Cheetah 3 Is Learning To Move Blindly Before Learning To See

July 11, 2018 by Steven Dufresne 23 Comments

Stand up right now and walk around for a minute. We’re pretty sure you didn’t see everywhere you stepped nor did you plan each step meticulously according to visual input. So why should robots do the same? Wouldn’t your robot be more versatile if it could use its vision to plan a path, but leave most of the walking to the legs with the help of various sensors and knowledge of joint positions?

That’s the approach [Sangbae Kim] and a team of researchers at MIT are taking with their Cheetah 3. They’ve given it cameras but aren’t using them yet. Instead, they’re making sure it can move around blind first. So far they have it walking, running, jumping and even going up stairs cluttered with loose blocks and rolls of tape.

Cheetah 3 jumping 30 inches onto a desk — Jumping 30 inches onto a desk

Two algorithms are at the heart of its being able to move around blind.

The first is a contact detection algorithm which decides if the legs should transition between a swing or a step based on knowledge of the joint positions and data from gyroscopes and accelerometers. If it tilted unexpectedly due to stepping on a loose block then this is the algorithm which decides what the legs should do.

The second is a model-predictive algorithm. This predicts what force a leg should apply once the decision has been made to take a step. It does this by calculating the multiplicative positions of the robot’s body and legs a half second into the future. These calculations are done 20 times a second. They’re what help it handle situations such as when someone shoves it or tugs it on a leash. The calculations enabled it to regain its balance or continue in the direction it was headed.

There are a number of other awesome features of this quadruped robot which we haven’t seen in others such as Boston Dynamics’ SpotMini like invertible knee joints and walking on three legs. Check out those features and more in the video below.

Of course, SpotMini has a whole set of neat features of its own. Let’s just say that while they look very similar, they’re on two different evolutionary paths. And the Cheetah certainly has evolved since we last looked at it a few years ago.

Continue reading “Cheetah 3 Is Learning To Move Blindly Before Learning To See” →

3D Printed Arduino Bot Is Limbo Master

July 10, 2018 by Tom Nardi 9 Comments

As if we didn’t have enough to worry about in regards to the coming robot uprising, [Ali Aslam] of Potent Printables has recently wrapped up work on a 3D printed robot that can flatten itself down to the point it can fit under doors and other tight spaces. Based on research done at UC Berkeley, this robot is built entirely from printed parts and off the shelf hardware, so anyone can have their own little slice of Skynet.

On display at East Coast RepRap Festival

The key to the design are the folding “wings” which allow the robot to raise and lower itself on command. This not only helps it navigate tight spaces, but also gives it considerable all-terrain capability when it’s riding high. Rather than wheels or tracks, the design uses six rotors which look more like propellers than something you’d expect to find on a ground vehicle. These rotors work at the extreme angles necessary when the robot has lowered itself, and allow it to “step” over obstructions when they’re vertical.

For the electronics, things are about what you’d expect. An Arduino Pro Mini combined with tiny Pololu motor controllers is enough to get the bot rolling, and a Flysky FS-X6B receiver is onboard so the whole thing can be operated with a standard RC transmitter. The design could easily be adapted for WiFi or Bluetooth control if you’d rather not use RC gear for whatever reason.

Want to build your own? All of the STL files, as well as a complete Bill of Materials, are available on the Thingiverse page. [Ali] even has a series of videos on YouTube videos walking through the design and construction of the bot to help you along. Outside of the electronics, you’ll need a handful of screws and rods to complement the 50+ printed parts. Better start warming up the printer now.

As an interesting aside, we got a chance to see this little critter first hand at the recent East Coast RepRap Festival in Maryland, along with a number of other engineering marvels.

Continue reading “3D Printed Arduino Bot Is Limbo Master” →

Disney’s Humanoid Stunt Robots Throw Multiple Backflips No Sweat

July 8, 2018 by Mike Szczys 19 Comments

What’s the biggest problem right now with humanoid robots? They fall down. Disney seems to have solved that problem here by making robots that are meant to fall down and be caught by a net. Disney’s research arm (you may know them as Imagineers) is showing off a robot called Stuntronic which can perform controlled somersaults as it flies through the air. Check the video below, you really have to watch a few times to make sure this is a robot and not a person.

It’s really interesting to follow the evolution of this robot. It began with BRICK, a limbless rectangular bot that could shift its center of gravity to control orientation while moving through the air. From there, Stickman adapted those concepts into a stick-shaped robot that had two hinged portions which allowed for controlled somersaults as it flew through the air. Stuntronic feels like a big leap from that design.

As with Stickman, it can bend to control somersaults mid-air, but with the addition of articulated arms, Stuntronic can also add twists to the acrobatic bag of tricks. To our eye, this is very lifelike and manages to completely escape the uncanny valley. This is a ringing endorsement since one of the proposed purposes of this research is for live performances at Disney’s theme parks.

The Hall of Presidents was a marvel of its time, as robots presented famous speeches while decorated in the likeness of the leaders who originally delivered them. But to stand and deliver is a trick of decades past. We hope this is a trick of next year and not something we’ll have to wake decades into the future to see in person.

Oh, and for those wondering if Stuntronic stuck the landing? The controlled delivery into the net’s warm embrace is equally impressive. Hopefully, successful landings are commonplace because they’re launching these bots with some really wicked force! In addition to the gyroscopes and accelerometers you’d expect to find in a motion-aware machine, the design uses a trio of laser rangefinders that triangulate ground position to spot the optimal landing. We haven’t seen a publication for this bot yet but check the Stickman info for more on these sensors.

Continue reading “Disney’s Humanoid Stunt Robots Throw Multiple Backflips No Sweat” →

An Optical Mouse Sensor For Robotic Vision

July 6, 2018 by Jenny List 22 Comments

Readers with long memories will remember the days when mice and other similar pointing devices relied upon a hard rubber ball in contact with your desk or other surface, that transmitted any motion to a pair of toothed-wheel rotation sensors. Since the later half of the 1990s though, your rodent has been ever significantly more likely to rely upon an optical sensor taking the form of a small CCD camera connected to motion sensing electronics. These cameras are intriguing components with applications outside pointing devices, as is shown by [FoxIS] who has used one for robot vision.

The robot in question is a skid-steer 4-wheeled toy, to which he has added an ADNS3080 mouse sensor fitted with a lens, an H-bridge motor driver board, and a Wemos D1 Mini single board computer. The D1 serves a web page showing both the image from the ADNS3080 and an interface that allows the robot to be directed over a network connection. A pair of LiPo batteries complete the picture, with voltage monitoring via one of the Wemos analogue pins.

The ADNS3080 is an interesting component and we’d love see more of it. This laser distance sensor or perhaps this car movement tracker should give you some more info. We’ve heard rumors of them being useful for drones. Anyone?

Junk Bin Self Balancing Bot With ESP8266

July 5, 2018 by Tom Nardi 10 Comments

As we all know, sometimes the projects we plan simply never materialize. You have an idea, maybe even buy some of the parts you need, and then…nothing. Maybe you changed your mind, or maybe the idea was never that good to begin with. In any event, time marches on, the parts pile up, and the ideas come and go. Such is the life of the hacker.

[Andrius Mikonis] writes in to tell us how his graveyard of abandoned projects ended up providing exactly what he needed to embark on a project he’s been fascinated with for years: the two-wheel self balancing robot. He started with a motor and wheel set that was originally intended to be part of a rover, added an accelerometer, and tied the whole thing together with an ESP-01 he had lying around. The final result certainly looks the part, and goes to show that projects don’t always need to be 1000 hour labors of love to accomplish their goals.

The construction of this little bot is simple in the extreme. A piece of plywood makes up the primary structure, with the wheels glued to the bottom and the electronics taking up residence in the top. It’s powered by two lithium battery cells that were salvaged out of an old laptop, with a DC-DC buck converter to provide a stable 3.3 VDC for the ESP-01 and MPU6050 accelerometer. To control the motors themselves, [Andrius] is using a cheap L293 controller that he found on eBay.

For interactive control, [Andrius] is making use of the ESP’s Wi-Fi to provide a web-based interface. This lets you control the bot from essentially any device that has a browser, rather than having to use a dedicated hardware transmitter.

Self-balancing robots of various levels of complexity are a relatively common project in the hacker world. There’s just something magical about the way they scoot around, seeming to defy gravity.

Wrangling RC Servos Becoming A Hassle? Try Serial Bus Servos!

July 5, 2018 by Roger Cheng 24 Comments

When we need actuators for a project, a servo from the remote-control hobby world is a popular solution. Though as the number of servos go up, keeping their wires neat and managing their control signals become a challenge. Once we start running more servos than we have fingers and toes, it’s worth considering the serial bus variety. Today we’ll go over what they are and examine three products on the market.

Continue reading “Wrangling RC Servos Becoming A Hassle? Try Serial Bus Servos!” →

Disney’s New Robot Limbs Trained Using Neural Networks

July 3, 2018 by Steven Dufresne 9 Comments

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” →