Connect with us

Gadgets

Robots learn to grab and scramble with new levels of agility – TechCrunch

Published

on

Robots are amazing things, but outside of their specific domains they are incredibly limited. So flexibility — not physical, but mental — is a constant area of research. A trio of new robotic setups demonstrate ways they can evolve to accommodate novel situations: using both “hands,” getting up after a fall, and understanding visual instructions they’ve never seen before.

The robots, all developed independently, are gathered together today in a special issue of the journal Science Robotics dedicated to learning. Each shows an interesting new way in which robots can improve their interactions with the real world.

On the other hand…

First there is the question of using the right tool for a job. As humans with multi-purpose grippers on the ends of our arms, we’re pretty experienced with this. We understand from a lifetime of touching stuff that we need to use this grip to pick this up, we need to use tools for that, this will be light, that heavy, and so on.

Robots, of course, have no inherent knowledge of this, which can make things difficult; it may not understand that it can’t pick up something of a given size, shape, or texture. A new system from Berkeley roboticists acts as a rudimentary decision-making process, classifying objects as able to be grabbed either by an ordinary pincer grip or with a suction cup grip.

A robot, wielding both simultaneously, decides on the fly (using depth-based imagery) what items to grab and with which tool; the result is extremely high reliability even on piles of objects it’s never seen before.

It’s done with a neural network that consumed millions of data points on items, arrangements, and attempts to grab them. If you attempted to pick up a teddy bear with a suction cup and it didn’t work the first ten thousand times, would you keep on trying? This system learned to make that kind of determination, and as you can imagine such a thing is potentially very important for tasks like warehouse picking for which robots are being groomed.

Interestingly, because of the “black box” nature of complex neural networks, it’s difficult to tell what exactly Dex-Net 4.0 is actually basing its choices on, although there are some obvious preferences, explained Berkeley’s  Ken Goldberg in an email.

“We can try to infer some intuition but the two networks are inscrutable in that we can’t extract understandable ‘policies,’ ” he wrote. “We empirically find that smooth planar surfaces away from edges generally score well on the suction model and pairs of antipodal points generally score well for the gripper.”

Now that reliability and versatility are high, the next step is speed; Goldberg said that the team is “working on an exciting new approach” to reduce computation time for the network, to be documented, no doubt, in a future paper.

ANYmal’s new tricks

Quadrupedal robots are already flexible in that they can handle all kinds of terrain confidently, even recovering from slips (and of course cruel kicks). But when they fall, they fall hard. And generally speaking they don’t get up.

The way these robots have their legs configured makes it difficult to do things in anything other than an upright position. But ANYmal, a robot developed by ETH Zurich (and which you may recall from its little trip to the sewer recently), has a more versatile setup that gives its legs extra degrees of freedom.

What could you do with that extra movement? All kinds of things. But it’s incredibly difficult to figure out the exact best way for the robot to move in order to maximize speed or stability. So why not use a simulation to test thousands of ANYmals trying different things at once, and use the results from that in the real world?

This simulation-based learning doesn’t always work, because it isn’t possible right now to accurately simulate all the physics involved. But it can produce extremely novel behaviors or streamline ones humans thought were already optimal.

At any rate that’s what the researchers did here, and not only did they arrive at a faster trot for the bot (above), but taught it an amazing new trick: getting up from a fall. Any fall. Watch this:

It’s extraordinary that the robot has come up with essentially a single technique to get on its feet from nearly any likely fall position, as long as it has room and the use of all its legs. Remember, people didn’t design this — the simulation and evolutionary algorithms came up with it by trying thousands of different behaviors over and over and keeping the ones that worked.

Ikea assembly is the killer app

Let’s say you were given three bowls, with red and green balls in the center one. Then you’re given this on a sheet of paper:

As a human with a brain, you take this paper for instructions, and you understand that the green and red circles represent balls of those colors, and that red ones need to go to the left, while green ones go to the right.

This is one of those things where humans apply vast amounts of knowledge and intuitive understanding without even realizing it. How did you choose to decide the circles represent the balls? Because of the shape? Then why don’t the arrows refer to “real” arrows? How do you know how far to go to the right or left? How do you know the paper even refers to these items at all? All questions you would resolve in a fraction of a second, and any of which might stump a robot.

Researchers have taken some baby steps towards being able to connect abstract representations like the above with the real world, a task that involves a significant amount of what amounts to a sort of machine creativity or imagination.

Making the connection between a green dot on a white background in a diagram and a greenish roundish thing on a black background in the real world isn’t obvious, but the “visual cognitive computer” created by Miguel Lázaro-Gredilla and his colleagues at Vicarious AI seems to be doing pretty well at it.

It’s still very primitive, of course, but in theory it’s the same toolset that one uses to, for example, assemble a piece of Ikea furniture: look at an abstract representation, connect it to real-world objects, then manipulate those objects according to the instructions. We’re years away from that, but it wasn’t long ago that we were years away from a robot getting up from a fall or deciding a suction cup or pincer would work better to pick something up.

The papers and videos demonstrating all the concepts above should be available at the Science Robotics site.

Source link

Continue Reading
Click to comment

Leave a Reply

Your email address will not be published.

Gadgets

iOS 15.5 and macOS 12.4 bring updates to Podcasts, digital payments, and more

Published

on

Enlarge / Apple’s Studio Display received a firmware update today to improve its webcam performance.

Andrew Cunningham

Apple released new software updates for all of its platforms on Tuesday. That includes the following:

  • iOS 15.5 for iPhones and the iPod touch
  • iPadOS 15.5 for iPads
  • macOS 12.4 for Macs
  • watchOS 8.6 for the Apple Watch
  • tvOS 15.5 for the Apple TV
  • HomePod Software 15.5 for HomePods
  • Studio Display Firmware 15.5 for the Studio Display
  • Swift Playgrounds 4.1 for iPad and Mac

These are almost certainly the last updates before the company’s annual developer conference, which is scheduled to kick off on June 6. Among other things, Apple will announce iOS and iPadOS 16, macOS 13, and watchOS 9 at the conference, but those updates won’t arrive until later this year.

iOS 15.5

Today’s iOS update offers just enough new user-facing features to earn that 15.x label instead of 15.x.x, which is usually reserved for bug fixes and the like.

All told, though, it’s a small update. The built-in Podcasts app gets “a new setting to limit episodes stored on your Mac and automatically delete older ones.”

And 15.5 allows the iPhone to be used as a point-of-sale device without any additional hardware, as reported in February.

Previously, vendors like farmer’s market stalls and home repair services used iPhones with attached add-on hardware from companies like Stripe to receive payments.

Now the iPhone doesn’t need those attachments; Stripe works just fine with an iPhone fresh out of the box.

Additionally, iOS 15.5 brings safety features in the Messages app meant to prevent children from being exposed to inappropriate content in the following new countries:

  • Australia
  • Canada
  • New Zealand
  • United Kingdom
  • United States

macOS 12.4

On the macOS side, Apple names just two user-facing changes in its release notes. It adds support for Studio Display Firmware Update 15.5 (which claims to improve webcam performance on Apple’s new monitor), and the built-in Podcasts app gets the same new feature that iOS did.

However, macOS 12.4 includes more than 50 security updates under the hood, according to Apple’s support documentation.

watchOS 8.6

watchOS 8.6 is a relatively minor update. It expands some of the Watch’s health features—namely irregular heart rhythm detection and the ECG—to Apple Watch users in Mexico.

Other updates

Studio Display Firmware 15.5 attempts to address some user complaints about the monitor’s webcam quality. Apple hasn’t shared any details about what’s in the HomePod firmware update or tvOS 15.5.

There’s also Swift Playgrounds 4.1 for Mac and iPad. It’s not an OS update, but it landed around the same time. It allows users to use Playgrounds to build apps with SwiftUI on the Mac, and it deepens App Store Connect integration for publishing apps, among other things.

Continue Reading

Gadgets

Google backtracks on legacy GSuite account shutdown, won’t take user emails

Published

on

Enlarge / An artist’s rendering of Google’s current reputation.

Google finally launched a solution for people with “legacy” GSuite Google accounts. After initially threatening to shut down free GSuite accounts if users didn’t start paying for the service, Google has completely backed off. Once users jump through some sign-up hoops, Google will allow their ~16-year-old accounts to continue functioning. You’ll even get to keep your email address.

The saga so far, if you haven’t been following, is that Google has a custom-domain user account service, currently called “Google Workspace” and previously called “G Suite” and “Google Apps.” The service is mostly a normal Google account that lets you use an email that ends in your custom domain name rather than “@gmail.com.” Today this service is aimed at businesses and costs money each month, but that was not always the case. From 2006 to 2012, custom domain Google accounts were free and were even pitched at families as a geeky way to have an online Google identity.

In January, some bean counter at Google apparently noticed this tiny group of longtime users was technically getting a paid service for free and decided this was unacceptable. Google posted an announcement in January declaring these people “Legacy GSuite users” and basically told them, ‘Pay up or lose your account.’ These users signed up for a free Google service and stored data on it for as long as 16 years, and there were no indications it would ever be charged. Google held this decade-plus of user data hostage, telling users to start paying business rates for Workspace or face an account shutdown.

A week later, after the inevitable public outcry, Google relented somewhat and said it would vaguely, eventually provide “an option for you to move your non-Google Workspace paid content and most of your data to a no-cost option.” Being told you’ll be able to keep “most of your data” that you’ve been accumulating for 16 years is a rather alarming statement. Google’s one bit of specifics in January was that “this new option won’t include premium features like custom email,” so you’d have to stop hosting your email with Google, and you’d presumably have to go through some wild Google account conversion process. It then let these users anxiously flap in the wind, with no further details, for six months.

How to save your free GSuite account

In May, Google finally told these users what would happen to their accounts. The new support page says, “For individuals and families using your account for non-commercial purposes, you can continue using the G Suite legacy free edition and opt out of the transition to Google Workspace.” The link for that is here or in your GSuite admin panel. You’ll need to confirm that your GSuite account is for personal use, and not business use, because businesses are still expected to pay for Workspace. If you already bent to Google’s will and started paying for Workspace because of the January announcement, Google says you should contact support.

That bottom
Enlarge / That bottom “Personal use” button is what you want.

Lee Hutchinson

The biggest news from this latest announcement is that Google has decided against taking people’s custom email away. A second support page says, “You can continue using your custom domain with Gmail, retain access to no-cost Google services such as Google Drive and Google Meet, and keep your purchases and data.” It now sounds like there will be no changes to your account, provided you click through the “self-transition” screen before the deadline.

The deadline to opt out of an account shutdown, which has changed several times now, is June 27, 2022. If you don’t complete this opt out by June 27, you will be automatically billed for Workspace. If you don’t have a card on file and don’t opt out, your account will be suspended on August 1 and shut down.

The automatic enrollment and billing, without explicit user consent, is one of the wilder parts of this story. If you don’t closely follow the tech news scene, there’s a good chance you won’t know this is coming, and you will either suddenly be billed without your consent or find that your Google account has suddenly stopped working.

For a company whose key business pillar is convincing users to store vast amounts of data, playing games like this is a bizarre decision. At least it came to a reasonable conclusion.

Continue Reading

Gadgets

Testing shows AMD’s FSR 2.0 can even help lowly Intel integrated GPUs

Published

on

Intel

There are two things to like about version 2.0 of AMD’s FidelityFX Super Resolution (FSR) upscaling tech, which finally began appearing in actual games late last week. The most important is that the quality of the upscaled image is dramatically better than in FSR version 1.0. The second is that FSR 2.0 is compatible with all kinds of GPUs, including not just AMD’s but older GeForce GPUs that aren’t compatible with Nvidia’s proprietary deep learning super sampling (DLSS).

New testing from Tom’s Hardware has also revealed another unlikely beneficiary: Intel’s recent integrated GPUs. Using an Iris Xe laptop GPU in a Core i7-1165G7, FSR 2.0 was able to bump the average frame rates in a 720p version of Deathloop by around 16%, nudging it from just under 30 fps to just over 30 fps and helping to offset the low resolution with its built-in anti-aliasing. Not bad for a nearly two-year-old laptop GPU playing a demanding modern game.

There are caveats, some of which apply to all upscaling technologies and some that are specific to Intel’s GPUs. FSR 2.0 and DLSS are generally good enough to let you bump up your resolution or quality settings a bit while maintaining a playable frame rate. They can also make borderline-unplayable games playable, and they can help you squeeze a little more life out of your current GPU if you don’t want (or can’t afford) to spring for an upgrade.

But upscaling also isn’t magical—the integrated GPU in a 10th-generation Intel Ice Lake CPU got nowhere near playable frame rates in Deathloop without FSR 2.0, and the low-double-digit performance improvement from FSR didn’t get it over that 30 fps line. Both Intel GPUs also showed lots of graphical corruption in most of the test runs, though this was inconsistent and could be fixed in future driver updates.

Wider, manufacturer-agnostic hardware compatibility could eventually help AMD accomplish with FSR what it did with FreeSync adaptive sync technology a few years ago. Nvidia’s G-Sync was technically superior, but it required more expensive monitors with an additional hardware module installed, and it only worked with Nvidia GPUs. FreeSync wasn’t as good initially, but it piggybacked on standard DisplayPort features that made it easier and cheaper to implement. A few years later, Nvidia enabled FreeSync support in its drivers, and today, FreeSync is by far the more prevalent of the two technologies.

Game developers could choose to support FSR 2.0 over Nvidia’s DLSS for the same reason: It provides good-enough results that cover a much broader range of GPU hardware from multiple manufacturers. AMD isn’t alone in trying to define a more widely compatible standard for high-quality upscaling, though—Intel’s upcoming XeSS standard can also be used with Intel, Nvidia, or AMD GPUs. DLSS support is also fairly entrenched, with relatively wide support across a long list of modern games.

Continue Reading

Trending