A simple robot that performs line-following or obstacle avoidance can fit all of its logic inside a single Arduino sketch. But as a robot’s autonomy increases, its corresponding software gets complicated very quickly. It won’t be long before diagnostic monitoring and logging comes in handy, or the desire to encapsulate feature areas and orchestrate how they work together. This is where tools like the Robot Operating System (ROS) come in, so we don’t have to keep reinventing these same wheels. And Open Robotics just released ROS 2 Dashing Diademata for all of us to use.
ROS is an open source project that’s been underway since 2007 and updated regularly, each named after a turtle species. What makes this one worthy of extra attention? Dashing marks the first longer term support (LTS) release of ROS 2, a refreshed second generation of ROS. All high level concepts stayed the same, meaning almost everything in our ROS orientation guide is still applicable in ROS 2. But there were big changes under the hood reflecting technical advances over the past decade.
ROS was built in an age where a Unix workstation cost thousands of dollars, XML was going to be how we communicate all data online, and an autonomous robot cost more than a high-end luxury car. Now we have $35 Raspberry Pi running Linux, XML has fallen out of favor due to processing overhead, and some autonomous robots are high-end luxury cars. For these and many other reasons, the people of Open Robotics decided it was time to make a clean break from legacy code.
The break has its detractors, as it meant leaving behind the vast library of freely available robot intelligence modules released by researchers over the years. Popular ones were (or will be) ported to ROS 2, and there is a translation bridge sufficient to work with some, but the rest will be left behind. However, this update also resolved many of the deal-breakers preventing adoption outside of research, making ROS more attractive for commercial investment which should bring more robots mainstream.
Judging by responses to the release announcement, there are plenty of people eager to put ROS 2 to work, but it is not the only freshly baked robotics framework around. We just saw Nvidia release their Isaac Robot Engine tailored to make the most of their Jetson hardware.
Robots that can dynamically reconfigure themselves to adapt to their environments offer a promising advantage over their less dynamic cousins. Researchers have been working through all the challenges of realizing that potential: hardware, software, and all the interactions in between. On the software end of the spectrum, a team at University of Pennsylvania’s ModLab has been working on a robot that can autonomously choose a configuration to best fit its task at hand.
We’ve recently done an overview of modular robots, and we noted that coordination and control are persistent challenges in this area. The robot in this particular demonstration is a hybrid: a fixed core module serving as central command, plus six of the lab’s dynamic SMORES-EP modules. The core module has a RGB+Depth camera for awareness of its environment. A separate downwards-looking camera watches SMORES modules for awareness of itself.
Combining that data using a mix of open robot research software and new machine specific code, this team’s creation autonomously navigates an unfamiliar test environment. While it can adapt to specific terrain challenges like a wood staircase, there are still limitations on situations it can handle. Kudos to the researchers for honestly showing and explaining how the robot can get stuck on a ground seam, instead of editing that gaffe out to cover it up.
While this robot isn’t the completely decentralized modular robot system some are aiming for, it would be a mistake to dismiss based on that criticism alone. At the very least, it is an instructive step on the journey offering a tradeoff that’s useful on its own merits. And perhaps this hybrid approach will find application with a modular robot close to our hearts: Dtto, the winner of our 2016 Hackaday Prize.
[via Science News]
Continue reading “SMORES Robot Finds Its Own Way To The Campfire”
Disney is working on modular, intelligent robot limbs that snap into place with magnets. The intelligence comes from a reasonable sized neural network that also incorporates some modularity. The robot is their Snapbot whose base unit can fit up to eight of limbs, and so far they’ve trained with up to three together.
The modularity further extends to a choice of three types of limb. One with roll and pitch, another with yaw and pitch, and a third with roll, yaw, and pitch. Interestingly, of the three types, the yaw-pitch one seems most effective.
In this age of massive, deep neural networks requiring GPUs or even online services for training in a reasonable amount of time, it’s refreshing to see that this one’s only two layers deep and can be trained in three hours on a single-core, 3.4 GHz Intel i7 processor. Three hours may still seem long, but remember, this isn’t a simulation in a silicon virtual world. This is real-life where the servo motors have to actually move. Of course, they didn’t want to sit around and reset it after each attempt to move across the table so they built in an automatic mechanism to pull the robot back to the starting position before trying to cross the table again. To further speed training, they found that once they’d trained for one limb, they could then copy the last of the network’s layers to get a head starting on the training for two limbs.
Why do training? Afterall, we’ve seen pretty awesome multi-limbed robots working with manual coding, an example being this hexapod tank based on one from the movie Ghost in the Shell. They did that too and then compared the results of the manual approach with those of the trained one and the trained one moved further in the same amount of time. At a minimum, we can learn a trick or two from this modular crawler.
Check out their article for the details and watch it in action in its learning environment below.
Continue reading “Disney’s New Robot Limbs Trained Using Neural Networks”
The greatest challenge of robotics is autonomy. Usually, this means cars that can drive themselves, a robotic vacuum that won’t drive down the stairs, or a rover on Mars that can drive on Mars. This project is nothing like that. Instead of building a robot with a single shape, this robot is made out of several modules that can self-assemble into different structures. It’s an organized fleet of robots, all helping each other, like an ant colony, or our future as Gray Goo.
If the idea of self-assembling modular robots sounds familiar, you’re right. The Dtto won the Grand Prize in the 2016 Hackaday Prize, and it’s a beast of a project. It’s an ouroboros of a robot that can assemble itself into a snake, a wheel, or an arm. It’s weird, but if you want a robot that can do anything, this is the kind of modularity that you need. One step closer to Gray Goo, at least.
Like Dtto, the noMad can transform itself into bridges, arms, snakes, and wheels by assembling each individual piece into one component of a massive structure. It’s something we rarely see, and it’s a difficult computational and engineering problem. Still, the progress the team behind noMad has been making is remarkable, and we can’t wait to see the finished project.