Hacker News Reader

Comment by gdsdfe | Hacker News Reader

so is this like the only programmable drone out there? or is this the only one that is open source ? how much something like would cost? anyone have an idea ?

View on HN

Crazyflie is a tiny flying PCB, original version from ~2013 (no camera) https://www.bitcraze.io/

And yes, it is not very robust, I broke one PCB arm but could fix it without affecting the balance too much.

The bigger problem was encountered when it flew out of reach. I had an xbox 360 controller connected to a xbox wireless dongle on my laptop, running crazyflie SW, connected to crazyfly via another dongle. Tested it in a city park (somebody attended to the laptop).

I moved a bit away from the laptop setup while trying to learn to fly it (manual leveling is hard). I don't know which wireless connection failed first. But the result was, it just kept running the last instruction, and flying up, up and away. Luckily the battery failed when it was up about 50-100 meters, and had drifted away about 100m horizontally due to wind.

I was sweating like hell, picturing scenarios of it coming down on a car windshield and causing a surprised driver to crash, etc. Luckily it came down within the same park. Without new damage to drone, or third parties!

Lessons learned: Don't fly an indoor drone outdoors, it's not built to resist wind and doesn't have SW safe for it. And never, ever fly any drone outdoors without model plane liability insurance.

> And never, ever fly any drone outdoors without model plane liability insurance.

Or be able to run fast ... ;-)

The ArduPilot software platform is used very commonly across the drone industry and this is definitely not the only programmable drone available! The ArduPilot site has a list of some ready-to-use drones that have ArduPilot installed: https://ardupilot.org/copter/docs/common-rtf.html.

A lot of the time ArduPilot is installed with a flight computer (separate from the flight controller) which sends control signals inline with the RC receiver and can be retrofit onto an existing drone.

You can also pick up pretty inexpensive kits to retrofit any DIY quadcopter or RC aircraft with autopilot capability. Two examples are the older APM2.8 hardware (no longer supported, dirt cheap) and the Pixhawk PX4 (actively supported, more expensive but still <$200.)

Ah so if I understand this correctly : on the software side I'm most probably gonna land on ArduPilot, and for the hardware that's were the bulk of decisions will be?

Like bri3d said in his comment, there are a few options -- ArduPilot is heavily focused on mission planning (Waypoints, GPS guidance, etc). There is other flight controller firmware like Betaflight which is more focused on the stabilization and PID control of a quad.

There's definitely a lot of different decisions you can make on the hardware side: Choosing what sensors you include, flight computer & flight controller, ESCs and motors, frame, and so on.

On the other hand, you could also choose to build a drone from a kit or ready-to-fly model, and experiment more with higher-level software. ArduPilot can be connected to ROS, which is the standard software platform for robotics. Much of the swarming behavior, visual tracking, and other experiments are done on a platform like ROS, which then sends waypoints or other instructions to the flight computer.

Some areas to research if you're into this:

ArduPilot (Firmware handling waypoint missions, guidance, etc. on an onboard flight controller): https://ardupilot.org/copter/index.html

MavLink (Standard serial protocol used to communicate with drones, used to communicate between an ArduPilot controller and a ground station): https://mavlink.io/en/

ROS (Software platform designed for robotics, commonly ArduPilot drones feed into a ROS system for swarm experiments etc.): https://www.ros.org/

No and not even close. Look up Betaflight (racing oriented) and iNav (navigation oriented). Ardupilot is more integration/mission oriented, but it's far from the only game in town.

A small <250g drone is in the $70-$100 price range, although you will also need a $70+ radio transmitter or some heavy hacking to control it. Most of the (hundreds) of supported control boards are not open hardware in the schematics sense, but some are.

The all-in-one PCB integration strategy is cute, but I really don't think it's a great approach from a crash-recovery point of view.

> The all-in-one PCB integration strategy is cute, but I really don't think it's a great approach from a crash-recovery point of view.

Spot on.

However, it's a nice way to lure you into building a quad: just put some motors on, and you're good to go.

Fly once, crash, and then go actually build a proper quad.

Small quads generally do remarkably little damage to themselves in most typical crashes. Remember this thing only weighs 102g, and half of that is a battery held on by springs.

In typical indoor use (can't get more than a few meters high, unlikely to go faster than a few m/s), this'll crash way more than one time before it show any signs of damage to anything except the (replaceable) props. (Sure, you won't be handing it down to your grandkids in your will, but it's not as disposable as "fly once, crash, throw it away"...)

I totally agree with you, it's the most compressed form, in a slim electronics layer, for the most lightweight design. Every grams are used for the best selected components, assembled with passion and fully supported by Ardupilot firmware.

I agree -- and beyond a catastrophic crash, I wonder how practical the PCB frame is for longevity of the quad, vibration dampening, etc.

Without catastrophic crashes, it's a flying PCB with the finest micro drone's build integrated in its frame.

About vibration dampening, if you look at the maiden flight here https://discuss.ardupilot.org/t/ardubee-a-ready-to-fly-micro...

Don't you think the untuned ArduBee was flying already nicely?

The graph of the postfilter acc and gyro is not so bad, the new Dynamic Notch Filter technics in Ardupilot is working so good. Furthermore, the best in his class IMU is put in the center of the frame, where the single 18650 battery, the most of the mass, is fixed in place.

I don't know much about the drone space to be quite honest but I always wondered what it would cost to mess around with swarm algorithms, on actual drones, without breaking the bank. Any pointers?

IMO, a lot. Moving things are hard, and drones especially, are _very_ damn hard. One mistake and the drone goes poof! It gets frustrating after a while.

Just some pointers if you want to do this:

- you need some companion computer on your drone.

- GPS alone is not precise enough, you'll get wierd drifts from it all the time, you want something like RTK GPS to be able to keep stuff together.

- If you want your positioning to relie on vision, get a global shutter camera, a rolling shutter will make your life harder than it needs to be.

With ArduBee initiative we are making our best to innovate creating our best solution for swarming, with micro drones, indoor and even outdoor, based on Ardupilot Open Source project. If you have a look at https://ardupilot.org/copter/ you could start to have an idea of the included potential.

Speaking of transmitters, what is a good entry level transmitter/receiver combo to get into for hacking servo-based RC projects? So far I like a Spektrum DX6 controller best I just wish it was a bit smaller for my kid's hands.

Look up the Jumper T16 - it’s a multiprotocol transmitter that supports most of the major brands, which means you’re not tied into any brand receiver.

The Jumper T176 is amazing, if you're going to get a good TX you want to last you a while, get that. If you want a cheap, small one with not too many capabilities but that a child can use, get one of the cheaper FlySky ones.

I've never built a drone yet but I've tried to do a lot of reading, and I don't understand why everything assumes that you have an R/C transmitter in your hand, and then at some point "throw the switch" and make it autonomous.

Isn't the entire point of these things that they fly themselves? That's why we call them drones and not R/C helis, right?

Why can't I just click "go" on my GUI, and never purchase a TX? Or can I and the distinction just isn't explained in a place that I've found it?

That's because Betaflight and INAV have evolved from manual/freestyle flying. ArduPilot is the more "autonomous" platform, which you can fly without a TX. You can add a USB telemetry module to your laptop and fly like that, although I've never done (and wouldn't do) that, there are many cases when you need manual control for recovery/failsafes.

I don't know where you got the idea that autonomy is why we call them drones. FPV drones are all analog and they've been around a while.

They assume you have a transmitter, because a majority of non parrot drone users are flying analog.

Yeah. There was a time maybe 8-9 years back where at least a section of the hobby tried to push back against the people using the term "drone" for any quadcopter/multirotor, and trying to educate people that only things capable of autonomous flight were "drones", while regular non-autonomous quadcopter really are just RC helicopters.

We lost that fight a long time ago.

(And I'm not even sure we were "right" to be honest. The term "drone" got used back in WW2 era for radio controlled aircraft used for target practice. There sure as hell were not autonomous... https://en.wikipedia.org/wiki/Radioplane_OQ-2 )

What drones are autonomous? The most well-known ones, used by the US military, aren't autonomous, AFAIK. A in UAV stands for Aerial, not Autonomous.

At least in the consumer space (which I suspect is what you're thinking of when you say "assumes that you have an R/C transmitter in your hand"), most people over the last decade (and even now in a lot of places) are flying under some interpretation of 50+ year old model aircraft rules.

Here (.au) I'm technically not allowed to fly FPV (where I'm watching a camera view from the air thru goggles or on a screen) without having another person ready to immediately take manual control who's watching/flying "LOS" (line of sight) and complying with the regular model aircraft pilot rules.

Practically nobody actually does this, but almost everybody holds some form of manual controller and is ready to take their goggles off or look away from the screen and fly the drone manually as a regular model aircraft.

If you _want_ to "just click go on the GUI", at least some of the DJI stuff will allow you to do that. I've got a DJI Spark that I can connect/control via wifi from an iPad, and software that lets me define a mission and just click fly without needing a controller. There's still an advantage to having the controller though, it's radio as way better range than an iPad's wifi, the wifi drops out and I lose the video link (while the drone keeps flying its uploaded mission) at a hundred or so meters. If the drone and the iPad both connect instead to the DJI controller, I'll get reliable video at well over 1km. One other reason I almost always use the controller is I'm much happier launching/landing in tight locations if I fly the last few meters manually. I'll sit on the back ledge of the car boot and land 1m away from the car flying manually, but I don't trust the drone/GPS quite enough to do that, and will always find a fairly large (at least 10mx10m or so if I can) area to let it land in if it's flying totally autonomously.

To be blunt: because they're not good enough. Even ArduPilot which is probably the best mission-flying software in the open source space, or DJI who are still ahead in the commercial space, will often need manual intervention to complete a mission.

You absolutely _can_ build an ArduPilot drone that takes commands and tries to fly a mission end-to-end without intervention, but without a real-time TX link of some form, you are in hot water when it fails.

I don't remember what was the price range but it was cheat. I think we speak about it in the end of their talk in the ArduPillot unconference. Check on ArduPilot YouTube channel for their talk.

What is impressive it the integration on this drone : that give a good flight time with low weight. That is a big advantage against current weight based regulation in numerous country. And of course ArduPilot integration give hight capabilities !

Excuse my noob question but : what do you mean by integration? And why is it important?

It probably means you can build firmware form ArduPilot and burn it into the drone. It's a big deal because ArduPilot is a whole platform and they have a lot of bells and whistles, communication with a companion computer, base station communication, a whole lot of possible configurations.

integration of all the different parts on board in a single base: WIFI with ESP8266 WiFi, brushless motors, 4x ESC last generation BLHeli32 with telemetry and DSHOOT up to 2400 -- FlighControl -- IMU -- baro -- compass -- power system with 1 easily swappable Li-Ion 18650 for maximum flight time -- rangeFinder for accurate low height navigation 0-3m -- sdcard slot for Log -- connector for common RC radio and telemetry + connector for all the external add-ons like GPS, indoor positioning system, optical flow, 8 horizontal rangefinder for collision advoiance,... All these hw connected with the Ardupilot features. Furthermore, for small copter its very hard to have good flight time, ArduBee try to have very good flight time, safe and ready for makers to develop, supported by one of the best Open Source project for drone, Ardupilot. It has been developed to be used safely for urban swarm robotics application development, education, creativity!

https://ardupilot.org/copter/

ESP8266 with Ardupilot (telemetry and control) https://ardupilot.org/copter/docs/common-esp8266-telemetry.h...

BLHeli_32 ESC Telemetryhttps://ardupilot.org/copter/docs/common-dshot-blheli32-tele...

A 5" ArduPilot quad will probably run you about $150, maybe $200 tops, depending.

It's not so easy to stay under $200 if you mean all parts apart the RC radio. I mean 4 motors + 4 ESC BLHEli32 with telemetry + FC + WIFI, compass, 1 down-facing rangefinder, .. in particular for the FC if you select these components with quality chips and sensors it starts to be quite challenging. Furthermore, if you would like to develop swarm applications the time spent to create the fleet is not to be underestimated. Last plus its a platform with 1 single common 18650 LI-Ion battery for higher flight time in comparison to the usual LiPo.

Yes, it depends on what kind of sensors you get. I meant just with GPS, 4 BLHeli ESCs will run you around $40, 4 motors another $40, Kakute F7 controller another $50, $30 for the frame, $8 for the GPS, $10 for the receiver, $5 if you want an extra PDB, it's not that bad.

If you add sensors it does add up though, you're right.

Reply on news.ycombinator.com