Tag Archives: change
#436180 Bipedal Robot Cassie Cal Learns to ...
There’s no particular reason why knowing how to juggle would be a useful skill for a robot. Despite this, robots are frequently taught how to juggle things. Blind robots can juggle, humanoid robots can juggle, and even drones can juggle. Why? Because juggling is hard, man! You have to think about a bunch of different things at once, and also do a bunch of different things at once, which this particular human at least finds to be overly stressful. While juggling may not stress robots out, it does require carefully coordinated sensing and computing and actuation, which means that it’s as good a task as any (and a more entertaining task than most) for testing the capabilities of your system.
UC Berkeley’s Cassie Cal robot, which consists of two legs and what could be called a torso if you were feeling charitable, has just learned to juggle by bouncing a ball on what would be her head if she had one of those. The idea is that if Cassie can juggle while balancing at the same time, she’ll be better able to do other things that require dynamic multitasking, too. And if that doesn’t work out, she’ll still be able to join the circus.
Cassie’s juggling is assisted by an external motion capture system that tracks the location of the ball, but otherwise everything is autonomous. Cassie is able to juggle the ball by leaning forwards and backwards, left and right, and moving up and down. She does this while maintaining her own balance, which is the whole point of this research—successfully executing two dynamic behaviors that may sometimes be at odds with one another. The end goal here is not to make a better juggling robot, but rather to explore dynamic multitasking, a skill that robots will need in order to be successful in human environments.
This work is from the Hybrid Robotics Lab at UC Berkeley, led by Koushil Sreenath, and is being done by Katherine Poggensee, Albert Li, Daniel Sotsaikich, Bike Zhang, and Prasanth Kotaru.
For a bit more detail, we spoke with Albert Li via email.
Image: UC Berkeley
UC Berkeley’s Cassie Cal getting ready to juggle.
IEEE Spectrum: What would be involved in getting Cassie to juggle without relying on motion capture?
Albert Li: Our motivation for starting off with motion capture was to first address the control challenge of juggling on a biped without worrying about implementing the perception. We actually do have a ball detector working on a camera, which would mean we wouldn’t have to rely on the motion capture system. However, we need to mount the camera in a way that it would provide the best upwards field of view, and we also have develop a reliable estimator. The estimator is particularly important because when the ball gets close enough to the camera, we actually can’t track the ball and have to assume our dynamic models describe its motion accurately enough until it bounces back up.
What keeps Cassie from juggling indefinitely?
There are a few factors that affect how long Cassie can sustain a juggle. While in simulation the paddle exhibits homogeneous properties like its stiffness and damping, in reality every surface has anisotropic contact properties. So, there are parts of the paddle which may be better for juggling than others (and importantly, react differently than modeled). These differences in contact are also exacerbated due to how the paddle is cantilevered when mounted on Cassie. When the ball hits these areas, it leads to a larger than expected error in a juggle. Due to the small size of the paddle, the ball may then just hit the paddle’s edge and end the juggling run. Over a very long run, this is a likely occurrence. Additionally, some large juggling errors could cause Cassie’s feet to slip slightly, which ends up changing the stable standing position over time. Since this version of the controller assumes Cassie is stationary, this change in position eventually leads to poor juggles and failure.
Would Cassie be able to juggle while walking (or hovershoe-ing)?
Walking (and hovershoe-ing) while juggling is a far more challenging problem and is certainly a goal for future research. Some of these challenges include getting the paddle to precise poses to juggle the ball while also moving to avoid any destabilizing effects of stepping incorrectly. The number of juggles per step of walking could also vary and make the mathematics of the problem more challenging. The controller goal is also more involved. While the current goal of the juggling controller is to juggle the ball to a static apex position, with a walking juggling controller, we may instead want to hit the ball forwards and also walk forwards to bounce it, juggle the ball along a particular path, etc. Solving such challenges would be the main thrusts of the follow-up research.
Can you give an example of a practical task that would be made possible by using a controller like this?
Studying juggling means studying contact behavior and leveraging our models of it to achieve a known objective. Juggling could also be used to study predictable post-contact flight behavior. Consider the scenario where a robot is attempting to make a catch, but fails, letting the ball to bounce off of its hand, and then recovering the catch. This behavior could also be intentional: It is often easier to first execute a bounce to direct the target and then perform a subsequent action. For example, volleyball players could in principle directly hit a spiked ball back, but almost always bump the ball back up and then return it.
Even beyond this motivating example, the kinds of models we employ to get juggling working are more generally applicable to any task that involves contact, which could include tasks besides bouncing like sliding and rolling. For example, clearing space on a desk by pushing objects to the side may be preferable than individually manipulating each and every object on it.
You mention collaborative juggling or juggling multiple balls—is that something you’ve tried yet? Can you talk a bit more about what you’re working on next?
We haven’t yet started working on collaborative or multi-ball juggling, but that’s also a goal for future work. Juggling multiple balls statically is probably the most reasonable next goal, but presents additional challenges. For instance, you have to encode a notion of juggling urgency (if the second ball isn’t hit hard enough, you have less time to get the first ball up before you get back to the second one).
On the other hand, collaborative human-robot juggling requires a more advanced decision-making framework. To get robust multi-agent juggling, the robot will need to employ some sort of probabilistic model of the expected human behavior (are they likely to move somewhere? Are they trying to catch the ball high or low? Is it safe to hit the ball back?). In general, developing such human models is difficult since humans are fairly unpredictable and often don’t exhibit rational behavior. This will be a focus of future work.
[ Hybrid Robotics Lab ] Continue reading
Posted in Human Robots
#436176 We’re Making Progress in Explainable ...
Machine learning algorithms are starting to exceed human performance in many narrow and specific domains, such as image recognition and certain types of medical diagnoses. They’re also rapidly improving in more complex domains such as generating eerily human-like text. We increasingly rely on machine learning algorithms to make decisions on a wide range of topics, from what we collectively spend billions of hours watching to who gets the job.
But machine learning algorithms cannot explain the decisions they make.
How can we justify putting these systems in charge of decisions that affect people’s lives if we don’t understand how they’re arriving at those decisions?
This desire to get more than raw numbers from machine learning algorithms has led to a renewed focus on explainable AI: algorithms that can make a decision or take an action, and tell you the reasons behind it.
What Makes You Say That?
In some circumstances, you can see a road to explainable AI already. Take OpenAI’s GTP-2 model, or IBM’s Project Debater. Both of these generate text based on a large corpus of training data, and try to make it as relevant as possible to the prompt that’s given. If these models were also able to provide a quick run-down of the top few sources in that corpus of training data they were drawing information from, it may be easier to understand where the “argument” (or poetic essay about unicorns) was coming from.
This is similar to the approach Google is now looking at for its image classifiers. Many algorithms are more sensitive to textures and the relationship between adjacent pixels in an image, rather than recognizing objects by their outlines as humans do. This leads to strange results: some algorithms can happily identify a totally scrambled image of a polar bear, but not a polar bear silhouette.
Previous attempts to make image classifiers explainable relied on significance mapping. In this method, the algorithm would highlight the areas of the image that contributed the most statistical weight to making the decision. This is usually determined by changing groups of pixels in the image and seeing which contribute to the biggest change in the algorithm’s impression of what the image is. For example, if the algorithm is trying to recognize a stop sign, changing the background is unlikely to be as important as changing the sign.
Google’s new approach changes the way that its algorithm recognizes objects, by examining them at several different resolutions and searching for matches to different “sub-objects” within the main object. You or I might recognize an ambulance from its flashing lights, its tires, and its logo; we might zoom in on the basketball held by an NBA player to deduce their occupation, and so on. By linking the overall categorization of an image to these “concepts,” the algorithm can explain its decision: I categorized this as a cat because of its tail and whiskers.
Even in this experiment, though, the “psychology” of the algorithm in decision-making is counter-intuitive. For example, in the basketball case, the most important factor in making the decision was actually the player’s jerseys rather than the basketball.
Can You Explain What You Don’t Understand?
While it may seem trivial, the conflict here is a fundamental one in approaches to artificial intelligence. Namely, how far can you get with mere statistical associations between huge sets of data, and how much do you need to introduce abstract concepts for real intelligence to arise?
At one end of the spectrum, Good Old-Fashioned AI or GOFAI dreamed up machines that would be entirely based on symbolic logic. The machine would be hard-coded with the concept of a dog, a flower, cars, and so forth, alongside all of the symbolic “rules” which we internalize, allowing us to distinguish between dogs, flowers, and cars. (You can imagine a similar approach to a conversational AI would teach it words and strict grammatical structures from the top down, rather than “learning” languages from statistical associations between letters and words in training data, as GPT-2 broadly does.)
Such a system would be able to explain itself, because it would deal in high-level, human-understandable concepts. The equation is closer to: “ball” + “stitches” + “white” = “baseball”, rather than a set of millions of numbers linking various pathways together. There are elements of GOFAI in Google’s new approach to explaining its image recognition: the new algorithm can recognize objects based on the sub-objects they contain. To do this, it requires at least a rudimentary understanding of what those sub-objects look like, and the rules that link objects to sub-objects, such as “cats have whiskers.”
The issue, of course, is the—maybe impossible—labor-intensive task of defining all these symbolic concepts and every conceivable rule that could possibly link them together by hand. The difficulty of creating systems like this, which could handle the “combinatorial explosion” present in reality, helped to lead to the first AI winter.
Meanwhile, neural networks rely on training themselves on vast sets of data. Without the “labeling” of supervised learning, this process might bear no relation to any concepts a human could understand (and therefore be utterly inexplicable).
Somewhere between these two, hope explainable AI enthusiasts, is a happy medium that can crunch colossal amounts of data, giving us all of the benefits that recent, neural-network AI has bestowed, while showing its working in terms that humans can understand.
Image Credit: Image by Seanbatty from Pixabay Continue reading
Posted in Human Robots
#436042 Video Friday: Caltech’s Drone With ...
Video Friday is your weekly selection of awesome robotics videos, collected by your Automaton bloggers. We’ll also be posting a weekly calendar of upcoming robotics events for the next few months; here’s what we have so far (send us your events!):
ISRR 2019 – October 6-10, 2019 – Hanoi, Vietnam
Ro-Man 2019 – October 14-18, 2019 – New Delhi, India
Humanoids 2019 – October 15-17, 2019 – Toronto, Canada
ARSO 2019 – October 31-1, 2019 – Beijing, China
ROSCon 2019 – October 31-1, 2019 – Macau
IROS 2019 – November 4-8, 2019 – Macau
Let us know if you have suggestions for next week, and enjoy today’s videos.
Caltech has been making progress on LEONARDO (LEg ON Aerial Robotic DrOne), their leggy thruster powered humanoid-thing. It can now balance and walk, which is quite impressive to see.
We’ll circle back again when they’ve got it jumping and floating around.
[ Caltech ]
Turn the subtitles on to learn how robots became experts at slicing bubbly, melty, delicious cheese.
These robots learned how to do the traditional Swiss raclette from demonstration. The Robot Learning & Interaction group at the Idiap Research Institute has developed an imitation learning technique allowing the robot to acquire new skills by considering position and force information, with an automatic adaptation to new situations. The range of applications is wide, including industrial robots, service robots, and assistive robots.
[ Idiap ]
Thanks Sylvain!
Some amazing news this week from Skydio, with the announcement of their better in every single way Skydio 2 autonomous drone. Read our full article for details, but here’s a getting started video that gives you an overview of what the drone can do.
The first batch sold out in 36 hours, but you can put down a $100 deposit to reserve the $999 drone for 2020 delivery.
[ Skydio ]
UBTECH is introducing a couple new robot kits for the holidays: ChampBot and FireBot.
$130 each, available on October 20.
[ Ubtech ]
NASA’s InSight lander on Mars is trying to use its robotic arm to get the mission’s heat flow probe, or mole, digging again. InSight team engineer Ashitey Trebbi-Ollennu, based at NASA’s Jet Propulsion Laboratory in Pasadena, California, explains what has been attempted and the game plan for the coming weeks. The next tactic they’ll try will be “pinning” the mole against the hole it’s in.
[ NASA ]
We introduce shape-changing swarm robots. A swarm of self-transformable robots can both individually and collectively change their configuration to display information, actuate objects, act as tangible controllers, visualize data, and provide physical affordances. ShapeBots is a concept prototype of shape-changing swarm robots. Each robot can change its shape by leveraging small linear actuators that are thin (2.5 cm) and highly extendable (up to 20cm) in both horizontal and vertical directions.
[ Ryo Suzuki ]
Robot abuse!
Vision 60 legged robot managing unstructured terrain without vision or force sensors in its legs. Using only high-transparency actuators and 2kHz algorithmic stability control… 4-limbs and 12-motors with only a velocity command.
[ Ghost Robotics ]
We asked real people to bring in real products they needed picked for their application. In MINUTES, we assembled the right tool.
This is a cool idea, but for a real challenge they should try it outside a supermarket. Or a pet store.
[ Soft Robotics ]
Good water quality is important to humans and to nature. In a country with as much water as the Netherlands has, ensuring water quality is a very labour-intensive undertaking. To address this issue, researchers from TU Delft have developed a ‘pelican drone’: a drone capable of taking water samples quickly, in combination with a measuring instrument that immediately analyses the water quality. The drone was tested this week at the new Marker Wadden nature area ‘Living Lab’.
[ MAVLab ]
In an international collaboration led by scientists in Switzerland, three amputees merge with their bionic prosthetic legs as they climb over various obstacles without having to look. The amputees report using and feeling their bionic leg as part of their own body, thanks to sensory feedback from the prosthetic leg that is delivered to nerves in the leg’s stump.
[ EPFL ]
It’s a little hard to see, but this is one way of testing out asteroid imaging spacecraft without actually going into space: a fake asteroid and a 2D microgravity simulator.
[ Caltech ]
Drones can help filmmakers do the kinds of shots that would be otherwise impossible.
[ DJI ]
Two long interviews this week from Lex Fridman’s AI Podcast, and both of them are worth watching: Gary Marcus, and Peter Norvig.
[ AI Podcast ]
This week’s CMU RI Seminar comes from Tucker Hermans at the University of Utah, on “Improving Multi-fingered Robot Manipulation by Unifying Learning and Planning.”
Multi-fingered hands offer autonomous robots increased dexterity, versatility, and stability over simple two-fingered grippers. Naturally, this increased ability comes with increased complexity in planning and executing manipulation actions. As such, I propose combining model-based planning with learned components to improve over purely data-driven or purely-model based approaches to manipulation. This talk examines multi-fingered autonomous manipulation when the robot has only partial knowledge of the object of interest. I will first present results on planning multi-fingered grasps for novel objects using a learned neural network. I will then present our approach to planning in-hand manipulation tasks when dynamic properties of objects are not known. I will conclude with a discussion of our ongoing and future research to further unify these two approaches.
[ CMU RI ] Continue reading
Posted in Human Robots
#436021 AI Faces Speed Bumps and Potholes on Its ...
Implementing machine learning in the real world isn’t easy. The tools are available and the road is well-marked—but the speed bumps are many.
That was the conclusion of panelists wrapping up a day of discussions at the IEEE AI Symposium 2019, held at Cisco’s San Jose, Calif., campus last week.
The toughest problem, says Ben Irving, senior manager of Cisco’s strategy innovations group, is people.
It’s tough to find data scientist expertise, he indicated, so companies are looking into non-traditional sources of personnel, like political science. “There are some untapped areas with a lot of untapped data science expertise,” Irving says.
Lazard’s artificial intelligence manager Trevor Mottl agreed that would-be data scientists don’t need formal training or experience to break into the field. “This field is changing really rapidly,” he says. “There are new language models coming out every month, and new tools, so [anyone should] expect to not know everything. Experiment, try out new tools and techniques, read, study, spend time; there aren’t any true experts at this point because the foundational elements are shifting so rapidly.”
“It is a wonderful time to get into a field,” he reasons, noting that it doesn’t take long to catch up because there aren’t 20 years of history.”
Confusion about what different kinds of machine learning specialists do doesn’t help the personnel situation. An audience member asked panelists to explain the difference between data scientist, data analyst, and data engineer. Darrin Johnson, Nvidia global director of technical marketing for enterprise, admitted it’s hard to sort out, and any two companies could define the positions differently. “Sometimes,” he says, particularly at smaller companies, “a data scientist plays all three roles. But as companies grow, there are different groups that ingest data, clean data, and use data. At some companies, training and inference are separate. It really depends, which is a challenge when you are trying to hire someone.”
Mitigating the risks of a hot job market
The competition to hire data scientists, analysts, engineers, or whatever companies call them requires that managers make sure any work being done is structured and comprehensible at all times, the panelists cautioned.
“We need to remember that our data scientists go home every day and sometimes they don’t come back because they go home and then go to a different company,” says Lazard’s Mottl. “That’s a fact of life. If you give people choice on [how they do development], and have a successful person who gets poached by competitor, you have to either hire a team to unwrap what that person built or jettison their work and rebuild it.”
By contrast, he says, “places that have structured coding and structured commits and organized constructions of software have done very well.”
But keeping all of a company’s engineers working with the same languages and on the same development paths is not easy to do in a field that moves as fast as machine learning. Zongjie Diao, Cisco director of product management for machine learning, quipped: “I have a data scientist friend who says the speed at which he changes girlfriends is less than speed at which he changes languages.”
The data scientist/IT manager clash
Once a company finds the data engineers and scientists they need and get them started on the task of applying machine learning to that company’s operations, one of the first obstacles they face just might be the company’s IT department, the panelists suggested.
“IT is process oriented,” Mottl says. The IT team “knows how to keep data secure, to set up servers. But when you bring in a data science team, they want sandboxes, they want freedom, they want to explore and play.”
Also, Nvidia’s Johnson pointed out, “There is a language barrier.” The AI world, he says, is very different from networking or storage, and data scientists find it hard to articulate their requirements to IT.
On the ground or in the cloud?
And then there is the decision of where exactly machine learning should happen—on site, or in the cloud? At Lazard, Mottl says, the deep learning engineers do their experimentation on premises; that’s their sandbox. “But when we deploy, we deploy in the cloud,” he says.
Nvidia, Johnson says, thinks the opposite approach is better. We see the cloud as “the sandbox,” he says. “So you can run as many experiments as possible, fail fast, and learn faster.”
For Cisco’s Irving, the “where” of machine learning depends on the confidentiality of the data.
Mottl, who says rolling machine learning technology into operation can hit resistance from all across the company, had one last word of caution for those aiming to implement AI:
Data scientists are building things that might change the ways other people in the organization work, like sales and even knowledge workers. [You need to] think about the internal stakeholders and prepare them, because the last thing you want to do is to create a valuable new thing that nobody likes and people take potshots against.
The AI Symposium was organized by the Silicon Valley chapters of the IEEE Young Professionals, the IEEE Consultants’ Network, and IEEE Women in Engineering and supported by Cisco. Continue reading
Posted in Human Robots