Hey guys, We built a quadcopter with makers/hackers in mind. What do you think?

[James_skyworksas]

Hey everyone, long time listener here, finally made an account to promote a new product from our start-up. (I hope thats okay) If this should be moved please let me know a better place to put it.

Anyways, Skyworks Aerial Systems is proud to announce a new product and we would love your feedback. You can check out the kickstarter here:

https://www.kickstarter.com/projects/dronesmith/eedutm-an-easy-educational-drone-kit-for-learning?ref=nav_search

But the main idea is that it is an easy to build quadcopter that has a Intel Edison onboard with connections for an Arduino shield. We are very excited to see what hackers/makers like yourselves will do with with the platform and are eager to hear your thoughts and ideas.

[cooper]

Sure, I'll bite.

I understand the Intel Edison isn't a Raspberry Pi but there are more suitable boards than a Pi out there to power a drone. Why did you pick this one, and what do you plan to do next year when you're about to release your product and the next Edison is released? You say you're working closely with Intel, but for how long do they guarantee availability of this specific platform?

Now, I know next to nothing about drones, but I'm pretty sure those rotors don't run off of 3.3-4.5 volts. You're going to have to have a converter circuit to take the rotor's battery power and use it to also power the board. While there are (far) less power hungry devices than a Pi2 out there, with the Edison admittedly being one of those, it will pale next to the power requirements from the rotors, so for this application the power consumption of the controller board, while not entirely devoid of importance, doesn't matter all that much really.

Let's make a more interesting comparison and pit the Edison against the Orange Pi 2:

CPU: The 1.6GHz quad core A7 runs circles around those 2 Atom cores, even with the Quark one thrown in which Intel isn't doing just yet.

RAM: Tie.

Connectivity: Tie - the Edison has bluetooth, the Orange Pi 2 has more USB, both have wifi using that shitty antenna connector, both have a stack of pins to play with and the extra shit on the Pi is just that much dead weight for this.

Power: Tie - The Edison lasts longer, but what's saved here will keep your drone afloat at most a second or 2 longer so even though the Edison wins the margin isn't enough to pick it over the other just for this.

Dimensions: The Edison is considerably smaller across all 3 axis which might matter considerably for your drone. Worse for the Orange Pi 2, the connectors that make it high have very little use when applied to the drone.

Programmability: The Orange Pi 2 wins this one easily. If you want to get the most out of your Edison, you're into heavy duty programming territory and since you want to use this drone as an educational tool of sorts, the Edison kinda sucks.

Cost: The price of an Orange Pi 2 is less than half the price of an Edison.

So please elaborate: why the Edison?

Edited June 1, 2015 by Cooper

[James_skyworksas]

Hey Cooper, excellent question

For the motors, there is a complicated electronic speed controller (ESC) for each of them, typical with brushless motors. This is built into the main motherboard where the Edison also plugs into, as well as the flight controller processor and IMU.

The Edison board is a lot more compact. The size is a big issue when trying to embed systems into a production product. It is also significantly lighter without the unnecessary connections the Pi2 has. Bluetooth is also conveniently integrated in. Quality is also important to us and Intel has better support for $10 more. The Edison is going to be supported for many years.

The program-ability of the Edison is also easy to do with the software (named Forge) we have developed to compliment Eedu. This will also be free and open source so that hackers can help build functionality into it. For the typical user, it is an easy to use high level programing environment. Our goal is to enable anyone with little programing experience to get started in Forge and programming drones.

[cooper]

Thanks for the useless comment on the motors. My uncertainty was about the voltage requirements of those motors specifically because you're going to get 1 power source for the whole thing and you're probably going to choose a power source such that it can feed the motors primarily, and convert this up or down to whatever the controller board needs. I believe but don't know for a fact that when you convert up or down, the higher the difference between the voltage IN/OUT, the higher the loss in that circuit.

You claim the cost of the Intel Edison is $10 more. I compared it against an Orange Pi 2 which is $35. Are you seriously saying you can get an Edison for $45? Because the prices I see mentioned are more like $70-80. Or are you now leaving out the breakout board which would make the Edison a lot less useful?

On your kickstarter page you claim that you need to compile your program directly on the device itself. While it's nice that you can, when you need to recompile your component 60 times to iron out the bugs the time wasted on that slow as hell CPU is going to be seriously off-putting to your target audience. Surely you can find a way to cross-compile or something?

You choose the lesser hardware capabilities of the Edison in exchange for the smaller device footprint and to reduce weight, which is a fair design decision to make.

The problem I have with this project is that, like half the planet, you're developing yet another drone. You claim yours is different because you can do more than simply plug a camera into it, but any platform that has a modest computer on board can manage that. So what's the unique selling point of your drone that should convince people to go for your kickstarter campaign?
Your drone costs $450 now as part of the early bird promotion. After that it'll be $500. For $450 I can in fact get a 6-arm multirotor platform that can trivially house a Pi-sized board and I personally think looks a lot more attractive. And I can get that one today. So why invest in your kickstarter now to get a similar drone at the end of the year? And it might be just me, but $500 for an educational toy is a tad excessive in my book (yes, I don't have kids, but my sisters do).

If I want to teach kids about sensors, I'd combine a sensor with an Arduino and be done for under $150. If I wanted to teach kids about drone flying, I'd get one of those hunks of Chinese plastic for $50 that lasts a year even when you don't crash it. If I wanted a smoke sensor, I wouldn't place it on a drone platform. Rumor has it those rotors displace air around them... So why are you so hell bent with viewing the drone as a useful sensory platform? Why not go for a boat? Or an RC car?

Mind you, I'm not expecting you to change your product. I'm just hoping you can explain why you made the choices you made as they don't make a heck of a lot of sense to me.

[metatron]

There has been a massive spate of educational programming toys coming out the last few years, a few car ones already exist. I can understand the need to keep weight down as motors and batteries aren't cheap, and it's easy to build a board with regulators to drop voltages as necessary from a single power source and integrate a charge control for the LiPo battery.

The plastics you're using look well thought out (metal inserts for screws) and injection moulded, and look like they would be structurally sound, although I would be concerned with schearing, but I'm guessing you have done modeling in software and looked at the structural properties of the plastics in relation to your design, as well as looking at the thermal issues of having hot components, i.e. power regulators near the plastic?

I really don't think it matters hugely, what dev-board you go with but there is no point using anything more than an Atmel microcontroller, with the airframe you've designed as even if you've over specced the motors and battery, as an educational toy or even a home gadget, it's unlikely anyone is going to be running more than GPS and a handful of sensors.

With your own programming environment, you can make it very simple to use, you also open yourself up to the huge amount of existing Drone development out there and reduce cost.

Now if you are going for something like a Edison or rPi, then I'm guessing you want to do things like object avoidance and basic image processing onboard. Object avoidance can be done on much lower hardware and you are going to need to over spec the motors and battery if you want to carry a suitable camera.

If I were you, I'd design a single board solution, with everything integrated based around a ATMEGA32U4. Just have angled pin headers for the motors and pins for sensors. Sell it as a kit, with optional sensor packs. Do a cheap two part injection molding body, as all the sensors/breakout boards will all be sold by you, you can keep there shape universal and have say four slots integrated into the body. As for a camera, I'd probably just sell a 2.4GHz unit as an optional extra and just power it for a pair of pins on the board.

As long as what everything is, is written on the board it should be fairly simple and you can prices it as starter kits and onwards. Tooling for the body is the most expensive part in manufacturing, £17,000 to £32,000 I've been quoted.

Edited June 2, 2015 by metatron

[James_skyworksas]

Cooper: our flight control system is built by us from the ground up. Part of this included getting rid of the Edison shield and incorporating it into our main board (while also getting rid of things that were unneeded and taking up excess space) We are using an Edison as our co-processor for handling all tasks that are not realtime critical. This makes it so that the user doesn't have to worry about PID tuning (unless they want to) and leaves the Edison to do all the heavy lifting of the programs running on it so that flight worthiness (multiple failsafes) is never effected. Edison has excellent performance/watt throughput, integrated BT + Wifi connectivity, and a flexible yocto recipe, which makes it very attractive for our applications.

I’m not sure what you mean here. Compilation time is first of all, very minimal, because most of the existing components are already compiled and provided as binaries by our image running on Edison. Secondly, because it is compiling on the device itself, it will not hinder your ability to develop on Forge. You’re not compiling a Linux image, or flight controller firmware, or sensor libs from scratch. You’re only compiling the code you wrote on Forge.

As for your final point: while yes, this is just another platform to learn how to use an Arduino, we feel that with drones becoming an estimated 80 billion dollar market by 2020 there is huge rewards for becoming familiar with these systems. This means users will greatly benefit from not only learning what drones can do, but what they can't do. All of this while also being the people who will be entering the workforce during this great spike.

Metatron: You're actually pretty on point and have pointed to a lot of things we did, with only minor designer differences. As far as heat concerns go, the electronics have no foreseeable reason to have any issue aside from actual shorts (which are less worrying then if we had a carbon fiber frame) We do realize that that we probably gave Eedu more power than it "needed" but when we we're doing efficiency testing we found that the drop in performance was ultimately not worth "cheeping out" for. Same with processor, we came at it with the idea that we never wanted someone to complain that they weren't able to complete a job because the processor wasn't capable. Also having friends at Intel who had not only put the Edison on a rover kit but had also shown interest in wanting to see one on a drone helped.

Yeah, tooling is what it is, and since we do have more then 2 pieces (battery pack, bottom body, top body, and legs) those prices are a little short, but also we are trying to hire more devs to focus on the other part of our product (Forge) the community driven developer platform.

[overwraith]

I want one with wifi injection radios!

...

But since that one isn't going to happen any time soon, I have to say that I completely understand going for the board which you did, especially with the bulkiness of the USB ports etc on the raspberry pi. Regardless of how awesome putting a USB port on a drone would be especially with the injection radio theme, the majority of roboticists are going to be more interested in the sensors you can attach to the drone.

By the way, how many sensors can you attach to this drone simultaneously?

[James_skyworksas]

Hey Overwraith,

I've talked to my engineers, who have informed me that if you are using i2c to interface with the Edison you are essentially more limited by weight/size than anything else as far as connectivity goes. If however you would prefer to use an Arduino shield you would be looking at the standard 13 Digital and 6 Analog inputs.

As for your suggestion for injection radio's, we were a little more interested in what exactly you wanted to use them for. If its for packet injection type stuff you would have full access to that Edison computer to be able to do as you would like. The forge environment is just a more user friendly view/access to that world. This may of course require more hardware but that's kind of the idea behind Eedu. We never would have thought of this idea and no one out there is going to make a drone do what specifically what your asking...we do however try to make it easier for anyone to do whatever they want with a drone...which may be what your looking for.

[overwraith]

I know, not many drone manufacturers are going to be interested in the implications of something like my suggestion. The only reason I bring it up is that a while back on one of these forums there was a discussion of putting a MKV wifi pineapple on a drone, having it park on somebody's roof, and return to base when the battery gets low. So essentially yes, there would be lots of packet injection and sniffing etc while the drone was stationary. This might not be something you could pursue at this point in time, if ever, I just thought I would bring it up. There have also been discussions on "drone" type processing and tracking of individuals via MAC addresses etc. Would be fun to have a fleet of drones with these types of capabilities cruising around.