Category Archives: Human Robots
#440827 Location key to improved autonomous ...
QUT robotics researchers working with Ford Motor Company have found a way to tell an autonomous vehicle which cameras to use when navigating. Continue reading
Posted in Human Robots
#440824 This Robot Dog Has an AI Brain and ...
Ever seen a baby gazelle learn to walk? A fawn, which is basically a mammalian daddy long-legs, scrambles to its feet, falls, stands, and falls again. Eventually, it stands long enough to flail its toothpick-like legs into a series of near falls…ahem, steps. Amazingly, a few minutes after this endearing display, the fawn is hopping around like an old pro.
Well, now we have a robot version of this classic Serengeti scene.
The fawn in this case is a robotic dog at the University of California, Berkeley. And it’s likewise a surprisingly quick learner (relative to the rest of robot-kind). The robot is also special because, unlike other flashier robots you might have seen online, it uses artificial intelligence to teach itself how to walk.
Beginning on its back, legs waving, the robot learns to flip itself over, stand up, and walk in an hour. A further ten minutes of harassment with a roll of cardboard is enough to teach it how to withstand and recover from being pushed around by its handlers.
It’s not the first time a robot has used artificial intelligence to learn to walk. But while prior robots learned the skill by trial and error over innumerable iterations in simulations, the Berkeley bot learned entirely in the real world.
In a paper published on the arXiv preprint server, the researchers—Danijar Hafner, Alejandro Escontrela, and Philipp Wu—say transferring algorithms that have learned in simulation to the real world isn’t straightforward. Little details and differences between the real world and simulation can trip up fledgling robots. On the other hand, training algorithms in the real world is impractical: It’d take too much time and wear and tear.
Four years ago, for example, OpenAI showed off an AI-enabled robotic hand that could manipulate a cube. The control algorithm, Dactyl, needed some 100 years’ worth of experience in a simulation powered by 6,144 CPUs and 8 Nvidia V100 GPUs to accomplish this relatively simple task. Things have advanced since then, but the problem largely remains. Pure reinforcement learning algorithms need too much trial and error to learn skills for them to train in the real world. Simply put, the learning process would break researchers and robots before making any meaningful progress.
The Berkeley team set out to solve this problem with an algorithm called Dreamer. Constructing what’s called a “world model,” Dreamer can project the probability a future action will achieve its goal. With experience, the accuracy of its projections improve. By filtering out less successful actions in advance, the world model allows the robot to more efficiently figure out what works.
“Learning world models from past experience enables robots to imagine the future outcomes of potential actions, reducing the amount of trial and error in the real environment needed to learn successful behaviors,” the researchers write. “By predicting future outcomes, world models allow for planning and behavior learning given only small amounts of real world interaction.”
In other words, a world model can reduce the equivalent of years of training time in a simulation to no more than an awkward hour in the real world.
The approach may have wider relevance than robot dogs too. The team also applied Dreamer to a pick-and-place robotic arm and a wheeled robot. In both cases, they found Dreamer allowed their robots to efficiently learn relevant skills, no sim time required. More ambitious future applications might include self-driving cars.
Of course, there are still challenges to address. Although reinforcement learning automates some of the intricate hand-coding behind today’s most advanced robots, it does still require engineers to define a robot’s goals and what constitutes success—an exercise that is both time consuming and open-ended for real-world environments. Also, though the robot survived the team’s experiments here, longer training on more advanced skills may prove too much for future bots to survive without damage. The researchers say it might be fruitful to combine simulator training with fast real-world learning.
Still, the results advance AI in robotics another step. Dreamer strengthens the case that “reinforcement learning will be a cornerstone tool in the future of robot control,” Jonathan Hurst, a professor of robotics at Oregon State University told MIT Technology Review.
Image Credit: Danijar Hafner / YouTube Continue reading
Posted in Human Robots
#440822 Video Friday: Build a Chair
Video Friday is your weekly selection of awesome robotics videos, collected by your friends at IEEE Spectrum robotics. We also post a weekly calendar of upcoming robotics events for the next few months. Please send us your events for inclusion.
IEEE CASE 2022: 20–24 August 2022, MEXICO CITYCLAWAR 2022: 12–14 September 2022, AZORES, PORTUGALANA Avatar XPRIZE Finals: 4–5 November 2022, LOS ANGELESCoRL 2022: 14–18 December 2022, AUCKLAND, NEW ZEALANDEnjoy today's videos!
This probably counts as hard mode for Ikea chair assembly.
[ Naver Lab ]
As anyone working with robotics knows, it’s mandatory to spend at least 10 percent of your time just mucking about with them because it’s fun, as GITAI illustrates with its new 10-meter robotic arm.
[ GITAI ]
Well, this is probably the weirdest example of domain randomization in simulation for quadrupeds that I’ve ever seen.
[ RSL ]
The RoboCup 2022 was held in Bangkok, Thailand. The final match was between B-Human from Bremen (jerseys in black) and HTWK Robots from Leipzig (jerseys in blue). The video starts with one of our defending robots starting a duel with the opponent. After a short time a pass is made to another robot, which tries to score a goal, but the opponent goalie is able to catch the ball. Afterwards another attacker robot is already waiting at the center circle, to take its chance to score a goal, through all four opponent robots.
[ Team B-Human ]
The mission to return Martian samples back to Earth will see a European 2.5-meter-long robotic arm pick up tubes filled with precious soil from Mars and transfer them to a rocket for a historic interplanetary delivery.
[ ESA ]
I still cannot believe that this is an approach to robotic fruit-picking that actually works.
[ Tevel Aerobotics ]
This video shows the basic performance of the humanoid robot Torobo, which is used as a research platform for JST’s Moonshot R&D program.
[ Tokyo Robotics ]
Volocopter illustrates why I always carry two violins with me everywhere. You know, just in case.
[ Volocopter ]
We address the problem of enabling quadrupedal robots to perform precise shooting skills in the real world using reinforcement learning. Developing algorithms to enable a legged robot to shoot a soccer ball to a given target is a challenging problem that combines robot motion control and planning into one task.
[ Hybrid Robotics ]
I will always love watching Cassie try very, very hard to not fall over, and then fall over.
[ Michigan Robotics ]
I don’t think this paper is about teaching bipeds to walk with attitude, but it should be.
[ DLG ]
Modboats are capable of collective swimming in arbitrary configurations! In this video you can see three different configurations of the Modboats swim across our test space and demonstrate their capabilities.
[ ModLab ]
How have we built our autonomous driving technology to navigate the world safely? It comes down to three easy steps: Sense, Solve, and Go. Using a combination of lidar, camera, radar, and compute, the Waymo Driver can visualize the world, calculate what others may do, and proceed smoothly and safely, day and night.
[ Waymo ]
Alan Alda discusses evolutionary robotics with Hod Lipson and Jordan Pollack on Scientific American Frontiers in 1999.
[ Creative Machines Lab ]
Brady Watkins gives us insight into how a big company like Softbank Robotics looks into the robotics market.
[ Robohub ] Continue reading
Posted in Human Robots
#440820 Amazon to Acquire iRobot For $1.7 ...
This morning, Amazon and iRobot announced “a definitive merger agreement under which Amazon will acquire iRobot” for US $1.7 billion. The announcement was a surprise, to put it mildly, and we’ve barely had a chance to digest the news. But taking a look at what’s already known can still yield initial (if incomplete) answers as to why Amazon and iRobot want to team up—and whether the merger seems like a good idea.
The press release, like most press releases about acquisitions of this nature, doesn’t include much in the way of detail. But here are some quotes:
“We know that saving time matters, and chores take precious time that can be better spent doing something that customers love,” said Dave Limp, SVP of Amazon Devices. “Over many years, the iRobot team has proven its ability to reinvent how people clean with products that are incredibly practical and inventive—from cleaning when and where customers want while avoiding common obstacles in the home, to automatically emptying the collection bin. Customers love iRobot products—and I'm excited to work with the iRobot team to invent in ways that make customers' lives easier and more enjoyable.”“Since we started iRobot, our team has been on a mission to create innovative, practical products that make customers' lives easier, leading to inventions like the Roomba and iRobot OS,” said Colin Angle, chairman and CEO of iRobot. “Amazon shares our passion for building thoughtful innovations that empower people to do more at home, and I cannot think of a better place for our team to continue our mission. I’m hugely excited to be a part of Amazon and to see what we can build together for customers in the years ahead.”There’s not much to go on here, and iRobot has already referred us to Amazon PR, which, to be honest, feels like a bit of a punch in the gut. I love (loved?) so many things about iRobot—their quirky early history working on weird DARPA projects and even weirder toys, everything they accomplished with the PackBot (and also this), and most of all, the fact that they’ve made a successful company building useful and affordable robots for the home, which is just…it’s so hard to do that I don’t even know where to start. And nobody knows what’s going to happen to iRobot going forward. I’m sure iRobot and Amazon have all kinds of plans and promises and whatnot, but still—I’m now nervous about iRobot’s future.
Why this is a good move for Amazon is clear, but what exactly is in it for iRobot?
It seems fairly obvious why Amazon wanted to get its hands on iRobot. Amazon has been working for years to integrate itself into homes, first with audio systems (Alexa), and then video (Ring), and more recently some questionable home robots of its own, like its indoor security drone and Astro. Amazon clearly needs some help in understanding how to make home robots useful, and iRobot can likely provide some guidance, with its extraordinarily qualified team of highly experienced engineers. And needless to say, iRobot is already well established in a huge number of homes, with brand recognition comparable to something like Velcro or Xerox, in the sense that people don’t have “robot vacuums,” they have Roombas.
All those Roombas in all of those homes are also collecting a crazy amount of data for iRobot. iRobot itself has been reasonably privacy-sensitive about this, but it would be naïve not to assume that Amazon sees a lot of potential for learning much, much more about what goes on in our living rooms. This is more concerning, because Amazon has its own ideas about data privacy, and it’s unclear what this will mean for increasingly camera-reliant Roombas going forward.
I get why this is a good move for Amazon, but I must admit that I’m still trying to figure out what exactly is in it for iRobot, besides of course that “$61 per share in an all-cash transaction valued at approximately $1.7 billion.” Which, to be fair, seems like a heck of a lot of money. Usually when these kinds of mergers happen (and I’m thinking back to Google acquiring all those robotics companies in 2013), the hypothetical appeal for the robotics company is that suddenly they have a bunch more resources to spend on exciting new projects along with a big support structure to help them succeed.
It’s true that iRobot has apparently had some trouble with finding ways to innovate and grow, with their biggest potential new consumer product (the Terra lawn mower) having been on pause since 2020. It could be that big pile of cash, plus not having to worry so much about growth as a publicly traded company, plus some new Amazon-ish projects to work on could be reason enough for this acquisition.
My worry, though, is that iRobot is just going to get completely swallowed into Amazon and effectively cease to exist in a meaningful and unique way. I hope that the relationship between Amazon and iRobot will be an exception to this historical trend. Plus, there is some precedent for this—Boston Dynamics, for example, has survived multiple acquisitions while keeping its technology and philosophy more or less independent and intact. It’ll be on iRobot to very aggressively act to preserve itself, and keeping Colin Angle as CEO is a good start.
We’ll be trying to track down more folks to talk to about this over the coming weeks for a more nuanced and in-depth perspective. In the meantime, make sure to give your Roomba a hug—it’s been quite a day for little round robot vacuums. Continue reading
Posted in Human Robots