Insect-inspired flying robot handles collisions, goes where other robots can’t | Hacker News Reader

Insect-inspired flying robot handles collisions, goes where other robots can’t

robohub.org

165 pointsrobotgal41 comments

With out trying to be childish, my second thought was, "I'd love to whack that thing with a bat".

I cant think what or how, but it seems to me that are more refined, more rugged version could give rise to a brilliant game / toy type thing. Yeah, Im sure their ideas for its uses are way more sensible. But, well, I want to whack it!!!

Quidditch?

Yeah, whatever invention is, playing with it seems to be the first wish. Quidditch should be so much nicer with this robot.

Yep!!!!

Jedi training

Hell yeah. Precisely what I couldn't think of!!!!!!!!!!

This looks awesome and I want one. I think Parrot (makers of the AR Drone) should license this immediately. Their product is fun, but it's still quite scary to pilot indoors with people around. With this I can imagine playing AR drone tag!

I agree. If they start producing this robot as an affordable toy -kids and parents would love it! And it should give them the best and the most numerous use cases.

Some more info on the project and a video of a similar project at the epfl:

http://lis.epfl.ch/airburr

http://www.youtube.com/watch?v=dm90IV3tMjI

This is such a seemingly simple idea, I'm surprised that many of the toymakers selling helis for indoor use haven't arrived at this conclusion already.

It's astonishingly elegant. I can't imagine nobody's thought of this before, so it's likely that the implementation is harder than it appears.

The added weight of the cage and gimbal system likely required them to use expensive ultra-light material and extra strong electronics. The motor on this thing looks pretty beefy, which in turn would require better driver components and a bigger (and/or denser) battery.

That said, I hope they can get a production version (made of syrofoam maybe?) so I can play with it. My main issue with hacking on quadcopters is how fragile they are.

I wonder if there's a "gimbal lock" equivalent where impacts in certain alignments relative to the gimbal mechanism still transmit a fairly large moment to the central unit.

If you have to take care of that situation, maybe it would be better to take care of all the impact compensation in software, and save the weight of the gimbal mechanism.

I've seen a few that looked a lot like this actually. For example:

http://www.amazon.com/ROBOTIC-3-Channel-Flying-Ball-Helicopt...

The main difference in the research here just seems to be the separation between the outer ball and the stabilized machinery inside it.

good lord this is awesome. i love the video of it bouncing off that guy's head...total contrast to whenever you're flying a parrot AR quadcopter and everyone takes notice and tries to get away from it.

we've been working with a professor at georgia tech about using a drone to aid in fruit-tree spotting and urban foraging, and i would be soo excited to have one of these...especially cool that they let it navigate off a magnetic direction and just say "figure it out". i just picture a future of artificial scents and pheromones broadcasting off of objects to lure flying robots to them.

I think this is a mature, well engineered approach to collisions - one that engineers can learn from.

Rather than throwing a hissy fit when it encounters an exception, a mature system should try to gracefully recover... Because we're never going to have total control over the inputs / environment. Especially when we're just a small bug in a big world.

Clever and simple. I like it. There are all sorts of serious uses for this, some scary, some exciting. Here's a silly one: I saw a large group of people playing a game of Quiddich in Central Park earlier in the summer. A smaller, faster version of this would make a great snitch(sp?) for them.

That would solve one of the many, many problems with "Muggle Quidditch". I have a lot of proposals for how they could turn that into a good game though.

First, ditch the brooms. You're riding unicycles from now on. Second, the bludgers need to be medicine balls and they need to be tethered to a fleet of quadrotors flying above using sophisticated algorithms to adjust their flight paths so that they attack players and veer off when hit with bats. And then yes, finally, the snitch needs to be a robot.

Then you'll have something I'd be willing to watch people play.

though unicycles? that sounds HARD :/ why not just run around? (or use those cool two-wheeled platform type bikes)

Well, a few reasons. Broomstick riding is supposed to be hard, so unicycle riding seems like a reasonable analogue. Another reason is that it has the similar aspect of falling off, while still being extremely maneuverable.

And lastly, I have this mental image of a nerdy guy peddling a unicycle like hell away from a 20 lb flying medicine ball, while another player cruises over to smack it out of the way with a baseball bat. There is no way any alternative is going to live up to that vision.

Very good idea this solve one of the biggest problems of this type of robots. However currently flying time rates are too short, so we need more power in order to use them in practical applications.i hope the graphene super capacitors can deliver that boost in power that can overcome this issue.

Neat idea! But what makes you think that graphene super capacitors are the best option? Why not some combination with solar battery(if it's going to fly outside) or some other kind of energy source?

Because the relation weight/power its outstanding. well i don't think using solar could help at all,because it will make the robot heavier and it will need a big area in order to power up the robot which for this type of robots its not possible i think, and also unless solar cells improve their energy efficiency this doesnt sound possible yet.

This is the second worst application of the "animal-inspired" moniker that I have seen.

Nike calling their sneakers "bio-inspired" because they had leopard spots on them was the worst.

I think the linkage was appropriate. While reading the title I immediately thought about the way a housefly repeatedly rams itself into windows (while not seriously damaging itself) to try and find a way out.

Granted, insects don't fly in bucky-ball-esque gyro cages, but are there any other types of flying organisms that fly in a similar manner? Definitely not birds -- as windex commercials have demonstrated.

I can assume there are people who look at this for it's military application. I'm trying to imagine what kind of search and rescue situation this can handle.

This also has the added benefit of keeping the rotors from causing harm. You'd have to actually stick your hand into the sphere in order to cut off a finger vs some quadcopters which force the operator to be aware.

Probably not a problem when there is a human flying it, but I wonder how difficult it would be to get an autopilot to recognise/deal with gimbal lock.

Recognizing gimbal lock wouldn't be very difficult. Using a rotary encoder on each axis, you could measure their angles and tell if any pair of gimbals was aligned. And apparently there are techniques for arbitrarily rotating one gimbal when lock is detected.

That would require making the gimbals higher tech than they currently are now though.

I'm wondering if you could perhaps detect it by detecting a sudden unplanned twist around one axis with an onboard accelerometer.

It doesn't really add that much complexity to the gimbals. Rotary encoders are extremely inexpensive, and the information would just be interpreted by the central controller.

I wonder why they went with a coaxial rotor and control surfaces rather than a typical multirotor (e.g. quadcopter), which seems much simpler.

The size (by cross-section surface area or pretty much any other metric) to thrust ratio is substantially improved, meaning they need a smaller, lighter ball encasing the unit.

I suspect that the thrust-to-weight ratio is also better since they're carrying fewer motors and ESCs around at the expense of some servos and surfaces, which are almost certainly lighter.

Plus the rotors are in a much more advantageous position for containment within a ball.

I'd be interested in evaluating the efficiency of vectoring prop wash using control surfaces versus modulating the speed of fixed-pitch rotors. There's almost certainly something there but I don't know which direction the comparison would go.

Anecdotally it's much easier to get good yaw authority using some type of thrust vectoring (control surfaces, tricopter style tilt-rotor, etc) than using counter-rotation like most quadcopters do as well.

I'd love to build one of these types of machine even sans ball and see how tough the trade-offs vs. a traditional multirotor are in real life; on the surface the design seems superior to me.

Yeah, I realize the advantages of their design, it just doesn't seem worth it with the added complexity and (I would guess) reduced maneuverability. But I guess maneuverability isn't a huge concern, since it's designed to basically bounce around the environment.

A quadcopter would require a larger radius protective sphere. The coaxial design is more compact.

Awesome concept.

But i wonder what if it went into a cave with icicles/sharp thin sticks that can perhaps go through the holes of the protection.

This is ridiculously cool.

How does no comments + 10 points get on the homepage of HN. System makes no sense.

Frequently threads with as few as 3 points make it to the front page. It's (among other things, obviously) about speed of upvotes on /newest- and given that posts rarely get more than a couple upvotes before they're knocked off the front page there, it's a good thing the threshold before things reach the front is low.

Also upvotes by YC peeps, probably have 3 times more weight than normal ones.

Looking at that video, I am inclined to say that this system(HN system) works very well. Almost like magic ;-)

IMO, a high upvote/comment ratio is definitely a sign of a good article.

Reply on news.ycombinator.com