Tag Archives: robot
#439592 Robot Shows How Simple Swimming Can Be
Lots of robots use bioinspiration in their design. Humanoids, quadrupeds, snake robots—if an animal has figured out a clever way of doing something, odds are there's a robot that's tried to duplicate it. But animals are often just a little too clever for the robots that we build that try to mimic them, which is why researchers at
Swiss Federal Institute of Technology Lausanne in Switzerland (EPFL) are using robots to learn about how animals themselves do what they do. In a paper published today in Science Robotics, roboticists from EPFL's Biorobotics Laboratory introduce a robotic eel that leverages sensory feedback from the water it swims through to coordinate its motion without the need for central control, suggesting a path towards simpler, more robust mobile robots.
The robotic eel—called AgnathaX—is a descendant of
AmphiBot, which has been swimming around at EPFL for something like two decades. AmphiBot's elegant motion in the water has come from the equivalent what are called central pattern generators (CPGs), which are sequences of neural circuits (the biological kind) that generate the sort of rhythms that you see in eel-like animals that rely on oscillations to move. It's possible to replicate these biological circuits using newfangled electronic circuits and software, leading to the same kind of smooth (albeit robotic) motion in AmphiBot.
Biological researchers had pretty much decided that CPGs explained the extent of wiggly animal motion, until it was discovered you can chop an eel's spinal cord in half, and it'll somehow maintain its coordinated undulatory swimming performance. Which is kinda nuts, right? Obviously, something else must be going on, but trying to futz with eels to figure out exactly what it was isn't, I would guess, pleasant for either researchers or their test subjects, which is where the robots come in. We can't make robotic eels that are exactly like the real thing, but we can duplicate some of their sensing and control systems well enough to understand how they do what they do.
AgnathaX exhibits the same smooth motions as the original version of AmphiBot, but it does so without having to rely on centralized programming that would be the equivalent of a biological CPG. Instead, it uses skin sensors that can detect pressure changes in the water around it, a feature also found on actual eels. By hooking these pressure sensors up to AgnathaX's motorized segments, the robot can generate swimming motions even if its segments aren't connected with each other—without a centralized nervous system, in other words. This spontaneous syncing up of disconnected moving elements is called entrainment, and the best demo of it that I've seen is this one, using metronomes:
UCLA Physics
The reason why this isn't just neat but also useful is that it provides a secondary method of control for robots. If the centralized control system of your swimming robot gets busted, you can rely on this water pressure-mediated local control to generate a swimming motion. And there are applications for modular robots as well, since you can potentially create a swimming robot out of a bunch of different physically connected modules that don't even have to talk to each other.
For more details, we spoke with
Robin Thandiackal and Kamilo Melo at EPFL, first authors on the Science Robotics paper.
IEEE Spectrum: Why do you need a robot to do this kind of research?
Thandiackal and Melo: From a more general perspective, with this kind of research we learn and understand how a system works by building it. This then allows us to modify and investigate the different components and understand their contribution to the system as a whole.
In a more specific context, it is difficult to separate the different components of the nervous system with respect to locomotion within a live animal. The central components are especially difficult to remove, and this is where a robot or also a simulated model becomes useful. We used both in our study. The robot has the unique advantage of using it within the real physics of the water, whereas these dynamics are approximated in simulation. However, we are confident in our simulations too because we validated them against the robot.
How is the robot model likely to be different from real animals? What can't you figure out using the robot, and how much could the robot be upgraded to fill that gap?
Thandiackal and Melo: The robot is by no means an exact copy of a real animal, only a first approximation. Instead, from observing and previous knowledge of real animals, we were able to create a mathematical representation of the neuromechanical control in real animals, and we implemented this mathematical representation of locomotion control on the robot to create a model. As the robot interacts with the real physics of undulatory swimming, we put a great effort in informing our design with the morphological and physiological characteristics of the real animal. This for example accounts for the scaling, the morphology and aspect ratio of the robot with respect to undulatory animals, and the muscle model that we used to approximately represent the viscoelastic characteristics of real muscles with a rotational joint.
Upgrading the robot is not going to be making it more “biological.” Again, the robot is part of the model, not a copy of the real biology. For the sake of this project, the robot was sufficient, and only a few things were missing in our design. You can even add other types of sensors and use the same robot base. However, if we would like to improve our robot for the future, it would be interesting to collect other fluid information like the surrounding fluid speed simultaneously with the force sensing, or to measure hydrodynamic pressure directly. Finally, we aim to test our model of undulatory swimming using a robot with three-dimensional capabilities, something which we are currently working on.
Upgrading the robot is not going to be making it more “biological.” The robot is part of the model, not a copy of the real biology.
What aspects of the function of a nervous system to generate undulatory motion in water aren't redundant with the force feedback from motion that you describe?
Thandiackal and Melo: Apart from the generation of oscillations and intersegmental coupling, which we found can be redundantly generated by the force feedback, the central nervous system still provides unique higher level commands like steering to regulate swimming direction. These commands typically originate in the brain (supraspinal) and are at the same time influenced by sensory signals. In many fish the lateral line organ, which directly connects to the brain, helps to inform the brain, e.g., to maintain position (rheotaxis) under variable flow conditions.
How can this work lead to robots that are more resilient?
Thandiackal and Melo: Robots that have our complete control architecture, with both peripheral and central components, are remarkably fault-tolerant and robust against damage in their sensors, communication buses, and control circuits. In principle, the robot should have the same fault-tolerance as demonstrated in simulation, with the ability to swim despite missing sensors, broken communication bus, or broken local microcontroller. Our control architecture offers very graceful degradation of swimming ability (as opposed to catastrophic failure).
Why is this discovery potentially important for modular robots?
Thandiackal and Melo: We showed that undulatory swimming can emerge in a self-organized manner by incorporating local force feedback without explicit communication between modules. In principle, we could create swimming robots of different sizes by simply attaching independent modules in a chain (e.g., without a communication bus between them). This can be useful for the design of modular swimming units with a high degree of reconfigurability and robustness, e.g. for search and rescue missions or environmental monitoring. Furthermore, the custom-designed sensing units provide a new way of accurate force sensing in water along the entirety of the body. We therefore hope that such units can help swimming robots to navigate through flow perturbations and enable advanced maneuvers in unsteady flows. Continue reading
Posted in Human Robots
#439576 Video Friday: Robot Opera
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!):
RO-MAN 2021 – August 8-12, 2021 – [Online Event]DARPA SubT Finals – September 21-23, 2021 – Louisville, KY, USAWeRobot 2021 – September 23-25, 2021 – Coral Gables, FL, USAIROS 2021 – September 27-1, 2021 – [Online Event]ROSCon 2021 – October 21-23, 2021 – New Orleans, LA, USALet us know if you have suggestions for next week, and enjoy today's videos.
The New National Theatre Tokyo presents Super Angel: “Witness the birth of a new opera, performed by Alter 3, an android with artificial life who makes friends with children in the chorus as they sing and perform together.”
Alter 3 is characterized by its body, in which all interior mechanisms are exposed, and a face from which it is impossible to determine gender or age, and it is an android robot designed to feel life, which is unprecedented in the field. Researchers from Osaka University and the University of Tokyo, which are famous for their work into androids and artificial life, have been collaborating up until now to create and study two Alter androids. The main challenges of this are whether or not it is possible for robots to acquire a sense of life independently through interactivity with the outside world, and to answer the basic question of exactly what life is through the course of this. [ NNTT ] via [ Robotstart ]
Running the bases at Dodger Stadium is a fun tradition that many children look forward to after most Sunday games. But not all children, especially those who are currently hospitalized or recovering from an illness at home, can physically experience it. That's why UCLA Health, the Dodgers and OhmniLabs teamed up to create a virtual run-the-bases experience for 10 pediatric patients at UCLA Mattel Children's Hospital.
[ UCLA ]
Thanks, Joseph!
The way to teach robots to move like animals is to collect data from animals, and it's surprising how much of a difference some little tweaks can make to a quadrupedal gait.
[ ETHZ CRL ]
Thanks, Fan!
Walker X had me at back massage.
[ Ubtech ]
I needed this video today.
[ Soft Robotics ]
MIT faculty and staff reimagine an iconic mechanical engineering class – 2.007 (Design and Manufacturing I) – so students can go head-to-head in the final robot competition from their dorm rooms, apartments, or homes across the country.
The full competition livestreams are at the link below.
[ MIT 2.007 ]
The world's best female flair bartender vs the most advanced bartending robot. Who's gonna win?
Kuka's last human versus robot challenge involving table tennis was a huge disappointment, so I really hope this one is better.
[ Makr Shakr ]
I know software compliance is all the rage, but there's still something to be said for robot arms that are inherently soft.
[ Motion Intelligent Lab ]
Thanks, Fan!
We present a versatile, adhesive, and soft material (called VENOM) with high dynamic friction and normal adhesion forces on various smooth and rough surfaces. VENOM is a dry adhesive material based on a simple mixture of super-soft, fast cure platinum-catalyzed silicone and iron powder. Our result demonstrates the use of VENOM for the feet of our sprawling posture robot.
[ Paper ]
Thanks, Poramate!
Hybrid security by humans and robots. Communication with humans is handled by security guards through the avatar security robot Ugo, and hybrid security takes advantage of the characteristics of each robot and security guard.
What's the head on the stick at the end? I want one of those!
[ Ugo ]
Check out more views of the MQ-25 T1 test asset's historic flight, when it became the first unmanned aircraft to ever refuel another aircraft—piloted or autonomous—during flight. During a June 2021 flight test, the MQ25 T1 test asset transferred fuel to an F/A-18 Super Hornet.
[ Boeing ]
It's definitely cool to be able to do this with a robot, but it really makes you realize how effortless these tasks are for humans, right?
[ Extend Robotics ]
GE Research's Robotics and Autonomy team, led by Senior Robotics Scientist, Shiraj Sen, successfully completed Year 1 of a project with the US Army through its Scalable Adaptive Resilient Autonomy Program (SARA) to develop and demonstrate a risk-aware autonomous ground vehicle that was capable of navigating safely in complex off-road test conditions.
[ GE Research ]
Here's one way to add some safety to your industrial robot, I guess?
[ Kuka ]
Okay but seriously how is a kitchen “fully robotic” if you have to do all the prep and cleaning?
Also you left all the good stuff in the pot.
[ Moley ]
Here are a couple of videos showing some recent research from the Brussels Human Robotics Research Center (BruBotics); check the YouTube descriptions for paper references.
[ BruBotics ]
Thanks, Bram!
A Michigan Robotics Colloquium, hosted by the Robotics Graduate Student Council (RGSC), was held on July 27, 2021 about assistive technologies.
[ Michigan Robotics ] Continue reading
Posted in Human Robots
#439559 MIT is Building a Dynamic, Acrobatic ...
For a long time, having a bipedal robot that could walk on a flat surface without falling over (and that could also maybe occasionally climb stairs or something) was a really big deal. But we’re more or less past that now. Thanks to the talented folks at companies like Agility Robotics and Boston Dynamics, we now expect bipedal robots to meet or exceed actual human performance for at least a small subset of dynamic tasks. The next step seems to be to find ways of pushing the limits of human performance, which it turns out means acrobatics. We know that IHMC has been developing their own child-size acrobatic humanoid named Nadia, and now it sounds like researchers from Sangbae Kim’s lab at MIT are working on a new acrobatic robot of their own.
We’ve seen a variety of legged robots from MIT’s Biomimetic Robotics Lab, including Cheetah and HERMES. Recently, they’ve been doing a bunch of work with their spunky little Mini Cheetahs (developed with funding and support from Naver Labs), which are designed for some dynamic stuff like gait exploration and some low-key four-legged acrobatics.
In a paper recently posted to arXiv (to be presented at Humanoids 2020 in July), Matthew Chignoli, Donghyun Kim, Elijah Stanger-Jones, and Sangbae Kim describe “a new humanoid robot design, an actuator-aware kino-dynamic motion planner, and a landing controller as part of a practical system design for highly dynamic motion control of the humanoid robot.” So it’s not just the robot itself, but all of the software infrastructure necessary to get it to do what they want it to do.
MIT Humanoid performing a back flip off of a humanoid robot off of a 0.4 m platform in simulation.
Image: MIT
First let’s talk about the hardware that we’ll be looking at once the MIT Humanoid makes it out of simulation. It’s got the appearance of a sort of upright version of Mini Cheetah, but that appearance is deceiving, says MIT’s Matt Chignoli. While the robot’s torso and arms are very similar to Mini Cheetah, the leg design is totally new and features redesigned actuators with higher power and better torque density. “The main focus of the leg design is to enable smooth but dynamic ‘heel-to-toe’ actions that happen in humans’ walking and running, while maintaining low inertia for smooth interactions with ground contacts,” Chignoli told us in an email. “Dynamic ankle actions have been rare in humanoid robots. We hope to develop robust, low inertia and powerful legs that can mimic human leg actions.”
The design strategy matters because the field of humanoid robots is presently dominated by hydraulically actuated robots and robots with series elastic actuators. As we continue to improve the performance of our proprioceptive actuator technology, as we have done for this work, we aim to demonstrate that our unique combination of high torque density, high bandwidth force control, and the ability to mitigate impacts is optimal for highly dynamic locomotion of any legged robot, including humanoids.
-Matt Chignoli
Now, it’s easy to say “oh well pfft that’s just in simulation and you can get anything to work in simulation,” which, yeah, that’s kinda true. But MIT is putting a lot of work into accurately simulating everything that they possibly can—in particular, they’re modeling the detailed physical constraints that the robot operates under as it performs dynamic motions, allowing the planner to take those constraints into account and (hopefully) resulting in motions that match the simulation pretty accurately.
“When it comes to the physical capabilities of the robot, anything we demonstrate in simulation should be feasible on the robot,” Chignoli says. “We include in our simulations detailed models for the robot’s actuators and battery, models that have been validated experimentally. Such detailed models are not frequently included in dynamic simulations for robots.” But simulation is still simulation, of course, and no matter how good your modeling is, that transfer can be tricky, especially when doing highly dynamic motions.
“Despite our confidence in our simulator’s ability to accurately mimic the physical capabilities of our robot with high fidelity, there are aspects of our simulator that remain uncertain as we aim to deploy our acrobatic motions onto hardware,” Chignoli explains. “The main difficulty we see is state estimation. We have been drawing upon research related to state estimation for drones, which makes use of visual odometry. Without having an assembled robot to test these new estimation strategies on, though, it is difficult to judge the simulation to real transfer for these types of things.”
We’re told that the design of the MIT Humanoid is complete, and that the plan is to build it for real over the summer, with the eventual goal of doing parkour over challenging terrains. It’s tempting to fixate on the whole acrobatics and parkour angle of things (and we’re totally looking forward to some awesome videos), but according to Chignoli, the really important contribution here is the framework rather than the robot itself:
The acrobatic motions that we demonstrate on our small-scale humanoid are less about the actual acrobatics and more about what the ability to perform such feats implies for both our hardware as well as our control framework. The motions are important in terms of the robot’s capabilities because we are proving, at least in simulation, that we can replicate the dynamic feats of Boston Dynamics’ ATLAS robot using an entirely different actuation scheme (proprioceptive electromagnetic motors vs. hydraulic actuators, respectively). Verification that proprioceptive actuators can achieve the necessary torque density to perform such motions while retaining the advantages of low mechanical impedance and high-bandwidth torque control is important as people consider how to design the next generation of dynamic humanoid robots. Furthermore, the acrobatic motions demonstrate the ability of our “actuator-aware” motion planner to generate feasible motion plans that push the boundaries of what our robot can do.
The MIT Humanoid Robot: Design, Motion Planning, and Control For Acrobatic Behaviors, by Matthew Chignoli, Donghyun Kim, Elijah Stanger-Jones, and Sangbae Kim from MIT and UMass Amherst, will be presented at Humanoids 2020 this July. You can read a preprint on arXiv here. Continue reading
Posted in Human Robots
#439555 Unitree’s Go1 Robot Dog Looks Pretty ...
In 2017, we first wrote about the Chinese startup Unitree Robotics, which had the goal of “making legged robots as popular and affordable as smartphones and drones.” Relative to the cost of other quadrupedal robots (like Boston Dynamics’ $74,000 Spot), Unitree’s quadrupeds are very affordable, with their A1 costing under $10,000 when it became available in 2020. This hasn’t quite reached the point of consumer electronics that Unitree is aiming for, but they’ve just gotten a lot closer: now available is the Unitree Go1, a totally decent looking small size quadruped that can be yours for an astonishingly low $2700.
Not bad, right? Speedy, good looking gait, robust, and a nifty combination of autonomous human following and obstacle avoidance. As with any product video, it’s important to take everything you see here with a grain of salt, but based on Unitree’s track record we have no particular reason to suspect that there’s much in the way of video trickery going on.
There are three versions of the Go1: the $2700 base model Go1 Air, the $3500 Go1, and the $8500 Go1 Edu. This looks to be the sort of Goldilocks pricing model, where most people are likely to spring for the middle version Go1, which includes better sensing and compute as well as 50% more battery life an an extra m/s of speed (up to 3.5m/s) for a modest premium in cost. The top of the line Edu model offers higher end computing, 2kg more payload (up to 5kg), as well as foot-force sensors, lidar, and a hardware extension interface and API access. More detailed specs are here, although if you’re someone who actually cares about detailed robot specs, what you’ll find on Unitree’s website at the moment will probably be a little bit disappointing.
We’ve reached out to Unitree to ask them about some of the specs that aren’t directly addressed on the website. Battery life is a big question—the video seems to suggest that the Go1 is capable of a three-kilometer, 20-minute jog, and then some grocery shopping and a picnic, all while doing obstacle avoidance and person following and with an occasional payload. If all of that is without any battery swaps, that’s pretty good. We’re also wondering exactly what the “Super Sensory System” is, what kinds of tracking and obstacle avoidance and map making skills the Go1 has, and exactly what capabilities you’ll be required to spring for the fancier (and more expensive) versions of the Go1 to enjoy.
Honestly, though, we’re not sure what Unitree could realistically tell us about the Go1 where we’d be like, “hmm okay maybe this isn’t that great of a deal after all.” Of course the real test will be when some non-Unitree folks get a hold of a Go1 to see what it can actually do (Unitree, please contact me for my mailing address), but even at $3500 for the midrange model, this seems like an impressively cost effective little robot.
Update: we contacted Unitree for more details, and they’ve also updated the Go1 website to include the following:
The battery life of the robot while jogging is about 1 hour
It weighs 12kg
The Super Sensory System includes five wide-angle stereo depth cameras, hypersonic distance sensors, and an integrated processing system
It’s running at 16 core CPU and a 1.5 tflop GPU
We also asked Wang Xingxing, Unitree’s CEO, about how they were able to make Go1 so affordable, and here’s what he told us:
Unitree Go1 can be regarded as a product that we have achieved after 6-7 years of iteration at the hardware level, only to achieve the goals of ultra-low cost, high reliability and high performance. Our company actually spent more manpower and money than software on the hardware level such as machinery. Continue reading
Posted in Human Robots