The software gimbal is a very nice trick: instead of moving the camera (with jerkiness and balance and extra control channels), they have a slightly down-angled 180-degree fisheye lens onto a big sensor. Then software stabilises and crops to present a slice of the view as if you were panning and tilting. Slick.
I've often had problem with stabilizing shaky GoPro footage where the footage has geometric distortion, and after the stabilization it gets very noticeable: as the image is panned around, at the edges it appears wavy and flowy.
A workaround for this is to account for geometric distortion when stabilizing. For example, in VirtualDub, use barrel distortion filter to "un-distort" the image, then do the Deshaker pass, then use barrel distortion filter to get the fisheye look back. This is a manual, error prone process, it takes time to get filter parameters right and to make sure no edges come into the view etc. Too much hassle if you want to process many videos.
When stabilization happens in-device, they can tune it for the exact lens used, and they can work on raw video data, before compression. They can also account for rolling shutter and avoid the "jello effect". Plus there is one less post-processing step afterwards. So this has good potential and I'd like to see in-device stabilization in more action cameras too (Sony actioncams do have steady shot, but they also have some other shortcomings--lower max bitrate than GoPro, kind of bulky with the waterproof case).
It is a nice trick, and the footage seems pretty solid. Will the quality match a real brushless gimbal with a good camera attached? I don't know, but for the majority of people this will do just fine I guess.
A twenty-degree down very high-res 180-degree fisheye isn't a very useful output. A distortion-corrected slice is much more valuable.
As for bandwidth, generally these things record best video onboard and the transmitted stream is allowed to get glitchy with distance / interference / alien jamming.
I've built several "hobby grade" multirotors with GPS and first-person video, and while I find the process fun and satisfying, it's far from accessible to mainstream consumers. I was never impressed by Parrot's original AR Drone, but this looks much more promising as an accessible drone with a good feature set. The biggest problem with the AR Drone, the control latency over WiFi, is probably covered up by the Bebop's GPS stabilization.
It doesn't look like there's any pricing information for the Bebop yet. They should definitely be able to keep it under $500, and if so (and if it works as well as their demos portray) I bet they'll sell well.
For another multicopter in the $500 price range, check out http://www.thepocketdrone.com - the most successful drone Kickstarter of all time (raised almost one million dollars). Disclosure: I work for them.
Shipping is due to begin late summer, but Kickstarter orders will be shipped first. I've been told there is going to be a price increase sometime this month, although I'm not privy to the exact date.
Your website is a bit withholding for my taste.. clicked around looking for a decent flight time approximation (vs. weight, conditions etc.).. Found nothing.
Are the blades proprietary? If so how much will they be?
I have quite a bit of quad/hexa-copter flying and the #1 problem I have had is replacing blades. Luckily my more expensive copter I have uses the same blades as RC planes so they can be ordered cheaply, but thats not the case with my other drone. I can foresee a huge headache for your users if the pocket drone uses proprietary blades.
I thought I read either in the comments the last time this was posted here or in the original article that they were aiming for $6-700. That seems right for the technology
I had a parrot drone before. It went up to 30 feet, disconnected from Wifi, dropped from the sky, and when I was just about to catch it, it re-connected, turned and crashed into pieces.... Ever since then, I am not a very big fan of their technology. For slightly more, you can get much better drones which would be worth the investment.
The first AR Drone was released before the tech was ready for an out of the box consumer experience, especially for outdoor flight. The technology has come a long way in a few short years, especially for hobby grade gear, and the Bebop looks pretty solid in the few demo videos I have seen.
My AR Drone had busted battery on arrival. But what really had disappointed me was very flimsy styrofoam construction. Looks like this version also continues to have styrofoam.
It's not a feature, its cost savings. You will never see good quality drones with styrofoams. The styrofoams makes drone lighter so that battery can be also cheap and manufacturer can claim bigger number for flight/charge. One advantage of styrofoams is that you can change it cheaply it breaks but in reality styrofoams makes drone much less stable because even small wind can now throw things off. It also makes it much more fragile.
The DJI Phantom is pretty much the standard in consumer-grade multirotors. It's plug & play, sold as a complete package with transmitter included, and for what you get it's actually not that expensive. If you're reluctant to fly a $500-$1000 UAV, get a cheap multirotor first and learn to fly. The Hubsan X4 is under $50, for instance.
If you're a tinkerer you can build your own for a lot less, but be prepared for a LOT of fiddling to get it right (and the crashes that come with the fiddling). I fly a DIY F330 (same dimensions as the Phantom, only 'naked' frame) with an Arduino-based flight controller (APM2.6). It has some advantages over the DJI system, such as unlimited waypoints for autonomous flight. It can carry a gopro, and replacement parts/batteries are a lot cheaper if you get them from China. I don't get anywhere near the battery duration of the DJI, though.
So: if you're looking for a new hobby: build your own. If you just want to fly: get the DJI.
Couldn't agree more. I had a Parrot 1.0 and it was of low overall quality. I have a Phantom (v1.1) and it's fantastic; a friend just bought the v2 which is even greater.
Among the differences: the Phantom has an actual radio, with ranges of over 500m (it would be very surprising if Wifi covered the advertised 300m of this new Parrot); you can change cameras as you wish (or, after a crash), it can fly really high and really fast, etc.
You can get a Phantom 1 for under $350 these days (without a gimbal), or a 2+ (camera and gimbal included) for around $1000. I would recommend buying the 2 without gimbal ($550) and the Zenmuse gimbal ($150-$200) -- and if you already have a GoPro you're all set.
Alternatively, you can buy a Hubsan H107D FPV which comes with a radio with a screen, for under $150! I don't own that one, but I have the Hubsan X4 which is a lot of fun (and completely safe to fly indoors).
If it is possible, you would be severely limited by the WiFi signal radius. I'd highly recommmend a normal 2.4 Ghz TX since they're designed for this stuff. The phantom comes with TX, but if you go the DIY route I can really recommend the Turnigy 9X from HobbyKing. It's only $60, has 9 channels, comes with receiver and people won't laugh at you because you're holding an iPad or a toy-like transmitter.
For learning how to fly in the sub $100 price range, get a Hubsan X4 H107C - don't be put off by the low price - they are crazy fun to fly. For the $1000+ price range, the DJI Phantoms are great. For the $500 price range, check out http://www.thepocketdrone.com - the most successful drone Kickstarter campaign of all time (raised almost one million dollars). Disclosure: I work for AirDroids who manufacture the Pocket Drone.
Before you even start, have a look. You'll need a frame kit, motors, ESC and a flight controller as well as a transmitter and a reciever. You might also want an FPV kit so you can see where your flying and a GoPro so that you can film what you see.
Have a look at the DJIF450 or DJIF550 frame kits, start with one of those cost me around £400GBP, you'll learn more with one of those than a Phantom and you'll be able to replace parts easier.
Although this is interesting, I think DIY multirotors are where the revolution really is. I've built my own and it really isn't that difficult, assuming you're buying a kit for the frame and not manufacturing your own it's hardly more involved than assembling a custom PC and can result in a drone able to carry a camera such as a GoPro for less than $200. If you are planning to do this though I'd recommend getting a small, cheap quadcopter like the Hubsan X4 to practise with first.
I have a Phantom and a Hubsan; I upgraded the Phantom with a new motherboard to install a gimbal; but I wouldn't have a clue of where to start for building my own from scratch?
The $200 figure sounds optimistic since the frame and the radio each cost around $100, and then you have to add the motherboard, flight controller, motors and props...?
Would you mind establishing a shopping list of what you built, and where you bought each part? I would love to give it a try! ;-)
- ESC's x4 (Turnigy 30A, $9.99, flash them with SimonK)
- Flight Controller (KK2.1, € 29.99)
- Transmitter + receiver (Turnigy 9x, € 59.99)
- Props (2CW, 2CCW, $2.46 per 2. Do get extras!)
- Battery (2200Mah 3S 25C, $10.57)
- Charger (Turnigy Balance charger, $22.95)
Total cost: a couple cents over $250. You can do even cheaper when shopping around Chinese webshops, but this should give you a pretty good idea. All of these parts are from Hobbyking, by the way.
(Edit: formatting)
If you're looking to get into this hobby I highly recommend a number of things instead of just putting those items in your shopping cart:
1. Instead of buying a cheap, easily-broken "flamewheel" frame (e.g. F330, F450, etc) get something that's much more difficult to break like a One-Piece Quad (OPQ) frame (http://witespyquad.gostorego.com/opq-one-piece-quad-frame.ht...) or learn how to make your own frame out of thin-walled PVC (which really is super quick and easy if you have a heat gun--takes a lot more time using the boiling water method; spend $20 and get a bog-standard heat gun). If you build your PVC frame with a "T fitting" field repairs usually take just two minutes to swap out a broken arm (flamewheel repairs take forever in comparison).
2. The Turnigy 9X Tx (aka "remote control" for the uninitiated =) is pretty good but the FrSky Taranis (~$200: http://www.alofthobbies.com/frsky-taranis-x8r-combo.html) is vastly superior. You will never "outgrow" a Taranis. It runs the open-source OpenTx firmware (infinitely hackable), supports 32 channels plus a new bi-directional communications system named, "Smart Port" (support for which was hacked into MulitWii by yours truly). Before the Taranis, to get a Tx that supported even 12 channels was a $1200-2000 investment. Not to mention one that does everything the Taranis can do (logging to SD card, audio feedback, jillions of switches/knobs/hats, etc). The Taranis has basically made every other Tx obsolete and every Tx over $200 is now a huge waste of money.
3. Get at least one LiPo "charge bag" (~$2). Lithium Polymer batteries, sitting by themselves on the shelf, hooked up to nothing, will occasionally EXPLODE INTO FLAMES THAT CANNOT BE PUT OUT. The chances are small, yes, but those chances go up every time you crash or if you let the charge of the battery get too low. They also go up every time you charge it. Stay safe: Charge your batteries in the bag and also keep them stored in the bag when not in use.
4. Always be aware of how much your quad weighs when flying. A 450mm frame outfitted with a GPS, 2200mAh battery, and a camera can weigh 5-6 pounds and yet can fly well above 400 feet (actually ~5x higher but don't do that!). It's enough to kill someone and/or do very expensive damage to a house, car, or whatever it happens to crash into. Always assume that at any point during a flight any ONE component could fail resulting in "death from the sky": A motor, a propeller, a speed controller, the battery, a connector, the flight controller, the receiver, the battery, the frame, etc. All it takes is just one of those components to fail and it's DOOM.
Admittedly for $200 you aren't going to be getting anything fancy, but you'll certainly get something that gets the job done. Since this was my first RC project I went with a $23 radio system, which for basic control is fine. You can get a decent frame for $60, I went with a wooden one since it's pretty sturdy and unlike something like carbon fibre, if you crash you only have to go to a DIY store for replacements. The rest of the parts added up to $120 and were from HobbyKing. At the moment I have mine carrying a Mobius camera which is great, but it could easily carry a GoPro too. I have a blog post about it if you want more detail:
I'm curious how easy is it to write software to control a multirotor you built yourself. One of the big appeals of the AR drone was a fairly easy to use Linux SDK.
Have a look at Rotor Bones from Flite Test
https://www.youtube.com/watch?v=x_bDJ-esG4A
Its a wooden kit you can buy, then have a look at Motors, ESC and a flight Controller and Tx/Rx you can pick them all up on Hobbyking.com.
If the aim is to reduce cost, is it wise to split the drone and the camera ? In a comment to this one, the DIY drone is around 250$, the GoPro itself is 200$, the Bebop drone (whose camera actually use the same sensor than a GoPro) includes drone and camera (allowing to use flight information to improve video, and video to improve flight control), cost does not seem to be the right reason to make DIY drone. Technical interest may be relevant, as well as specific requirements, but cost is rarely a benefit from DIY.
My comment isn't neutral, as I work at Parrot, but I think it is honest. I also make my comparison with an expected price of 500$ for the Bebop drone, but I have truly no idea of what the actual price will be.
[...] Bebop Drone navigation computer features a Parrot P7 dual-core CPU, quad-core GPU and 8 GB of Flash Memory[...]
was intriguing. I really don't believe that they really have a custom "Parrot P7" CPU since that makes very little sense considering their use for the CPU. Also, even if they had designed their own CPU, why would they include a quad-core GPU for this application? That doesn't make any sense, either.
Does anyone know more about that CPU? I don't follow Parrot closely so perhaps I'm misjudging them.
I worked on the development of the P7, and yes, it is a dual core CPU powered by ARM cores with a quad-core GPU. The CPU is meant to be used for other projects too, so a quad core GPU makes sense.
I guess I kind of assumed it was an ARM, that does make it feel a lot less magic since (from my understanding) designing a SoC based on ARM is quite far from the same as designing a CPU from scratch in terms of complexity.
I wish they had included the word "ARM" in their description.
They've been releasing SDKs for their drones for many years. Although the PR material doesn't make it obvious if you get a chance to download the SDK it becomes very apparent.
My last AR.drone was ARM based with RTLinux as the kernel, not sure if it is the same now. You could SSH directly to the onboard OS and into Busybox. I was able to modify the wpa-supplicant so it could connect to the house wifi and be controlled from my PC, instead of using the crappy iOS client. From the SDK I wrote a small .Net app that used my Xbox controller, gave me telemetry, and recorded video.
One of my startup ideas couple of years ago was cheap stabilized mini dolly/crane system for casual videographers. This space, I thought, was ripe for disruption considering tons of new professional grade cameras sporting HD video but other equipment remaining as expensive as ever.
Shooting with this kind of system really makes even bad video interesting.
Now I think drones could be just that dolly/crane system. I hope people make their control software open source. The key would now be software that can produce super stabilized flights in artistic trajectories automatically.
That was exactly what was going through my mind as well.
I just make short films as a hobby, and I've been so excited over the years to watch these huge quality/price inversions for high quality video. There's still a ways to go, but the movement has been amazing; it's been thrilling.
And now this?! There are so many shots I have in my head that I could pull off with this. The idea that I -- I alone! -- could essentially afford my own dolly/crane system blows my mind.
I feel like I have just seen the future.
As for the software you mentioned: it's clear, the stability has to get better (but there was clearly a lot of wind on that beach, as you would expect, so maybe it would be better in different conditions). But for the trajectories, you could also contract someone that worked remotely, to whom you could draw out the trajectory you want or model in another way, and be your camera operator. (Wouldn't that be mind-blowing? "The camera operator on this film never left their home in Madagascar.") Really -- I've just seen the future, and as a (hobbyist) film maker, I'm ecstatic at the potential!
This with the Oculus Rift combined is going to make for some seriously awesome "tourism demos". Someone has to make a Bebop-as-a-service thing where you have warehouses full of Bebops across the world's largest tourist centers, then a paying user will have his/her Rift linked up with that Bebop, which will then fly high above the city's ground and let the tourist have a look at the city from the comfort of his/her home.
Wouldn't make much sense combined with the Rift though, as the camera on the Bebop can't move. You could move the Bebop itself based on head motion, but i am not sure if that would give a good feel of control. Besides you would still be quite limited in wifi range and i doubt people want drones falling into the pools or on their heads ;)
The camera has a very wide handle so you can actually move the visible window by simply adjusting the software cropping. So you can actually track the head movements without moving the drone on a roughly 180degree angle in every direction.
There's only one camera though, so you don't have any stereoscopic effect.
I had a Parrot AR.drone. After one too many crashes the motherboard wouldn't signal the motors to start. I sent it back to the store I bought it from, who promptly gave me $300 in store credit rather than another drone. Although I would rather have had another drone. When I asked the CSrep she explained they stopped carrying them because the high return rate. My son and high had a lot of fun flying it indoors and I'm looking for another.
I'm fascinated by this drone[1]. Variable pitch rotors means it has the agility of a helicopter, including inverted flight (upside-down). Only downside is there is 1 motor so 1 single point of failure.
The Stingray is awesome, although I wouldn't trust myself flying it until my skills are much better (Curtis Youngblood is a beast). The motor as a point of failure probably isn't a huge deal, because those motors are themselves very reliable, it can autorotate like a traditional helicopter, and a traditional quadcopter can hardly survive a motor failure anyway (although software should be helping that soon).
This is amazing. With some camera smoothing in post production you could have shots that were previously only possible with a helicopter shot. I can see this being just as influential if not more than the gopro.
yes, but many drones like the DJI Phantom already do that, and better than AR Drones and possibly the Bebop. They also feature GoPro Mounts but are obviously more expensive if you add the cost of the GoPro.
The big innovation here over comparable camera-stablizied GPS-locked drones (eg. DJI Phantom II) is that there is no mechanical gimble which significantly reduces weight and battery consumption of the Bebop.
It's a great innovation, but I still prefer my mechanical gimbal. It offers a lot wider angles and I won't have to crop my footage. When I don't need it I'll just leave off, the reduced weight makes the drone ripping through the skies like a ferrari.
But a mechanical gimbal is awesome; it can rotate to a horizontal position, so you can get satellite-like views, or simply look at the top of something; with a fixed front camera you will never be able to get those kinds of shots.
I'm more interested to see a small drone that can carry a small load, say of a couple to a few pounds. This could start to disrupt the small package/food delivery industry on a massive scale while cutting emissions and increasing service quality. We have seen plenty of small drones meant to be used for surveillance; it's not very interesting anymore.
I actually have an idea of indoor robotic drones inside fulfillment ware houses, that can automatically pick items as orders arrive (like Kiva systems, but with drones). These drones will need capability of carrying couple of pounds.
Current law doesn't allow commercial drones - meaning nothing that makes a profit or supports a business. Hobby photography is the only application that makes sense right now.
If you do your homework or even a little googling you will realize that what you are saying isn't true. Hopefully this is good news for you! I think the link below may have even been submitted to HN, though I haven't checked.
What law are you referring to? Columbian law? Swiss law? Russian law! I really have no idea, being that I am from Earth and there are so many countries. It does make this thread hard to follow, just too much guesswork.
Owning an AR.Drone 2.0 the thing I most hate is using their mobile app to control them. The Bebop looks small enough to make it worth using my mobile to control it. I can see why the introduced the Skycontroller which looks pretty interesting. Saw a guy build something like this for his homemade drone.
I'm surprised no one has mentioned the IRIS by 3DRobotics. It uses a brushless gimbal to provide the image stabilization and is made largely of plastic components that can be replaced easily. I have their hexa-copter (built from a kit) and have been very happy with it.
Two things worth pointing out that I think are very important:
* The IRIS uses RC plane blades so they are cheap to replaces
I think this has the potential to be the next big thing. Not necessarily this particular device, but the idea itself.
There are lots of applications for these types of devices.
- News reporting
- Surveillance and recon
- Security (couple it with a wireless charger)
- Entertainment
All of these can be broken down even further into different industries etc.
Did anyone else catch the "distorsion" typo in the video? What is it with typos that makes you want to point them out? Is it that important to us (me) to feel superior? Or is there a tiny bit of altruism hidden somewhere...
Parrot is built by a French company. In French, distortion is spelled distorsion. You should not look any further for the reason of the mistake, which is (very) hard to spot for a Frenchman.
Sure did. At least for me it's about identifying the error to the editor especially in the case of a business selling a product and trying to cast a sense of professionalism. Spelling like that just taints their image IMHO.
