Robot Uses Fake Raspberry to Practice Picking Fruit

Robot Uses Fake Raspberry to Practice Picking Fruit

By Popular Science, June 26, 2023

Lab equipment with the raspberry mockup.

A growing number of farmers are interested in using robots for time-intensive tasks such as harvesting strawberries, sweet peppers, apples, lettuce, and tomatoes.

Credit: EPFL Create

Scientists at the Swiss Federal Institute of Technology, Lausanne (EPFL) Computational Robot Design & Fabrication Laboratory engineered a robot that practiced raspberry-picking on a silicone raspberry mockup with an artificial stem to ensure it learned how to handle the fragile fruit.

The researchers said the stem "can 'tell' the robot how much pressure is being applied, both while the fruit is still attached to the receptacle and after it's been released."

Experiments showed the robot could harvest 60% of the fruits while keeping them intact versus the human average of 90%.

An improved raspberry model could help enhance the robot's performance, while an extended setting that simulates "environmental conditions such as lighting, temperature, and humidity could further close the Lab2Field reality gap," according to the researchers.

From Popular Science

View Full Article   

Integrated Photonic Circuits Could Help Close the 'Terahertz Gap'

Towards improved chip Fabrication.

Integrated Photonic Circuits Could Help Close the 'Terahertz Gap'

EPFL News (Switzerland),Celia Luterbacher, December 1, 2023

Scientists at the Swiss Federal Institute of Technology, Lausanne (EPFL), ETH Zurich, and Harvard University created a thin-film circuit that generates custom-tailored terahertz-frequency waves in conjunction with a laser beam. This achievement could help bridge the so-called terahertz (THz) gap situated between approximately 300 gigahertz (GHz) and 30,000 GHz (30 THz) on the electromagnetic spectrum. The chip features an integrated photonic circuit fabricated from lithium niobate. Said EPFL's Cristina Benea-Chelmus, "The fact that our device already makes use of a standard optical signal is really an advantage, because it means that these new chips can be used with traditional lasers, which work very well and are very well understood. It means our device is telecommunications-compatible." ... ' 

Catheter-shaped Steerable Robot robotic devices

 Catheter-shaped robotic devices could be guided remotely within the body, or exercise semi-autonomous control. Steerable Soft Robots Could Enhance Medical Applications

EPFL (Switzerland), November 28, 2022

Scientists at the Swiss Federal Institute of Technology, Lausanne (EPFL) and the U.K.'s Imperial College London have produced catheter-shaped soft robots with sophisticated motion control. When used as catheters, the fiber-based robots could be guided remotely or semi-autonomously to a specific destination within a patient's body. The researchers used thermal drawing to generate fibers from thermoplastic elastomers, with multiple micrometer-scale channels grooved along their length to accommodate mechanisms like tendons. "In addition to the tendons, the fibers can integrate optical guides, electrodes, and microchannels that enable drug delivery, imaging, electrical recording and stimulation, and other tools commonly used in robotics and medical applications," explained EPFL's Andreas Leber. ... ' 

Edible Drones

Narrow and novel  application here.

ACM NEWS

Stranded Without Food? Edible Drone Has Snackable Wings  By CNET, November 17, 2022

The wings of this drone are nutritious and, depending on what you think about rice cakes, delicious.

The drone's design uses some familiar-looking airplane-like components; the big difference is that its fixed wing is made from rice cakes and gelatin.

It's a nightmare scenario. You're on an ambitious mountain hike when you get lost, injured or stranded. The good news is help is finally on the way, but the bad news is it's going to take time to reach you and you're out of food. That's when a buzzing drone comes flying in for a landing. Not only do you get the snacks, medicine or water it's carrying, you can eat the wings to tide you over until the rescue team arrives.

This scenario could become real. A team with the Swiss Federal Institute of Technology, Lausanne (EPFL) has developed a prototype edible drone. The munchable machine is part of a broader project called RoboFood. RoboFood is about investigating edible robots for humans and animals as well as foods that behave like robots.

The team published its work online with the title "Towards edible drones for rescue missions: design and flight of nutritional wings." The study is tackling the problem of getting commercial drones to carry enough of a payload to help people in emergency situations.

  Full Article.   

Understanding of Cellular Metabolism

Home/News/Better Understanding of Cellular Metabolism With AI/Full Text

ACM TECHNEWS

Better Understanding of Cellular Metabolism With AI   By EPFL

October 6, 2022  Scientists at EPFL, the Swiss Federal Institute of Technology, Lausanne, have developed REKINDLE, a deep learning-based computational framework that replicates dynamic metabolism in cells.

"REKINDLE will allow the research community to reduce computational efforts in generating kinetic models by several orders of magnitude," says EPFL's Ljubisa Miskovic. "It will also help in postulating new hypotheses by integrating biochemical data in these models, elucidating experimental observations, and steering new therapeutic discoveries and biotechnology designs."

Researchers envision the framework optimizing the metabolic network of microbes to generate industrial-scale chemical compounds, as well as unifying the use of kinetic modeling in the scientific community. To that end, EPFL's Subham Choudhury said REKINDLE uses popular Python libraries to promote accessibility and ease of use. ... 

An EPFL framework that leverages deep learning paves the way for the efficient and accurate modeling of metabolic processes. ... 

full Article

AI for Handwriting Skills

 Always looking for things that will look for patterns that need to be selectively taught.  Here a good example, teaching children to improve handwriting.  Other examples?  Coding?  AI Management?

Innovative application helps students learn to write

EPFL startup School Rebound has developed a revolutionary application that uses artificial intelligence to help students improve their handwriting in a fun and personalized way. Nearly 15,000 students in Switzerland, France, and Italy are already putting it to work.

Handwriting problems affect nearly 25% of children aged 5 to 12. These problems, if not managed early on, can negatively impact them throughout their school years. EPFL startup School Rebound has provided a concrete solution to this problem by developing an application that uses tablets and artificial intelligence to better detect potential handwriting problems and support children as they learn. The subscription app is called Dynamilis, and can be used by all children learning to write – with difficulties or without – at home or at school. Dynamilis has already been downloaded more than 10,000 times and will be released soon in England and the United States.

While Dynamilis was free during its testing phase, it switched to a subscription model this past March. After a free trial week, parents, therapists, and schools are offered a monthly or annual subscription. Costs depend on the number of children using the application.

L. Boatto, A. Peguet, T. Asselborn, S. Viquerat, P. Dillenbourg © 2022 Dynamilis / Sven Viquerat

Strengths and weaknesses at a glance

School Rebound was founded in 2021 based on the research done by CEO Thibault Asselborn for his PhD thesis at EPFL’s Computer-Human Interaction in Learning and Instruction Laboratory (CHILI). “Some children have handwriting problems that may stay hidden for months. Because the problems are not obvious, parents and teachers may hesitate to consult specialists,” says Asselborn. “During this time, the students can accumulate learning difficulties as these handwriting problems monopolize their concentration and prevent them from developing other skills. This could cause students to lose confidence and develop significant educational blocks.”

As part of his PhD research, Asselborn helped develop an algorithm that can rapidly analyze a child’s handwriting. The child only has to write for 30 seconds on an iPad with an Apple Pencil for the application to establish their handwriting profile. “The tests we ran in the lab lasted about 20 minutes, but they didn’t take certain factors into account, which may reduce the accuracy of the analyses,” says Asselborn. Dynamilis evaluates dynamic aspects of handwriting that the human eye can’t see such as stability, pressure, speed, and angle. “Our app gives detailed analyses about the motor aspects of handwriting. This can give parents an initial indication before determining whether their child has difficulties with handwriting and, if so, to what degree. If the difficulties surpass a certain level, they’re recommended to go see a specialist.”

Learning dressed up like a game

Based on a child’s handwriting profile, the app recommends personalized activities to practice the fundamental aspects of handwriting – all while playing games. “The in-app tests don’t look like medical tests in the strictest sense,” says Asselborn. “It’s important to have an air of fun to avoid making children feel like they’re taking an exam, which can put them on edge.” The playful aspect during the testing phase is crucial for the School Rebound team and their Chairman, Pierre Dillenbourg, also head of CHILI. “While we were developing Dynamilis, our aim was to help children,” says Dillenbourg. “To do that, we knew we had to go beyond a simple handwriting-analysis program and give them activities to support learning and, for the more severe cases, correction. Games are an effective solution for children who are having problems in school because of their issues with handwriting.”  ... ' 

Social Tracking Animals with Deep Learning

Swiss Federal Institute of Technology  EPFL   Deep Learning Tracking 

Time to get social: tracking animals with deep learning

Researchers at EPFL have made strides in computer-aided animal tracking by expanding their software, DeepLabCut, to offer high-performance tracking of multiple animals in videos.

The ability to capture the behavior of animals is critical for neuroscience, ecology, and many other fields. Cameras are ideal for capturing fine-grained behavior, but developing computer vision techniques to extract the animal’s behavior is challenging even though this seems effortless for our own visual system.

One of the key aspects of quantifying animal behavior is “pose estimation”, which refers to the ability of a computer to identify the pose (position and orientation of different body parts) of an animal. In a lab setting, it’s possible to assist pose estimation by placing markers on the animal’s body like in motion-capture techniques used in movies (think Gollum in the Lord of the Rings). But as one can imagine, getting animals to wear specialized equipment is not the easiest task, and downright impossible and unethical in the wild.

For this reason, Professors Alexander Mathis and Mackenzie Mathis at EPFL have been pioneering “markerless” tracking for animals. Their software relies on deep-learning to “teach” computers to perform pose estimation without the need for physical or virtual markers.

Their teams have been developing DeepLabCut, an open-source, deep-learning “animal pose estimation package” that can perform markerless motion capture of animals. In 2018 they released DeepLabCut, and the software has gained significant traction in life sciences: over 350,00 downloads of the software and nearly 1400 citations. Then, in 2020, the Mathis teams released DeepLabCut-Live!, a real-time low-latency version of DeepLabCut that allows researchers to rapidly give feedback to animals they are studying.

Now, the scientists have expanded DeepLabCut to address another challenge in pose estimation: tracking social animals, even closely interacting ones; e.g., parenting mice or schooling fish. The challenges here are obvious: the individual animals can be so similar looking that they confuse the computer, they can obscure each other, and there can be many “keypoints” that researchers wish to track, making it computationally difficult to process efficiently.

To tackle this challenge, they first created four datasets of varying difficulty for benchmarking multi-animal pose estimation networks. The datasets, collected with colleagues at MIT and Harvard University, consist of three mice in an open field, home-cage parenting in mice, pairs of marmosets housed in a large enclosure, and fourteen fish in a flow tank. With these datasets in hand, the researchers were able to develop novel methods to deal with the difficulties of real-world tracking. ..... 

A VR exploration of the Universe

With my background in astronomy I relish the thought.    An ideal application given all the data we have gathered. General availability? Will be looking for this.

Explore the Universe with VR

EPFL News (Switzerland)

Hillary Sanctuary, October 12, 2021

Researchers at Switzerland's École polytechnique fédérale de Lausanne (EPFL) used the latest astrophysical and cosmological data to create a virtual reality (VR) experience of outer space. The VIRUP (Virtual Reality Universe Project) open-source software allows users to navigate through a detailed map of the universe. It can visualize data from more than eight databases, including the Sloan Digital Sky Survey, comprised of more than 50 million galaxies and 300 million objects. Users need VR glasses, a computer to run the VIRUP engine, and sufficient storage space to take advantage of the fully immersive, three-dimensional experience. EPFL's Yves Revaz said, "VIRUP is precisely a way of making all of our astrophysical data accessible to everyone."  ... '

Swimming Robotic Locomotion

Have seen a few swimming robot examples, now a deeper take via EPFL.

Swimming robot gives fresh insight into locomotion and neuroscience  by Ecole Polytechnique Federale de Lausanne

Thanks to their swimming robot modeled after a lamprey, EPFL scientists may have discovered why some vertebrates are able to retain their locomotor capabilities after a spinal cord lesion. The finding could also help improve the performance of swimming robots used for search and rescue missions and for environmental monitoring.

Scientists at the Biorobotics Laboratory (BioRob) in EPFL's School of Engineering are developing innovative robots in order to study locomotion in animals and, ultimately, gain a better understanding of the neuroscience behind the generation of movement. One such robot is AgnathaX, a swimming robot employed in an international study with researchers from EPFL as well as Tohoku University in Japan, Institut Mines-Télécom Atlantique in Nantes, France, and Université de Sherbrooke in Canada. The study has just been published in Science Robotics.

"Our goal with this robot was to examine how the nervous system processes sensory information so as to produce a given kind of movement," says Prof. Auke Ijspeert, the head of BioRob. "This mechanism is hard to study in living organisms because the different components of the central and peripheral nervous systems are highly interconnected within the spinal cord. That makes it hard to understand their dynamics and the influence they have on each other."  .... ' 

Real-Time Marker-Less Motion Capture for Animals

Real time feedback for animal movement and posture. 

Real time Studies of animal Motion by Neural activity.By EPFL (Switzerland)

Nik Papageorgiou, December 10, 2020

An updated deep learning software toolbox developed at the Swiss Federal Institute of Technology, Lausanne (EPFL) facilitates real-time feedback studies on animal movement and posture. DeepLabCut-Live! (DLC-Live!) is designed to enable computers to track and predict these factors free of motion-capture markers, by controlling or stimulating the animals' neural activity. DLC-Live!'s tailored networks predict posture from video frames, combined with low latency so researchers can supply real-time feedback and assess behavioral functions of specific neural circuits; the system also interfaces with hardware used in posture studies to deliver feedback to animals. EPFL's Mackenzie Mathis said, "It's economical, it's scalable, and we hope it's a technical advance that allows even more questions to be asked about how the brain controls behavior."   .. '

High Quality Images of Moving Objects

I recall having to solve this problem for diagnosing from images of manufacturing machine parts.

Capturing Moving Subjects in Still-Life Quality
EPFL News (Switzerland)
June 18, 2020

Researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL) Advanced Quantum Architecture Laboratory and the University of Wisconsin-Madison (UW-Madison) Wision Laboratory have developed a method for capturing extremely clear images of moving subjects. UW-Madison's Mohit Gupta borrowed EPFL's SwissSPAD camera, which generates two-dimensional binary images at a resolution of 512 x 512 pixels. EPFL's Edoardo Charbon said SwissSPAD captures 100,000 binary images per second, as an algorithm corrects for variations; the researchers built a high-definition image of a moving subject by combining these photos. The team aims to repeat the experiment with the MegaX camera, which Charbon said "is similar to SwissSPAD in many ways; it's also a depth-sensing camera, thus it can generate [three-dimensional] images."   ... " 

Bioprinting the Very Small

Impressive capability, with many possible healthcare applications.

Printing Tiny, High-Precision Objects in Seconds
EPFL (Switzerland)
Sarah Perrin
February 13, 2020

Researchers at the Swiss Federal Institute of Technology in Lausanne (EPFL) have developed a high-precision technique for three-dimensionally (3D) printing small, soft objects in seconds. The method employs the principles of tomography, in which models of objects are constructed from surface scans. The printer transmits a laser through a translucent gel that is either organic or liquid plastic, hardening the material as algorithms calculate the areas the laser targets, the beam's angles, and intensity. The system currently produces 2cm structures with 80-micrometer precision, and new devices should be able to print larger objects, potentially up to 15 centimeters. The researchers partnered with a surgeon to test 3D-printed arteries fabricated with this method, and the technology could potentially have bioprinting applications due to its ability to print solid objects of different textures. ... "

Improving Neural Models

So many groups working on how to make such models smaller and faster.  Not different from the very long hunt for faster models to solve optimization problems over the years.   Discussion is technical.

A Tool to Simplify Complex Neuron Models
EPFL News (Switzerland)
January 15, 2020

Researchers at the Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland and Israel's Hebrew University of Jerusalem have developed a computational tool to streamline complex neuron models of any type of cell, while retaining their input/output properties and accelerating the run-times of cell simulations. The Neuron_Reduce tool maps a dendritic computation tree into a simpler multi-cylindrical tree, mapping synapses and ion channels into the reduced model to preserve their transfer impedance to the cell body. EPFL's Pramod Kumbhar said, "Neuron_Reduce ... opens the path for a novel type of reduced models that crucially maintain important details of the model but possibly run 40 to 250 times faster."

Predicting Lightning: Where/When

Skeptical.   If we are talking about predicting the likelihood of lighting in an area an hour before it hits, thats done already in weather predictive maps.    Suggests here that the accuracy is within a 30 km radius.   Specifics of goal are slightly incomplete,  but perhaps that is not being done that accurately yet.  "When it may strike", versus when/where?   AI aspect may well using machine learning in an area that might attract lightning such as tall trees or buildings

Using AI and machine learning to forecast lightning  By Techcrunchx 

As one of the most irregular phenomena in nature, lightning is very disturbing. Scientists have lately made an AI system that forecasts lightning up to 30 minutes before it strikes.

Lightning regularly kills animals and people, initiates fires, destroys power lines and keeps aircraft stranded. Till now, it has been almost out of the question to predict lightning, with no simple technology for predicting where and when it will strike the earth.

Engineers at the Ecole Polytechnique Federale de Lausanne’s (EPFL) School of Engineering built a simple and cheap system to forecast when lightning will strike. Farhad Rachidi led the research, which resulted in a technique of predicting lightning between 10 and 30 minutes before it hits, inside a 30km radius.  ....  "

Epidemics and use of Personal GPS Data

Recall our work with bioterrorism modeling, link below.  Not just a matter of disease epidemics.

During Epidemics, Access to GPS Data from Smartphones Can Be Crucial
Ecole Polytechnique Fédérale de Lausanne (Switzerland)
By Sandrine Perroud

Researchers at Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland and the Massachusetts Institute of Technology have found that human mobility is a major factor in the spread of vector-borne diseases such as malaria and dengue. The researchers used mobile phone data and census models to effectively predict the spatial distribution of dengue cases in Singapore, based on data from actual reported cases in 2013 and 2014. The team also demonstrated that the types of data used in their study could be obtained without infringing on people's privacy. Said EPFL's Emanuele Massaro, "We need to think seriously about changing the law around accessing this kind of information – not just for scientific research, but for wider prevention and public health reasons." ... '

Insect Scale Robotics

And is a bit frightening too.

The Deep Learning Way to Design Fly-Like Robots
EPFL (Switzerland)
October 11, 2019

Researchers at Switzerland's Swiss Federal Institute of Technology, Lausanne (EPFL) have developed motion-capture software that uses deep learning to model a fruit fly's movements in three dimensions (3D), in the hope of designing fly-like robots. The DeepFly3D system uses multiple cameras to record images of a fly crawling atop a small floating ball, which the software processes. DeepFly3D can infer the insect's pose in three dimensions, predicting and calculating behavioral measurements at ultra-fine resolution without the need for manual calibration. The system also employs active learning to improve its own performance. EPFL's Pavan Ramdya said, "If we learn how [the fly] does what it does, we can have important impact on robotics and medicine and, perhaps most importantly, we can gain these insights in a relatively short period of time." ..... 

Artificial Skin and Augmented Reality

Links between skin and augmented reality of interest

Artificial Skin Could Help Rehabilitation, Enhance Virtual Reality
Ecole Polytechnique Fédérale de Lausanne
Laure-Anne Pessina

Researchers at Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland have developed a soft, flexible artificial skin, made of silicone and electrodes and equipped with soft sensors and actuators, which provides haptic feedback in the form of pressure and vibration. Strain sensors continuously measure the skin's deformation so that the haptic feedback can be adjusted to produce a realistic sense of touch. The actuators help form a membrane layer and can be tuned to varying pressures and frequencies. On top of the membrane layer sits a sensor that contains soft electrodes that measure the skin's deformation continuously and send the data to a microcontroller, which uses the feedback to fine-tune the sensation transmitted to the user. For now, the scientists have tested the technology on users' fingers. The next step, said EPFL's Harshal Sonar, "will be to develop a fully wearable prototype for applications in rehabilitation and virtual and augmented reality."

Robo-Ants from Switzerland

Yet more smaller robotics, with claim of some ability to work together.   Note in particular the integration of sensor and autonomous capabilities..

Robot Ants Robot-Ants Can Jump, Communicate, and Work Together
Ecole Polytechnique Fédérale de Lausanne
By Laure-Anne Pessina

Researchers at Ecole Polytechnique Fédérale de Lausanne (EPFL) in Switzerland have developed robots inspired by ants that can communicate with each other, assign roles among themselves, and complete complex tasks together. The robots can jump and crawl to explore uneven surfaces, and can quickly detect and overcome obstacles much larger and heavier than themselves. The Tribots are completely autonomous and untethered, and are equipped with infrared and proximity sensors for detection and communications purposes. Said EPFL researcher Jamie Paik, "With their unique collective intelligence, our tiny robots can demonstrate better adaptability to unknown environments; therefore, for certain missions, they would outperform larger, more powerful robots." ... ' 

Talk on Advances in Image Recognition

I note this is an advanced technical talk on elements of image recognition ...

Invitation to the ISSIP Cognitive Systems Institute Group Webinar

Full series list, past and present  recordings are here:  http://cognitive-science.info/community/weekly-update/

Date and Time: September 06, 2018 - 10:30am US Eastern
Talk Title: Learning to Find Good Correspondences

Speaker: Eduard Trulls, EPFL

Talk Description:  

In this talk, Eduart will present a novel deep architecture to learn to find good correspondences for wide-baseline stereo. Our solution is based on putative keypointmatches, which we learn to label as inliers or outliers while simultaneously using them to recover the camera pose, a fundamental Computer Vision problem. Our solution is simple (no convolutional or fully-connected layers), small (4 Mb), easy to train (state of the art matching outdoors scenes with only 59 images) and generalizes well, particularly in contrast to dense networks which require the entire image.

Bio: 
Eduard Trulls is currently a post-doc at the Computer Vision Lab at EPFL in Lausanne, Switzerland. He obtained his PhD from the Institute of Robotics in Barcelona, Spain, in 2015. His thesis explored novel strategies to enhance local, low-level features (e.g. SIFT, HOG) with global, mid-level data such as motion or segmentation cues. His current work focuses on designing novel approaches to apply deep learning techniques to classical, low-level computer vision problems such as local feature extraction and matching for 3D reconstruction.  ... 

Date and Time : September 06 2018 - 10:30am US Eastern
Zoom meeting Link: https://zoom.us/j/7371462221
Zoom Callin: (415) 762-9988 or (646) 568-7788 Meeting id 7371462221
Zoom International Numbers: https://zoom.us/zoomconference
(Check the website in case the date or time changes: http://cognitive-science.info/community/weekly-update/ )

Please retweet  - https://twitter.com/sumalaika/status/1036933287640489984

Join LinkedIn Group https://www.linkedin.com/groups/6729452

Blurring the Lines Between Virtual and Reality

In EPFL News.  Could be a way to transfer gaming, and thus gaming engagement to real spaces.  Say offices, but perhaps also to industrial spaces. Possible then also for use with data, or analytic  solutions to data.  We examined some possibilities related to this before tech existed.    Note also the real-time aspects of such interactions.
 
Blurring the Lines Between Virtual and Reality 

Swiss Federal Institute of Technology in Lausanne

A next-generation virtual reality (VR) headset created by Hugo Hueber at the Swiss Federal Institute of Technology in Lausanne, Switzerland (EPFL) enables wearers to manipulate both real and virtual objects with tactile sensations, using hand avatars that replicate even the slightest movements. Hueber says an enhanced VR video game he is designing "combines the latest technology with the [three-dimensional] interactive research we're carrying out at the lab. That will let video gamers interact physically with a virtual environment that can be transferred to any location—a living room, office, or even classroom—instantly." Real-world objects are modeled and calibrated in the game, and then players can use them at the same time they physically touch them. In addition, players can see and precisely move virtual versions of fingers, and wear bodily sensors to see themselves move within the game in real time.  ... "