors.

UC Berkeley's Jitendra Malik said the robot employs absolutely no visual input, instead closely monitoring itself.

The RMA system uses a constantly running main gait-control algorithm and a parallel adaptive algorithm that watches internal readings and provides the main model adjustment data in response to terrain changes.

From TechCrunch   Full article.

IOT Security from CyLab

Looks to be a good effort underway.   In general IOT devices have high levels of security danger, because they are minimally protected to begin with, are often placed in networks where they are open to external threats, and in practice are rarely patched against newly discovered danger.   There is also a tendency for consumer IOT to be cheaply coded and developed, with inadequate testing for security.  All this driven by their cost being needed to sell in the consumer market.

CyLab's IoT Security, Privacy Label Effectively Conveys Risk

Carnegie Mellon University CyLab Security and Privacy Institute

By Daniel Tkacik, May 26, 2021

Researchers found that Carnegie Mellon University CyLab's prototype security and privacy label adequately conveys the risks associated with the use of Internet-connected devices. Their study involved 1,371 participants who were given a randomly assigned scenario about buying a smart device, and asked whether information on the label would change their risk perception and their willingness to purchase. The label detailed a device's privacy and security practices, like the purpose of data collection and with whom data is shared. Most of the attributes on the label resulted in accurate risk perceptions, although the study found some misconceptions. Researcher Pardis Emami-Naeini said, "Our findings suggest that manufacturers need to provide consumers with justifications as to why patching may be necessary, why it takes them a specific amount of time to patch a vulnerability, and why it might not be practical to patch vulnerabilities faster."

On TAB Overloads in Chrome

 Well yes, will give this a try.

Overcoming Tab Overload   By Carnegie Mellon University School of Computer Science

In a study of Internet browser tab usage, computer scientists at Carnegie Mellon University (CMU) found that tab overload is an issue for many people.

The researchers assessed tab use via surveys and interviews, asking why people kept tabs open and why they closed them. They found that despite being overwhelmed by the number of open tabs, people did not want them hidden for fear they would not go back to them.

The researchers also created a Google Chrome browser extension, dubbed Skeema, to turn tabs into tasks. Skeema leverages machine learning to suggest how open tabs could be grouped into tasks and allows users to organize, prioritize, and switch between them.

CMU's Joseph Chee Chang said, "Our task-centric approach allowed users to manage their browser tabs more efficiently, enabling them to better switch between tasks, reduce tab clutter, and create task structures that better reflected their mental models."

From Carnegie Mellon University School of Computer Science

Open Source Intrusion Detection

First I have seen such a capability offered.   Worth a look to see about following: 

 World's Fastest Open-Source Intrusion Detection Is Here

Carnegie Mellon University CyLab Security and Privacy Institute by Daniel Tkacik

Researchers in Carnegie Mellon University's CyLab Security and Privacy Institute have developed an open source intrusion detection system that achieves speeds of 100 gigabits per second on a single server. The team programmed a field-programmable gate array (FPGA) for intrusion detection, and crafted algorithms that cannot run on traditional processors. CyLab's Justine Sherry said the server’s five cores are necessary because the FPGA processes an average 95% of data packets when placed in a network, with the remaining 5% shunted to central processing units when the array is overwhelmed. The system consumes 38 times less power than hundreds of processing cores would in executing the same tasks.   ... '

Automated Math Reasoning

In our earliest AI courses,we learned about theorem proving using AI. And yes, it was not automated math reasoning.  But it gave you the hope that it could be done, if only you could state the problem at hand as purely mathematical.   Or even parts of it.  But it was never so.  Like the article says, it rarely intersects exactly with the real world, except for elements of the real world that are also approximations within contexts.  Bottom line, its still hard.    Good article explains it, with hopes for the next generation.

How Close Are Computers to Automating Mathematical Reasoning? in Quanta Mag.  Stephen Ornes
Contributing Writer 

AI tools are shaping next-generation theorem provers, and with them the relationship between math and machine.

n the 1970s, the late mathematician Paul Cohen, the only person to ever win a Fields Medal for work in mathematical logic, reportedly made a sweeping prediction that continues to excite and irritate mathematicians — that “at some unspecified future time, mathematicians would be replaced by computers.” Cohen, legendary for his daring methods in set theory, predicted that all of mathematics could be automated, including the writing of proofs.

A proof is a step-by-step logical argument that verifies the truth of a conjecture, or a mathematical proposition. (Once it’s proved, a conjecture becomes a theorem.) It both establishes the validity of a statement and explains why it’s true. A proof is strange, though. It’s abstract and untethered to material experience. “They’re this crazy contact between an imaginary, nonphysical world and biologically evolved creatures,” said the cognitive scientist Simon DeDeo of Carnegie Mellon University, who studies mathematical certainty by analyzing the structure of proofs. “We did not evolve to do this.”

Computers are useful for big calculations, but proofs require something different. Conjectures arise from inductive reasoning — a kind of intuition about an interesting problem — and proofs generally follow deductive, step-by-step logic. They often require complicated creative thinking as well as the more laborious work of filling in the gaps, and machines can’t achieve this combination.

Computerized theorem provers can be broken down into two categories. Automated theorem provers, or ATPs, typically use brute-force methods to crunch through big calculations. Interactive theorem provers, or ITPs, act as proof assistants that can verify the accuracy of an argument and check existing proofs for errors. But these two strategies, even when combined (as is the case with newer theorem provers), don’t add up to automated reasoning.  .... "

IOT Devices in the Hotel Room

Intriguing aspects of IOT in hospitality.   Has been less in the area of assistants than I expected.

Shedding Light (and Sound) on Hidden IoT Devices in Your Next Hotel Room

Carnegie Mellon University Human-Computer Interaction Institute  By Daniel Tkacik

A study by researchers at Carnegie Mellon University (CMU)'s Human-Computer Interaction Institute and CyLab Security and Privacy Institute, working with colleagues from China’s Xi'an Jiaotong University, explored methods for detecting hidden Internet of Things (IoT) devices, using light and sound. The authors considered three locator/detector designs—placing a light-emitting diode (LED) on a device; placing an LED and beeping mechanism on a device; and a contextualized picture that showed the device in position, taken by the hospitality host. Participants pinpointed devices much faster with locator designs than without them, and about two-thirds of participants preferred the LED-plus-beeper design. CMU's Jason Hong said, "Our hope is that the findings in this paper can help industry and policymakers in adopting the idea of locators for IoT devices ... " 

Robots as Army Assistants

Robots following orders.   We know this is still hard, and depends much upon the nature of an order.  Its good that many consumers are getting used to the idea of talking to smart speakers, and getting used to the implications of understanding, planning, sensors and results.

U.S. Army Creating Robots That Can Follow Orders
Technology Review
by David Hambling

Researchers at the U.S. Army Research Laboratory (ARL), the Massachusetts Institute of Technology, Carnegie Mellon University, the National Aeronautics and Space Administration's Jet Propulsion Laboratory, and Boston Dynamics have developed software that allows robots to understand verbal instructions, carry out those instructions, and report back. The robot can accept verbal instructions, interpret gestures, or be controlled via a tablet to return data in the form of maps and images. The researchers used deep learning to teach the system to identify an object, as well as providing a knowledge base for more detailed information that helps the robot carry out its orders. Said ARL's Ethan Stump, "The robot can make maps, label objects in those maps, interpret and execute simple commands with respect to those objects, and ask for clarification when there is ambiguity in the command." ..... '

Matching for Identification of Antibiotics

Refreshing here is that there is no claim for AI.  We need all kinds of analytics to enhance our skills.

Computational 'Match Game' Identifies Potential Antibiotics
Carnegie Mellon News    By Byron Spice

Carnegie Mellon University (CMU) computational biologists collaborated with researchers at seven other institutions to develop a software tool that identifies bioactive molecules and the microbial genes that generate them, for assessment as potential antibiotics. The team demonstrated that MetaMiner can detect such molecules at least 100 times faster than was possible with previous techniques. MetaMiner applies genome mining methodology, analyzing gene clusters to deduce molecules the genes produce. CMU's Hosein Mohimani and Liu Cao bypassed genome mining's high susceptibility to error by building an error-tolerant search engine that finds matches between databases of microbial DNA and databases that classify molecular products according to mass spectra. With MetaMiner, the researchers identified 31 known and seven previously unknown ribosomally synthesized and post-translationally modified peptides in about 14 days; Mohimani said obtaining those results manually likely would have taken decades. ... "

NELL and Machine Learning

In a re-examination of 'machine learning', the broader definition that includes but is not restricted to deep learning.   I recalled our look the CMU effort called NELL,  Never Ending Language Learner'.  At the time it was insufficiently oriented to our needs, any information about the actual use of its learning?  Here is the intro to it on the Carnegie Mellon site.

NELL: The Computer that Learns,   Professor Tom Mitchell

Tom Mitchell's two daughters are grown but watching his newest 'baby' learn to read is an unprecedented achievement.

Professor Mitchell leads the team that developed the Never-Ending Language Learner – NELL – a computer system that, over time, is teaching itself to read and understand the web.
"I've been interested for many, many years in how machines learn because I'm also interested in how humans learn," explained Mitchell, who heads Carnegie Mellon's Machine Learning department – the first and only department of its kind in the world. "NELL comes naturally out of that. The current machine-learning algorithms are very different in style than how you and I learn. They analyze a single data set, output an answer, and then you turn them off. That's not like us at all! The idea of NELL is to capture a style more like the on-going learning of humans."

Understanding language – the way humans do – depends on both context and background knowledge gained over time. So NELL scans the web – attempting to "read" hundreds of millions of web pages on a fact-finding mission.

For example, the repeated combination of a phrase like "New York City Marathon" in combination with other words has taught NELL to learn that it's a "race" and a "sports event." ...  "

See also:
Project website:  http://rtw.ml.cmu.edu/rtw/
On twitter:           https://twitter.com/cmunell

Voices in AI Podcast: A Conversation with Norman Sadeh

We had early AI connects with Carnegie Mellon.

Voices in AI – Episode 90: A Conversation with Norman Sadeh
By Byron Reese

Episode 90 of Voices in AI features Byron speaking with Norman Sadeh from Carnegie Mellon University about the nature of intelligence and how AI effects our privacy.

Listen to this episode or read the full transcript at www.VoicesinAI.com

Transcript Excerpt:

Byron Reese: This is Voices in AI brought to you by GigaOm I’m Byron Reese, today my guest is Norman Sadeh. He is a professor at Carnegie Mellon School of Computer Science. He’s affiliated with Cylab which is well known for their seminal work in AI planning and scheduling, and he is an authority on computer privacy. Welcome to the show.

Carnegie Mellon has this amazing reputation in the AI world. It’s arguably second to none. There are a few university campuses that seem to really… there’s Toronto and MIT, and in Carnegie Mellon’s case, how did AI become such a central focus?

Norman Sadeh: Well, this is one of the birthplaces of AI, and so the people who founded our computer science department included Herbert Simon and Allen Newell who are viewed as two of the four founders of AI. And so they contributed to the early research in that space. They helped frame many of the problems that people are still working on today, and they helped recruit also many more faculty over the years that have contributed to making Carnegie Mellon as the place that many people refer to as being the number one place in AI here in the US.

Not to say that there are not other many good places out there, but CMU is clearly a place where a lot of the leading research has been conducted over the years, whether you are looking at autonomous vehicles – for instance, I remember when I came here to do my PhD back in 1997, there was research going on autonomous vehicles. Obviously the vehicles were a lot clumsier than they are today, not moving quite as fast, but there’s a very, very long history of AI research, here at Carnegie Mellon. The same is true for language technology, the same is true for robotics, you name it. There are lots and lots of people here who are doing truly amazing things. ..... "