Robot Leaps Uncanny Valley On Backward Knees

February 3, 2017 by Mike Szczys 32 Comments

We’ve covered a ton of Boston Dynamics robots but this is the second one in a row that has shown a departure from what a lot of people’s notion of an ‘advanced’ robot should look like. It’s a cellphone camera clip of a video played at a conference, but at least it isn’t vertical video — kudos to [juvertson]. At about 3:40 seconds into the video you get a good look “Handle” at a four-limbed robot with backwards joints and wheel.

This design makes a lot of sense and it’s good to see Boston Dynamics thinking about unique robot kinematics alongside the realities of motion. The result is something that appears neither human nor animal — it’s definitely not natural. Despite the presenter’s assertion that this will be nightmare-inducing, we think it’s the opposite, since it doesn’t tweak that string in your brain that cries “predator”.

Obviously this is what we’d call a self-balancer. But two-wheels-plus-rigid-frame it is not. The articulated lower limbs allow it to shift its mass over the wheels. The upper limbs play their part in balancing, at one point acting in the same way a figure skater’s arms would during a spin. And its dexterity in hopping over an obstacle is only made better by [juvertson’s] commentary. This is a really good balance between purely wheeled and purely humanoid designs and a nice addition to the evolution of robotics.

Continue reading “Robot Leaps Uncanny Valley On Backward Knees” →

Tiny Robot Clings To Leaves With Static Electricity

January 31, 2017 by Lewin Day 15 Comments

Flying is an energy-intensive activity. The birds and the bees don’t hover around incessantly like your little sister’s quadcopter. They flit to and fro, perching on branches and leaves while they plan their next move. Sure, a quadcopter can land on the ground, but then it has to spend more energy getting back to altitude. Researchers at Harvard decided to try to develop flying robots that can perch on various surfaces like insects can.

Perching on surfaces happens electrostatically. The team used an electrode patch with a foam mounting to the robot. This allows the patch to make contact with surfaces easily even if the approach is a few degrees off. This is particularly important for a tiny robot that is easily affected by even the slightest air draft. The robots were designed to be as light as possible — just 84mg — as the electrostatic force is not particularly strong.

It’s estimated that perching electrostatically for a robot of this size uses approximately 1000 times less power than during flight. This would be of great use for surveillance robots that could take up a vantage point at altitude without having to continually expend a great deal of energy to stay airborne. The abstract of the research paper notes that this method of perching was successful on wood, glass, and a leaf. It appears testing was done with tethers; it would be interesting to see if this technique would be powerful enough for a robot that carries its own power source. Makes us wonder if we ever ended up with tiny flyers that recharge from power lines?

We’re seeing more tiny flying robots every day now – the IMAV 2016 competition was a great example of the current state of the art.

Continue reading “Tiny Robot Clings To Leaves With Static Electricity” →

Ping Pong Ball-Juggling Robot

January 22, 2017 by Bryan Cockfield 6 Comments

There aren’t too many sports named for the sound that is produced during the game. Even though it’s properly referred to as “table tennis” by serious practitioners, ping pong is probably the most obvious. To that end, [Nekojiru] built a ping pong ball juggling robot that used those very acoustics to pinpoint the location of the ball in relation to the robot. Not satisfied with his efforts there, he moved onto a visual solution and built a new juggling rig that uses computer vision instead of sound to keep a ping pong ball aloft.

The main controller is a Raspberry Pi 2 with a Pi camera module attached. After some mishaps with the planned IR vision system, [Nekojiru] decided to use green light to illuminate the ball. He notes that OpenCV probably wouldn’t have worked for him because it’s not fast enough for the 90 fps that’s required to bounce the ping pong ball. After looking at the incoming data from this system, an algorithm extracts 3D information about the ball and directs the paddle to strike the ball in a particular way.

If you’ve ever wanted to get into real-time object tracking, this is a great project to look over. The control system is well polished and the robot itself looks almost professionally made. Maybe it’s possible to build something similar to test [Nekojiru]’s hypothesis that OpenCV isn’t fast enough for this. If you want to get started in that realm of object tracking, there are some great projects that make use of that piece of software as well.

Taking Control Of Your Furby

January 21, 2017 by Rich Hawkes 17 Comments

Furbys have been around for a while and they are an interesting (if annoying) toy that will teach the kids to be okay with their eventual robotic overlords. In the meantime, the latest version of the robotic companion/toy/annoyance uses Bluetooth LE to communicate with the owner and [Jeija] has been listening in on the Bluetooth communication, trying to reverse engineer the protocol in order to run code on Furby.

[Jeija] has made a lot of progress and can already control the Furby’s actions, antenna and backlight color, and change the Furby’s emotional state by changing the values of the Furby’s hungriness, tiredness, etc. [Jeija] has created a program that runs on top of Node.js and can communicate with the Furby and change its properties. [Jeija] has also discovered, and can bring up, a secret debug menu that displays in the Furby’s eyes. Yet to be discovered is how to run your own code on the Furby, however, [Jeija] is able to add custom audio to the official DLC files and upload them into the Furby.

[Jeija] points out the all this was done without taking a Furby apart, only by sniffing the Bluetooth communication between the robot and the controlling app (Android/iOS device.) Check out a similar hack on the previous generation of Furbys, as well as a replacement brain for them. We just hope that the designers included a red/green LED so that we will all know when the Furbys switch from good to evil.

Continue reading “Taking Control Of Your Furby” →

Lego Boosts Their Robotic Offering

January 21, 2017 by Al Williams 22 Comments

Kids often have their first exposure to robots in school using Lego Mindstorm kits. Now Lego is rolling out Boost — a robotic kit targeting all Lego builders from 7 years old and up. The kit is scheduled to be on the market later this year (it appeared at the recent CES) and will sell for about $160.

[The Brothers Brick] had a chance to try the kit out at CES (see the video below) and you might find their review interesting. The kit provides parts and instructions to build five different models: a cat, a robot, a guitar, a 3D printer, and a tracked vehicle. You can check out the official page, too.

Continue reading “Lego Boosts Their Robotic Offering” →

Brain Controlled Tracked Robot

January 20, 2017 by Pedro Umbelino 1 Comment

[Imetomi] found himself salvaging a camera from a broken drone when he decided to use it in a new project, a tracked robot with a live video feed from the mounted camera.

… I had a cheap Chinese drone that was broken, but its camera seemed to be operating and when I took apart my drone I found a small WiFi chip with a video transmitter. I (decided) that I will use this little circuit for a project and I started to buy and salvage the parts.

Being a tracked robot, it can negotiate most types of terrain and climb hills up to 40 degrees. It is powered by two 18650 lithium-ion batteries with a capacity of 2600 mAh and the remote control is based on the HC-12 serial communication module. You can control it with a joystick and watch the camera’s live-stream in a virtual reality glass. That’s pretty neat but it’s not all.

[Imetomi] also used a hacked Nacomimi Brainwave Toy to make a brain controlled version of his robot. The brainwaves are detected using sensors placed on the scalp. To actually control it the operator has to focus on the right hand to move right, focus on the left hand to move left, blink to move forward and blink again to stop. There is also an ultrasonic sensor to help navigation so the robot doesn’t bump into things. It’s not very precise but you can always build the joystick version or, even better, make a version with both controls.

Continue reading “Brain Controlled Tracked Robot” →

Sentry Robot Turns Bad Cat To Good

January 20, 2017 by Donald Papp 48 Comments

The household of [James Watts] has cats, and those cats have decided that various spots of carpet are just great for digging up with their claws. After some efforts at training the cats, [James] enlisted a robotic cat trainer with remote wireless sensors. The automated trainer does only one job, but it does that one job reliably and tirelessly, which is just what is needed in this case. A task like “automate training the cats to stop clawing the carpet” is really made up of many smaller problems, and [James] implemented a number of clever ideas in his solution.

First of all, the need for an automated solution has a lot to do with how pets form associations, and the need to have the negative reinforcement be in the right place at the right time to be effective. A harmless spritz of water in this case is used for correction and needed to be applied immediately, consistently, and “from out of nowhere” (instead of coming from a person.) Otherwise, as [James] discovered, spraying water when the cats clawed the carpet simply meant that they stopped doing it when he was around.

There were a number of tricky problems to solve in the process. One was how to reliably detect cats actually clawing the carpet. Another was how to direct the harmless spray of water to only the spot in question, and how to rig and manage a water supply without creating another mess in the process. Finally, the whole thing needed to be clean and tidy; a hackjob with a mess of wires strung everywhere just wouldn’t do.

To achieve all this, [James] created a main sprayer unit that is wirelessly connected to remote sensor units using NRF24L01+ serial packet radios. When a remote senses that a trouble spot is being clawed, the main unit uses an RC servo to swivel a spray nozzle in the correct direction and give the offending feline a watery reminder.

The self-contained remote sensors use an accelerometer to detect the slight lifting of the carpet when it’s being clawed. [James] programmed the MMA8452Q three axis accelerometer to trigger an external pin when motion is sensed above a certain threshold, and this event is sent over the wireless link.

For the main sprayer unit itself, [James] cleverly based it around an off-the-shelf replacement windshield washer tank. With an integrated pump, tubing, and assortment of nozzles there was no need to design any of those elements from scratch. If you want to give the project a shot, check out the github repository — probably worth it it since one night is all it took to change the cat behavior which explains the lack of any action video.

Pet projects usually center around automating the feeding process, but it’s nice to see other applications. For something on the positive-reinforcement end of training, check out this cat exercise wheel that integrates a treat dispenser to encourage an exercise regimen.