Dangerous flying car drone zoomed into UK's Gatwick Airport airspace after killswitch failed A “flying car” drone being shown off at Goodwood Aerodrome in England was using unlicensed spectrum for safety-critical radio commands – and when its operators lost control the craft climbed into airspace reserved for Gatwick Airport, an accident report has revealed. The homebuilt Alauda Airspeeder Mk.II crashed into a field …
Essentially, yes. If you squint a little.
On the other hand, while I've seen C5s re-powered with everything up to and including a jet engine, to the best of my knowledge nobody's ever been daft enough to actually try to take off in one of the things. There might be a reason for that ...
On the gripping hand, Sir Clive's li'l hotrod weighs in at about half of the thing in TFA.
Ohhhhh, and that's not all, folks! Mashable's[1] just been showing off footage from the manufacturers of this thing the Mk3 version. Yes, you heard that right.
So now my question is whether that Mk3 version includes the required changes the AAIB has raised... if anything, the AAIB will have told the CAA, and the CAA will be looking at *any* documentation and request for permission to fly the thing with exceptionally close scrutiny. As they should. Once burnt, twice shy and all that.
Not sure how the RC works but it should be fairy easy to implement a gentle setdown if the signal is lost for a few seconds or so. We get quite a few requests on a local FB for people to look out for drones that got out of range and kept going. Perhaps they should put pingers in them too - when I've been to amateur rocket days many of them need some serious hunting.
it should be fairy easy to implement a gentle setdown if the signal is lost for a few seconds or so
Agreed. the on-board IMU [necessary for something like a quad rotor] would tell you when the thing is hovering, and then you simply have it drop at a known safe velocity.
From what i oould tell, though, they were using an Arduino Nano (or a clone) and they only have about 32k of NVRAM. Maybe the nav stuff was on a separate Pi board but still, there's _not_ a lot of code space in an Arduino Nano and, well, results were what they were...
The photos of shoddy electronics workmanship didn't make it any better. It's not THAT hard to visually QA a circuit board, and things that fly should be getting extra scrutiny. When I was in the Navy I was an ET and even went to a special soldering school where we learned to solder to NASA spec. Burned insulation and improperly mounted components would have obviously failed the visual inspection. In fact the NASA spec took into consideration a number of things from thermal and mechanical stress on wiring and solder joints, to use of a 'conformal coating' to seal it all up. As I understand it, current NASA specs also mandate leaded solder, due to unleaded solder always looking like a 'cold joint', and to help mitigate the threat of 'whiskering'.
Unfortunately it isn't that easy. The most common failure mode on drones (or any aircraft) is a sensor failure, which results in the aircraft misunderstanding the state it is in. Many drone accidents are caused by drones with "fail-safe" procedures designed to "gently land them" causing them to impact objects at high speed because a failed sensor causes a mismatch between what the drone thinks is a gentle landing and the speed and direction it is actually traveling.
While still allowing horizontal maneouvering. Nothing less simple than a simple uncontrollable descent onto a busy motorway / tracks of an approaching high speed train / active car crusher / lit bonfire / preferred inescapable-death-event of choice.
Even model aircraft over 3.5kg in weight must be fitted with a fail safe (I'm not totally sure about the 3.5kg limit as everyone in the club sets up a fail safe regardless of their model's size).
If they had just stuck to "toy plane" safety rules there would not have been a problem.
Model flyers have to stick to a comprehensive set of CAA rules like not being permitted to overfly a crowd, or fly within 50m of a gathering of people.
More information starting to trickle out: Apparently the RC gear being used was a well known commercial brand. HOWEVER, it was set up to use the Australian frequencies (same as US FCC) in the 915 MHz band. These are illegal for model control in the UK, where we use the EU 868MHz band. In addition, the max power output permitted in the UK is 25mW on this band - one reason it is rarely used. However, the FCC systems doesn't have a 25 mW setting, so they reduced it to 10mW! This is the kind of power level you would use for an indoor toy!
Also, although the commercial system incorporates a failsafe (mandatory for models over a certain weight in the UK) this had not been set!
The more you read about it, the more you realise what an incompetent bunch of idiots were operating this thing!
--
Pete
The FCC 915MHz band is in fact a 26MHz wide band from 902MHz to 928MHz.
So the unit could have been trying to talk over Vodafone or Telefonica (O2) uplink or the start of the Vodafone down links. Given the number of people at Goodwood on the day and that the GSM uses 2W TX power, it probably wasn't going to be much use!
The build quality is below that of many hobby projects I have seen.
Well, the last Escort Mexico I saw looked like it had been maintained by a 15 year old, had stray bits of burned wire sticking out at odd angles near the power plant, badly fitted and mismatched components, and had obviously been operated outside it's design envelope because it was upside down in a ditch.
Sounds pretty close to me.
As a side note, "sombrero" is the Spanish word for hat or cap. Pretty much any hat or cap, not a specific make or model.
The AAIB report is here:
https://www.gov.uk/aaib-reports/aaib-investigation-to-alauda-airspeeder-mk-ii-uas-registration-n-slash-a-040719
Very long, but very complete and well worth reading. As usual, the AAIB did an excellent job of analysing the crash and what caused it.
Neither the CAA nor the CASA (Aussy CAA) exactly cover themselves with glory over this, while the Alauda gang, who 'built' this chunk of junk just come across as a bunch of incompetent fuckwits.
> bunch of incompetent fuckwits.
Indeed. At least they've had their arses kicked before anyone got hurt.
Video of the drone that crashed is here:
https://www.youtube.com/watch?v=tzLzeBJpofQ
Racing manned drones looks about as much fun (as a spectator sport) as watching paint dry (or watching Formula 1 from the last 10+ years, for that matter).
Airspeeder seems like a completely pointless and unnecessary "sport" - let's hope Sky TV pay them a fortune to put it behind an exclusive pay-wall as that's certain to kill it stone dead before anyone has time to care.
> F1 has been excruciatingly boring for decades.
Yep. F1 has been rubbish since James Hunt stopped racing. Or when they banned the Tyrrell 6-wheeler and other vehicle innovations.
The drivers now lack all personality while contriving a faux edgy persona (if they can bothered at all), there's no actual competition between the teams, and the sport itself is no longer the pinnacle it claims to be where the technology innovations supposedly fed into road cars - it's now more likely to be going in the opposite direction these days assuming the stupidly restrictive F1 rules allow it.
F1 selling the rights to Pay TV broadcasters is just the final nail in the coffin once the sponsors work out nobody is watching. It won't be missed.
This could be quite fun actually: loading à volunteer into a light GRP shell surrounded by 4 unprotected meat sliders - each one within arms reach - with a questionable hand built control system and get them to fly around a track for 10 minutes.
A safer alternative sport would be soap box carts going down Bury Hill with swords stuck in the grass verge and barbed wire randomly stretched across the road.
while the Alauda gang, who 'built' this chunk of junk just come across as a bunch of incompetent fuckwits.
When you invent something, and want to show it off to the public, don't make the kinds of mistakes that Tesla did, on occasion. Mad science is cool, but doesn't play well with the public when things go horribly wrong.
It's probably best to hire a safety consultant (one that's familiar with that particular field) to come in and make sure you're following regulations and taking necessary precautions, before doing anything in public.
but yeah that's the difference between "mad science" and "industry". If you can do BOTH, like Elon Musk, then you're gonna do well if you can build the thing you're dreaming up (and it actually works and doesn't kill people or break things unintentionally). But if all of the lights are going to go out for 3 counties in all directions, or airplanes might crash, nobody out there wants to be told "RUN!". Best to avoid that, yeah.
[I've got these ideas about nuclear fusion that I'd like to try. It wouldn't be all that hard, nor even expensive, to collect enough parts and build it. Problem is radiation, neighbors' safety, and stuff like that. So I haven't]
That looks very much like an Arduino nano v3 clone with an Atmel ATMEGA328P MCU next to the wireless board on the kill switch board. The AAIB report also shows the mainboard with two such devices,and the kill switch transmitter also has one. They can be purchased for under £4 each, including delivery. The type of relay board on the kill switch, around £2.50 each.
Nothing wrong with Arduino and ATMega kit in and of itself. I've found it quite robust and very reliable. Would I fly with it? Possibly in an experimental, yes. Low and slow, though. Very, very low and slow. If I were intimately familiar with the code. Maybe.
How these particular examples are assembled, on the other hand ...
Agree. I use the cheaper clone devices for prototyping, then switch to slightly more expensive ones assembled in Italy for anything more permanent.
The 2 units bought over for the demo at Goodwood were 3/4 scale versions, so very much a development prototype - but that does not excuse the issues with the design and fabrication.
Fabricated? Bollocks... I could swallow a handful of 0805s and throw up a better build than that.
There isn't a problem per se with bolting mass produced modules together, nor even necessarily with using perfboard to hold it - if it's done right. Doing it righter would require a proper PCB being designed to some sort of recognised best practice, and significant testing of failure modes. It would also include some sort of quality control, like, say, looking at the product.
In particular the most obvious requirement is, as pointed out in the article and above, don't use the presence of a signal command as a kill switch. Use the absence of a signal...
FFS there was more safety and lockout on the boards we built for PARIS.
<idle speculation> I wonder if this was 'designed' by simply scaling up an existing drone design? With perhaps little understanding at the basic level of how some of the modules really worked? </idle speculation>
<edit> And they used a normally closed relay as a kill switch? Did they go to school? Did it rain that day and the teacher didn't come?
No excuses, all drone software I've used has comprehensive fail safe features:
Bad signal - drop to floor...
Even if they'd used cheap and nasty drone racing hardware, there would have been more intrinsic safety. I can only cringe when looking at the build quality, I wouldn't even plug a flight battery into this to test it!
"nor even necessarily with using perfboard to hold it - if it's done right."
That's what I came here to say too. The build method is not relevant so long as it's done properly. Not using a custom PCB for a prototype is absolutely normal. It just requires the correct soldering skills, attention to detail and proper testing.
Now, if they'd used the re-usable "push-in contact" type breadboard I'd be a lot more concerned.
Was reading up on the Mars helicopter. The designer was telling how he ordered parts for the flight control system from SparkFun (US vendor of electronic parts for experimenters and hobbyists). "If they work OK, we'll get them certified for the mission" was his attitude.
Since the SparkFun stuff is (like Arduino sensors) essentially commercial parts placed onto pluggable PCBs or "eval boards" as they're known in the industry, I see nothing at all wrong with that. However, for flight hardware, I would expect to see purpose built PCBs where those parts are integrated into a complete system, with regulated power, latching connectors, wire harnessing, etc.
The picture looks like something I might do in the lab for testing, not something I would bet anyone's life on, which is what they did...and came away very lucky.
I was going to say the same thing. I still waiting on a couple I ordered from Aliexpress where they go for $1-2 a piece. They're probably super useful for drones for controlling servos and suchlike.
That said I think the main issue here is not so much the cobbled together bits, but the mindset of the designers. Devices that go up in the sky need to be fail to safe, i.e. the device should go into a safe mode in the absence of a command signal, i.e. if it loses contact with the command device it should hover and start descending until signal is regained or it touches down and shuts off.
The drone also needs a kill switch on top of a fail safe mode but the kill switch needs to trigger reliably through the command signal and a sidechannel, either of which causes it to cut power.
It started with F1s, I still fly controllers with F3s as well...
The newer versions of the flight software have reduced ability on the lower power chips, and you run out of UARTS for talking to pheripherals as well.
I think the Aduino in this case was just a cobbled together 433MHz killswitch, though there is Ardupilot as flight software as well.
"That flies with a person inside it @200KPH and only weighs 95 kilos?"
FTFY
Actually, the breadboard doesn't really bother me. The bloody awful assembly exhibited by these so-called "experts", on the other hand, very definitely does. I wonder if the FAA knows what they are planning on doing in the Mojave?
The Mojave is a very big place with not a whole lot going on in it. The FAA has a benign attitude towards people who want to kill themselves in experimental flying contraption -- their primairly requirement is that they do it where they're not going to harm anyone or anything.
Over the sea, or if they *really* want to do it over land then hire part of the defence training estate. (aka artillery firing ranges etc; they are actually available to rent when the military is not using them)
Or the Mach Loop; it'd be good practice for dealing with other aircraft in airspace for a flying car. ;)
"The Mojave is a very big place with not a whole lot going on in it."
That's intentional. There are quite a few rules and regs that attempt to keep it that way (unless you have a valid mining claim, of course, which is a rant for another day).
People who see the desert as a whole lot of nothing can't see the forest for trees.
I read the AAIB report with great interest as we go to Chichester regularly to visit family (and intended to retire to that area), and I have customers in the unmanned aviation industry (hence I hold both EU and UK remote pilot certificates).
To me, the most telling sentence in the report was:
"Statements made by the operator and the findings of this investigation showed that they did not appear to have any knowledge or understanding of airworthiness standards."
I also got the impression that they don't even know what they don't know.
The crash aircraft was a Mk2 demonstrator.
Current developments (source: https://airspeeder.com/) :
They are now working on the Mk3 for remotely piloted races.
I think Mk4 is intended for races with a pilot on board.
Methinks their website design is more competent than their aircraft design. Seems they've taken the "move fast and break things" approach to design a bit too literally.
> Methinks their website design is more competent than their aircraft design.
And their web design might be competent if you have decent broadband. All of the site appears to be chucked into one page which has all to be loaded. It then betrays some of the most irritating "clever" tricks which web designers with little or no user experience skills adopt.
This might be their design; certainly there does not seem to be any attribution.
... an RC model pilot, he would have had the book thrown at him - and quite rightly too! In over 50 years of flying RC models, I don't think I've ever seen such an abysmally sloppy or incompetent installation!
As an aside, I've just received my new CAA registration number for my models. It comprises: "GB-OP-" (which is fair enough) followed by a random string of 12 numbers and letters. Assuming they've done the usual 25-letter alphabet thing, that gives a total of 35 possible characters. If I've done my sums right (maths was never my strong point!), that yields 3,379,220,508,056,640,625 possible permutations.
In 2020, the total global population was estimated at 7,800,000,000 (7.8 Billion)
That means that every human alive on the face of the planet could have approximately 433 MILLION CAA registration numbers! For model aircraft!
You couldn't make it up!
I guess I'll have to tow a banner behind my model to have enough space for the registration number....
--
Pete
@Peter Christy - your number might include some coding to make it difficult to fake registrations - a sort of checksum - which means that some of the characters are needed for the coding. Credit card numbers use the Luhn algortithm - which is used for validation rather than fake-proofing. That's how websites know instantly when your credit card number isn't valid.
More likely, the number isn't random and has other information coded in like location, date, registration type, etc. Given it's the CAA, this wouldn't be surprising.
Most likely, however, is that someone who wasn't very good at maths picked the format.
soldering that bad was when I was a 16yr old apprentice... and it was me doing it shortly before having my instructor kick my arse and being made to do it properly.
And that was'nt even on stuff expected to fly (except when the instructor got really pissed off and threw it out of the window)
Quite part from there only being one kill switch system, both the transmitter (report page 10 fig. 8) and receiver (report page 13 fig. 10) are abominations. A safety device must be more robust than the thing it's there to protect, and also ideally multiply redundant. These were clearly mere afterthoughts by someone who does not understand engineering in the slightest. Unpardonable considering a 95kg drone could easily kill someone in falling or destroy an aircraft engine.
This unfortunately typifies IoT. The negligent attitude common to software development has permeated the more established branches of engineering, dragging their standards down to its level.
"Safety Recommendation 2021-012
It is recommended that the Civil Aviation Authority, before issuing an Operational Authorisation to operate an Unmanned Aircraft System they have not previously had experience with, carry out a physical examination of the Unmanned Aircraft System to ensure that it is designed and built to suitable standards, and observe a test flight to confirm operation in accordance with the Operating Safety Case." [incident report page 54]
Good idea.
Approval for this flight was given without sight of a completely unknown craft from a completely unknown manufacturer, based solely on information provided by that manufacturer.
A bit like at Boeing?
"Approval for this flight was given without sight of a completely unknown craft from a completely unknown manufacturer, based solely on information provided by that manufacturer."
It also had approval from the Australian regulator, which was accepted by the CAA. After the Boeing/FAA debacle, maybe the CAA (and by extension, all air worthiness regulators) should be seriously considering not blindly accepting certifications from other countries until they've had another look at said procedures and at least spot checking some random samples to see if reality matches up with theory.
Incidentally do authorities have an explanation for the *other* Gatwick drone incident?
It seems that the unidentified object was hovering for hours at a time, and apparently a certain spooky agency (tm) sent up to investigate may or may not have clear pictures of the object taken on a Police drone which (surprise!) were classified for reasons of national security due to the nature of its propulsion system.
"No doubt your reaction to that will depend on your preconceptions..."
My reaction is mostly predicted by the fact that nobody has presented one, single, solitary shred of evidence that any drones existed in that airspace at that time. Given the very few facts that we do know, it was most likely a minor case of mass hysteria. The authorities should just admit they over-reacted, are continuing to over-react, and have spent inordinate amounts of the Public's money on a wild goose chase. It's time to drop it and move on, already.
The UK is not as relaxed about "experimental" aircraft as the USA and calling the planned manned version a "drone" is an attempt to dodge very strict regulations
Not being required to have a GNSS is remarkable. You would hope that loss of signal shoul make the aircraft freeze and hold position for a while to get gignal back and failing that land gently. This machine is way to heavy to just drop
I'm not intending to excuse the shortcomings of the operators, but on a point of (geographical) order: Goodwood Festival of Speed (GFoS) is held in front of Goodwood House, which is at least a mile north of the Goodwood motor race circuit which surrounds the aerodrome.
On that day therefore there would have been a minimal crowd at the circuit.
There is no proper 'aerodrome' near the House, simply some temporary areas set up to facilitate rotary wing ops for the duration of the event.
The aerial view on page iv of the AAIB report is misleading because it shows three arrows pointing to what it describes as the "Event Area", but in fact the event is predominantly located to the north of the northernmost arrow.
I would have expected better of the AAIB, and frankly the reporting of The Register seems to conflate the circuit and the hillclimb.
As a young lad I would occasionally accompany my dad to watch him fly RC planes and sometimes have a twiddle on the controls myself before getting bored and wandering off to drop stones in ponds, explore the abandoned buildings from when it was an active flying strip or other more interesting things.
One of his models had the on/off switch for the electrics mounted on the side of the fuselage for easy access. The usual take-off method for this plane was to get the motor running, then carefully pick it up and throw it from one hand then quickly grab the transmitter from an unwilling assistant and start flying. Except on this one occasion he managed to catch the on/off switch with his finger mid-throw and was rewarded with the sight of it flying perfectly straight and climbing slightly over the trees at the far edge of their field and away across the countryside, ignoring all control inputs. From this I learned two things:
1. Ergonomics is a useful science. If the only places you can put the on/off switch means it can be accidentally caught by an errant finger, make sure the forward position is on.
2. Never repeat the new words learned from your father and his flying buddies after an incident in front of your mother later that day.
Happy Ending 1: A couple of weeks later he got a call from a farmer who'd found the plane in a field 20 miles away, slightly trampled by cows. The plane never flew again but at least he got the receiver, servos and engine back which were still usable and quite expensive.
Happy Ending 2: In later years he had a job involving the design of cockpits and flight instruments on real aircraft. He never said where the master on/off switch for them was located.
reading the AAIB report it looks like the CAA asked the CASA for information. CASA said get the operators permission to release information but the CAA didn't follow up and ask operator for permission and so didn't find out what CASA knew or thought.
Between that and not actually looking at the drone, on the question of inspection the CAA seem to have not only dropped the ball, but knocked it into their own net.
That said the CAA had made some strict limitations to the permitted flight (eg 20 m max height) on the basis of the lack of geofence, multiple points of failure etc.
This is a bit of a red herring: the AIP contains nominal distances such as runway lengths and taxiway widths along with a schematic of the aerodrome layout.
It doesn’t contain the information the drone operators needed, which was the distance between the runway in use and the crowd line. That could have been obtained from Goodwood Aerodrome itself.
Part of the information contained in the AIP are the position and elevation of the runway thresholds (https://www.aurora.nats.co.uk/htmlAIP/Publications/2021-01-28-AIRAC/html/eAIP/EG-AD-2.EGHR-en-GB.html#AD-2.EGHR), and it is in fact the official source for such information.
That data does usually originate from the airfield operator, via an approved survey, but the official source is the AIP (in its printed or CD-ROM version, if one wants to be pedantic).
Not so much. The operator was using Google Earth satellite images, and presumably the ruler function, to draw his display box and safety buffer zone. He didn't check with the aerodrome if they had any up to date survey maps he could have used. The Google Earth image was out of date because the runway had been reduced in width after it was taken.
The AIP doesn't contain distances from runway edges to buildings, which in this case was the measurement the operator needed. See pp xx and xxi of the report.
Hi Gaz, I see your train of thought but:
> He didn't check with the aerodrome if they had any up to date survey maps he could have used.
He would have been directed to the AIP. Or to be fair, he should have known to look in the AIP if he had an inkling of an idea of what he was doing.
From the AIP he could also have obtained the runway centreline and width. Then he could have offset the centreline of RWY 14/32 Eastward by the semiwidth (15 m) to find the runway edge (though for grass runways the edge is hardly ever identifiable on the ground) and that would have given him the edge of what he claimed was to be his display area, from which he then had to distance the guests somehow.
Incidentally, did he actually publish his video of the Chichester demo, omitting one rather crucial aspect? https://www.youtube.com/watch?v=tzLzeBJpofQ
The aircraft was designed for high speed, high performance operations and the importance
of the control systems functioning correctly was reflected in the risk assessment. The flight
control circuit boards were mounted on Velcro with a foam lining, in an IP55 box. The circuit
boards were not subject to any vibration, shock, RF or temperature testing and the in-house
developed software was not developed to any level of assurance.
Masterpieces of the art of sentence juxtaposition. :)
I for one find it quite amazing what they've accomplished. Think of it: about 20 years ago it would have been prohibitively expensive to create such a flying "car". And nowadays you get some ready-to-use components for a few quid at Maplin, stick and solder them together and there you go: you've got a flying crap. Surely, they did give exactly one flying fuck about safety, reliability and risk in general. Surely, it got out of control and crashed. But it was flying ffs.
</sarcasm>
I'm deeply interested in what these RC models, representing something that will eventually carry a human, are accomplishing for the company today. It looks like fundraising at best and at worst, a massive cash burn and/or fraud.
Quad-motor RC piloted software and hardware already exist, so nothing novel here. And from the investigation, they're not collecting any engineering data to help them identify issues and inform them when they've fixed past problems with these flights.... so what is the purpose for these flights?
As a result of the investigation, it looks like they had to solve a crap-ton (technical phrase) of engineering and safety issues that are likely explicit for an unmanned vehicle, which take them no closer to a manned flight.
Once they put a meatbag in this machine, this becomes something wholly different from a regulatory point of view.
You answered it for yourself. They exist as a "Wow!" factor to hopefully help put a crap-ton[0] of investor money into the direct, personal control of the officers of the company, with little to no oversight as to how those funds are to be distributed. In other words, advertising from the marketing department. (Don't look at that man behind the curtain!)
Exciting investment opportunity or scam? You decide with your pocketbook.
Caveat emptor, quia ignorare non debuit quod jus alienum emit
[0] If you're into the Metricals, that would be a crap-tonne. The conversion factor is 1:1
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
