back to article Poor software design led to second £1m Army spy drone crash

An Army Watchkeeper drone flown by the Royal Artillery crashed on landing after its crew selected the self-flying craft’s “master override” function, according to the official report into the accident. Thales Watchkeeper drone WK006 flopped into the runway at Boscombe Down airfield in November 2015 when its two-man crew, …

  1. W4YBO

    "The panel concluded, in its analysis, that WK006 incorrectly sensed its altitude because the two onboard laser altimeters both recorded altitudes of less than 1m above ground level thanks to reflections from the low cloud it was flying through."

    Swap one for an infrared laser that'll punch through clouds. Fixed.

    1. Simon Harris

      Maybe they shouldn't have relied on cloud computing.

    2. Anonymous Coward
      Anonymous Coward

      I'm no expert, but I'd guess that in order to provide correct readings over lakes, rivers and other bodies of water, laser altimeters probably intentionally use a wavelength that is going to be reflected by water. If that's the case, then given what clouds are made of, that's always going to be a problem.

      1. SkippyBing

        The 'interesting' thing is that the Laser Altimeters are first tested at the CP at which point there is no way, according to the manufacturer, that they could actually get a ground return. Despite this the software will accept the height reported which doesn't really help.

        The technical parts of the report are quite an interesting read, you get the feeling Thales logic was cobbled together as they noticed things that didn't work...

        1. Anonymous Coward
          Anonymous Coward

          The technical parts of the report are quite an interesting read, you get the feeling Thales logic was cobbled together as they noticed things that didn't work...

          It's Semi-Autonomous Vehicle Not In Space Stuff (SAVNISS?). There is no doctrine on how to do that....

        2. cantankerous swineherd

          test driven development?

          1. Anonymous Coward
            Anonymous Coward

            DevOps. Or like we call it DevOOOoooooooops!

          2. Vic

            test driven development?

            Crash early, crash often.

            Vic.

        3. Alan Brown Silver badge

          "you get the feeling Thales logic was cobbled together as they noticed things that didn't work..."

          This is more than likely to be the case, most software end up like this even if it was more-or-less straightforward in the first place.

          The question is whether it was designed to have software modules added/removed as required or if the programmers wrote a Spaghetti Monster from Hell (I'm betting the latter)

          Now factor in that Thales not only write drone software, but also do stuff for transportation systems with critical safety requirements (such as railways and civil aviation) and that substandard programming ethos tends to be company-wide.

      2. John Smith 19 Gold badge
        FAIL

        "laser altimeters probably intentionally.. wavelength that is going to be reflected by water"

        Maybe so.

        But radar altimeters have been around for decades and for UAV use can be as light as 0.5Kg.

        And WTF did the GPS not confirm it or rather, deny the data was right.

      3. Trigonoceps occipitalis

        When I was last over the Möhne Lake we used two angled spot lights. I wouldn't have been able to see laser spots even then. Now I'm old and gray I have no chance.

      4. razorfishsl

        Or they could just use the time validated systems used by real aircraft........

      5. Voland's right hand Silver badge

        I'm no expert, but I'd guess that in order to provide correct readings over lakes

        A primitive radar altimeter which is used for landing and emergencies only is only a few hundred grams and a few thousand pounds worth of money. Landing uses radar guidance anyway so not like the drone needs to maintain radio silence when landing.

        Not including it in a 1M aircraft has only two words for it - criminal incompetence.

    3. Anonymous Coward
      Anonymous Coward

      Swap one for an infrared laser that'll punch through clouds

      Or even a radar altimeter, which have worked perfectly well in military aircraft for decades. As have weight-on-wheels switches for that matter, even during rough ground operations. You have to wonder how much military aviation experience the designers of this thing have.

      1. AlbertH

        You have to wonder how much military aviation experience the designers of this thing have.

        None whatsoever. I know two ex-members of the Thales team who left because of the fundamental ineptitude of the management. Many obvious, simple and inexpensive fixes were proposed for these vehicles, but management didn't want to know.

        After all, the more of these drones that the military destroy, the more Thales will sell - especially since they can always blame "pilot error" when the wetware decides that intervention is necessary!

      2. Anonymous Coward
        Anonymous Coward

        This software...

        >>>You have to wonder how much military aviation experience the designers of this thing have.

        ...was probably written by the same jokers that wrote the software in my (well known brand German) car.

        It repeatedly (8 times) identified the wrong components over period of a year for a speed sensing fault racking up £750 in repair costs that did nothing to cure the problem. At one point I was told it may need a £4,000 replacement gearbox! Eventually a grunt in the dealers garage said "Why don't we just try changing skibberly-bob-dubree, we've tried everything else that the fcuking software has told us?" He changed it and bingo - fault fixed! The skibberly-bob-dubree cost £50 and took 10 minutes to fit.

        1. Alan Brown Silver badge

          Re: This software...

          "...was probably written by the same jokers that wrote the software in my (well known brand German) car."

          Something similar happened in a friend's Mercedes(*). The onboard computer said about £2500 pf parts were busted, so the shop replaced them - and the computer still said they were busted, but friend was out £2500 in parts plus the labour.

          I borrowed the car and took it to my friendly local wizard. We discovered very quickly that the thermostat was jammed open, the engine remained stone cold and this was confusing the hell out of the computer.

          The stealership refused to accept responsibilty for this amazing level of servicing incompetence and tried to claim their warranty was voided by someone else looking at the vehicle. My friend didn't have the stomach for a legal fight and just replaced the thermostat (coincidentally also a £50 part on a merc), whereupon all the computer complaints went away.

          Moral: If there are a lot of systems supposedly wrong, there's usually a common cause. You don't just replace everything that "computer says", but tales like this are depressingly common in the motoring industry (My own car blew an O ring in the steering rack at 3 years old. Nissan's answer was a new rack. A reconditioning outfit removed, serviced, replaced the O ring and machined a second groove for a second O ring(*), then refitted it for less than the stealership's quoted labour cost, which would have been 1/3 of the total bill)

          (*) Unless AC has XX chromosomes, it's unlikely to be the same car.

          (**) Apparently Input shaft o-ring weakness is a known issue that most makers have never addressed.

          1. DropBear
            Facepalm

            Re: This software...

            Well there's this time when I fixed the AC in my car simply not turning on at all by replacing a water temperature sensor, while the computer insisted there are no error codes whatsoever AND said sensor was reading out perfectly normally right on my dashboard...? Clue: it was a dual-element sensor, and only the half going to the ECM was busted - the engine thought it was way past boiling and refused to engage the AC clutch (while utterly failing to voice this concern of his in any way)... fun times. I shudder to think what a "service fix" would have cost me, and how many failed attempts...

  2. Anonymous Coward
    WTF?

    OK here's a fix....

    If a switch breaks due to rough terrain (find that hard to believe it they used better switches).

    How about a simple, the wheel is spinning at the correct speed sensor?

    Now, if you can pay me a few hundred million for my consultancy time, much appreciated,

    1. frank ly

      Re: OK here's a fix....

      Or a magnetically actuated switch, or ... or ... or.......

      1. tqla

        Re: OK here's a fix....

        Motocross teams use something along those lines for suspension datalogging....

      2. Alan Brown Silver badge

        Re: OK here's a fix....

        I'd say both.

        Hall effect sensors for spin are rugged and lightweight - and weight on wheels can trivially be measured with a strain gauge stuck onto the landing struts - which would only break if the struts did, so it'd have to be one hell of a rough field.

        1. wellthen

          Re: OK here's a fix....

          I was thinking about wheel speed sensors. I think they tend to turn in the air making it a bit fiddlier (especially after a bounce). That said you can have one on each wheel to try correlate.

          Can't argue with strain gauges though. Cheap, reliable, really cheap. Only someone with literally no engineering experience wouldn't think of that. I have nothing but contempt for their chief engineer.

  3. Sureo

    "...train them as instructors ..."

    Now they've learned what not to do.

    Let's hope self-driving autos have more reliable sensors and software.

    1. Anonymous Coward
      Anonymous Coward

      No they don't.

      http://www.pcmag.com/news/350343/autonomous-uber-car-caught-running-red-light

  4. allthecoolshortnamesweretaken

    "...it struck the ground at a 35.29 degree nose down angle..."

    I'm a stckler for detail, precision and accuracy, but this is bordering on being ridiculous.

    1. Comments are attributed to your handle
      Trollface

      "stckler for detail" - yeah, sure you are.

      1. allthecoolshortnamesweretaken

        Oh deer...

        1. Steve Davies 3 Silver badge
          Joke

          Oh Deer?

          Is that a Sitka or a Red Deer in your headlights?

          1. Anonymous Coward
            Joke

            Re: Oh Deer?

            > Is that a Sitka or a Red Deer in your headlights?

            Hinds. As in the kind of deer that you can only see with hind sight.

    2. DropBear
      Trollface

      "...it struck the ground at a 35.29 degree nose down angle..."

      Ehhh, it's all just a big cover-up. What actually happened is the drone went sentient, saw a curvy B-2 on the ground and just told her "hey gorgeous, wanna see a cool Stuka impression...?"

  5. lafnlab
    Black Helicopters

    Outsourcing

    Maybe they can hire Amazon to get their drones to drop grenades on people.

    1. Anonymous Coward
      Anonymous Coward

      Re: Outsourcing

      For dropping shit, yes. But if you want grenades throwing over a fence, or IEDs leaving in dustbins, then Yodel have more expertise than Amazon.

      1. CrazyOldCatMan Silver badge
        Stop

        Re: Outsourcing

        throwing over a fence, or IEDs leaving in dustbins, then Yodel have more expertise than Amazon.

        Fortunately, my $WINESUPPLIER of choice appears to have ditched Yodel as a carrier. I suffered at least a 70% complaint rate with my deliveries given their habit of utterly failing to read delivery instructions (because walking an extra 8m to put the easily-nickable case of wine round the side of the house out of the view of the general public walking past my house is a concept their delivery drivers failed to grasp - as the result of which at least 2 £100-plus cases of wine went walkabout).

        1. Alan Brown Silver badge

          Re: Outsourcing

          "Fortunately, my $WINESUPPLIER of choice appears to have ditched Yodel as a carrier."

          Perhaps it's the same $WINESUPPLIER who used them to deliver a shipment to the right address (mine) in the wrong town (ie, I didn't order it). I refused delivery twice in 2 days then came home a few days later to find the previously rejected delivery dumped on my doorstep with an obviously broken bottle inside.

          $WINESUPPLIER was duly informed and promised to arrange a pickup, but never did so.

          I use the anecdote (and Yodel dumping stuff in the bushes outside $orkplace security gates because they can't be bothered using the intercom) as good reasons to warn people off such companies. It gets amusing when they try to use "Yodel have recorded the delivery" as a defence for breaching distance selling regulations.

  6. Hollerithevo

    Fog of war?

    "...two onboard laser altimeters both recorded altitudes of less than 1m above ground level thanks to reflections from the low cloud..."

  7. Scott Broukell

    Do I need to thpell it out for them . . .

    Thales should only be used at sea, not in the air and not on the ground! - except maybe for the New Year Thales in thopping thentres.

  8. Cynic_999

    Lack of communication

    So it seems that basically the manufacturers did not design the drone to land under such a low cloudbase, but the operators were not informed of that fact.

    Reminds me of the pilot's defect report, "Autoland requires adjustment - touchdown was extremely firm!" The engineer's note stated, "This aircraft is not fitted with autoland."

  9. Anonymous Coward
    Anonymous Coward

    When R/C hobby models now come with autoland, these folks are simply pathetic.

    I recently saw an ad for a US$350 model airplane that features autolanding capability: https://www.hobbyzone.com/rc-airplanes/parkflyer/HBZ8400.html

    Perhaps these folks should buy one and use it as a follow-me to guide their megabuck marvel to a landing when the going gets rough.

    1. Adam 52 Silver badge

      Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

      I have this feeling that a 950g model may have substantially less kinetic and potential energy than the aircraft in question.

      1. Gordon 10
        FAIL

        Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

        Whats that got to do with it? They are simply variables that should plug into the same auto landing module.

        The characteristics of landing will be different but only in the details. The same basic flight parameters have to achieved.

        1. Anonymous Coward
          Anonymous Coward

          Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

          Yes, it autolands by GPS.

          So how well does it perform really? Are small errors important or are they caught by the physics of the system? Can you think of a reason why autolanding-by-GPS would NOT be appropriate for a largish drone?

          The characteristics of landing will be different but only in the details. The same basic flight parameters have to achieved.

          Frankly unlikely.

          1. dajames

            Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

            Can you think of a reason why autolanding-by-GPS would NOT be appropriate for a largish drone?

            GPS can be spoofed or jammed. I seem to recall reading that an American drone was captured somewhere in the Middle East, some time ago, and was thought to have been induced to crash-land by feeding it bogus GPS signals. Were I designing a drone -- especially a military one -- I would certainly not use GPS as the sole means of determining position or altitude.

        2. Adam 52 Silver badge

          Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

          "The characteristics of landing will be different but only in the details"

          I'm not sure why I'm even responding to this, but you got upvotes so...

          Below a certain weight you can drop anything and it'll land safely. A foam child's toy will be OK if it smashes head first into a wall. A model aircraft will often land safely if you just fold the wings and tail.

          You can stall a 1kg model in from 10cm height just fine. Scale that up and you certainly can't stall a 300,000kg Jumbo jet in from 30km in the air.

          The margin of error in a model is several orders of magnitude bigger than in a military aircraft.

        3. Rich 11

          Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

          The characteristics of landing will be different but only in the details.

          Really? Consider a biological equivalent:

          You can drop a mouse down a thousand-yard mine shaft; and, on arriving at the bottom, it gets a slight shock and walks away, provided that the ground is fairly soft. A rat is killed, a man is broken, a horse splashes. -- JBS Haldane

          1. Anonymous Coward
            Anonymous Coward

            Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

            Why does JBS bloke that sound like a physco in the making?

            1. Alan Brown Silver badge

              Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

              "Why does JBS bloke that sound like a physco in the making?"

              Which order did he drop them in? :)

          2. Anonymous Coward
            Anonymous Coward

            Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

            Actually I think the RC plane analogy is perfectly fine, the difference in the landing is in the detail. For a model, a drone, a jumbo jet, the idea is to get close to the ground and then reduce speed so you touch down gently. The aerodynamics are the same, the data sensing requirements are the same. The detail is in speed and how much g the components can take. Take your mouse, rat man and horse and ask them all to jump off a chair. All use their eyes to estimate the drop, flex their knees (do mice have knees?) on landing in a similar manner. Hopefully all walk away.

          3. CrazyOldCatMan Silver badge

            Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

            You can drop a mouse down a thousand-yard mine shaft; and, on arriving at the bottom, it gets a slight shock and walks away, provided that the ground is fairly soft. A rat is killed, a man is broken, a horse splashes. -- JBS Haldane

            How about a whale? (With or without a bowl of petunias)..

            1. Kubla Cant

              Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

              You can drop a mouse down a thousand-yard mine shaft; and, on arriving at the bottom, it gets a slight shock and walks away, provided that the ground is fairly soft. A rat is killed, a man is broken, a horse splashes. -- JBS Haldane

              How about a whale? (With or without a bowl of petunias)..

              The whale would get stuck. For it is written: "Again I tell you, it is easier for a camel to go through the eye of a needle than for a whale to fit down a mine shaft."

              1. TRT Silver badge

                Re: When R/C hobby models now come with autoland, these folks are simply pathetic.

                You could fit a mouse to the front of the drone... see if it starts panicking when the engine cuts out because the drone thinks it has landed.

  10. Gene Cash Silver badge

    Are these the same blokes that designed Schiaparelli's landing system? Seems "I'm on the ground! (when I'm not!)" is making the rounds...

  11. Anonymous Coward
    Anonymous Coward

    This is the get out for NATO starting WW3. It was the drone wot done it.

  12. Anonymous Coward
    Anonymous Coward

    Many moons ago, when I was a young graduate, I worked on the UAV for a system called Phoenix. Phoenix went into service with the British Army back in the 80's, and it didn't bother trying to land its UAV on a runway. Instead the UAV would return to earth under a parachute and by cunning means it would be flipped on to its back so as to protect its expensive & venerable sensor package that it carried.

    Except sometimes it didn't, and the expensive & venerable sensor package would get trashed.

    Nothing changes.

    Phoenix is now a museum exhibit, boy does that make me feel old.

    1. Anonymous Coward
      Anonymous Coward

      Phoenix is now a museum exhibit, boy does that make me feel old.

      Hahhahahaa! Obsolescence boy! I supported Tornadoes at the same time, and they're still in service.

      Admittedly they need the aerospace equivalent of a zimmer frame, but, yeah, well.......can I offer you a Werther's Original?

      1. Swiss Anton

        You think the Tornado is old.

        I've seen B52s flying in the not too distant past around where I live (Swindon, SN1), and the Russkies still have their TU-92s. If the Tornado is zimmer frame, are these aircraft zombies?

        1. Destroy All Monsters Silver badge

          Tomorrow in the Mail:

          PUTIN commandeers ZOMBIE AIRPLANES!

        2. CrazyOldCatMan Silver badge

          I've seen B52s flying in the not too distant past around where I live (Swindon, SN1)

          That'll be because of the air shows at RAF Fairford..

          (I'm also in Swindon - SN5)

        3. Anonymous Coward
          Anonymous Coward

          If the Tornado is zimmer frame, are these aircraft zombies?

          No. There's a story on the Reg today about a Tri-star launching a satellite (=missile), and what that shows is that relatively dumb weapons platforms can be ages old, but still effective. But fast jets are different, because they are supposed to operate in much tighter envelopes, closer to hazard.

        4. John Brown (no body) Silver badge

          "are these aircraft zombies?"

          The zombie aircraft are eating the dronesBRAAAAaaaiiiinnnnzzzzzz hence the crashes.

    2. DropBear
      Trollface

      ...wherein "cunning means" actually just means the eye bolt for the parachute was on the drone's belly. C'mon, admit it...

  13. Magani
    Boffin

    Old quote might have made sense

    If God had wanted the Army to fly,

    He would have painted the sky brown.

    RAF/RAAF/USAF/etc saying.

    1. CrazyOldCatMan Silver badge

      Re: Old quote might have made sense

      He would have painted the sky brown.

      Ahh brown. The colour of fear..

  14. Norman Nescio Silver badge

    GPS - not as good as you might think

    Anyone thinking of using GPS for landing might need to be aware of a couple of things:

    1) GPS altitudes are heights above a calculated shape that approximates the earth's surface (a geoid). Local deviations from the calculated geoid can be large. Knowing the GPS altitude of the landing surface before landing on it would be a good idea, but that information may not be available to the UAV. You either need to take-off from the same surface and remember what your GPS height was; or have an extremely detailed topographic map available to you.

    2) GPS altitudes are considerably less accurate than horizontal co-ordinates. This reference (to civilian GPS, but the maths works for military too) http://gpsinformation.net/main/altitude.htm gives a possible vertical error of 40 metres, which in the context of landing a UAV is significant. That reference is from 2001, but the current GPS performance standards (issued in 2008) back it up: see page 34 of this document http://www.gps.gov/technical/ps/2008-SPS-performance-standard.pdf where the vertical accuracy worst case 95th percentile is 37 metres. The global average is 15 metres, so you probably want to augment GPS with something to make sure the UAV doesn't stop descending 40 metres too soon; or attempts to continue descending until 40 metres too late (aka CFIT).

    There's probably a good reason why a radar altimeter isn't used, but I don't know what it is.

    1. Alan Brown Silver badge

      Re: GPS - not as good as you might think

      "There's probably a good reason why a radar altimeter isn't used, but I don't know what it is."

      Probably the same reason why Thales deleted the WoW sensor instead of going to a more reliable one.

      Cheapeset bidder, built to specification even if that spec is obviously missing critical parts - all such additions are off-contract extras at some exorbitant price.

      1. Anonymous Coward
        Coat

        Re: GPS - not as good as you might think

        Thales deleted the WoW sensor instead of going to a more reliable one.

        You can get sensors for World of Warcraft?

    2. John Brown (no body) Silver badge

      Re: GPS - not as good as you might think

      "2) GPS altitudes are considerably less accurate than horizontal co-ordinates."

      And I can confirm that from experience. The SatNav I use in the car shows altitude and it's far from accurate, varying by up to 10' up and down, sometimes second by second. It regularly show altitude as negative when I'm clearly still on the road by the sea, even with the tide in and I'm not even a little bit wet :-)

      As you say, civilian GPS access is not as accurate as what the military have access to, but altitude is clearly not safe or reliable for a landing system on it's own without a LOT of extra topography data or a guarantee of landing back at the start point or some other point known in advance.

  15. bep

    Re the weight-on-wheels thing

    Shirley with modern technology and computers and things, it should be possible to have a programmable weight-on-wheels thing that you switch on or off depending on whether you are landing the thing on a concrete runway or a dirt strip?

    1. Alan Brown Silver badge

      Re: Re the weight-on-wheels thing

      It's not the matter of turning it on or off. It's the matter that Thales are apparently still using fragile mechanical switches to detect such things and they get broken on rough ground.

      A strain gauge simply sticks to the strut and measures the flex that occurs as the aircraft settles - if you look at the pictures you'll see it's U-shaped. It's that way because it's a spring and springs flex.

      http://uk.rs-online.com/web/c/automation-control-gear/sensors-transducers/strain-gauges/

      This kind of thing isn't exactly new. Aircraft have been using strain-guage based WoW sensing for some time - precisely due to the fragility of oleo sensor switches (WoW sensors on manually flown aircraft are necessary for civil autoland systems but are more frequently used as lockout devices to prevent someone accidentally hitting the gear-up switch whilst on the ground. It happens occasionally and causes much embarassment to everyone concerned)

      1. TRT Silver badge

        Re: Re the weight-on-wheels thing

        Thales idea of a WoW sensor is a bleeding V4 lever microswitch... their idea of an advanced version is the sort with a roller end.

  16. Pompous Git Silver badge

    An Army Watchkeeper drone flown by the Royal Artillery crashed on landing
    Er, isn't that what artillery's for? Sending a projectile through the air that crashes and explodes on landing (preferably nearer to the enemy than yourself). Perhaps they need someone whose expertise includes landing without exploding.

    1. Vinyl-Junkie
      Thumb Up

      An Army Watchkeeper drone flown by the Royal Artillery....

      My thoughts exactly. The Royal Artillery has centuries of experience of throwing stuff at the enemy so it goes bang on landing. Landing safely? Not so much!

      Should have given them to the AAC.

  17. razorfishsl

    company said landings on rough surfaces would be likely to break the switch, a problem it had noticed on its other drones.

    Which just shows how shite the design is for not physically decoupling the switch from the gear.......

    1. Richard 12 Silver badge

      Or for thinking of a switch at all

      As has been mentioned above several times - strain gauges are both damn near indestructible and incredibly cheap.

      Plus you get a "How hard did the wheel hit the deck" for free.

    2. John Brown (no body) Silver badge
      Facepalm

      "company said landings on rough surfaces would be likely to break the switch, a problem it had noticed on its other drones."

      Exactly. A proper engineer would design or specify a working solution. Thales answer to a known problem was to simply not bother sensing that action at all.

  18. Joe 35

    So someone in Britain......

    ..... bought a deadly aircraft that doesn't work if its a bit cloudy????

    1. Peter Christy

      Re: So someone in Britain......

      Well, we've bought two very expensive aircraft carriers, for which we have no aircraft! And the aircraft that we've ordered to (eventually) fly from them are not only exceedingly expensive, but according to reports, less capable than the aircraft they were supposed to replace!

      I wouldn't be surprised if we never see the F-35s. Just like the F-111, the contract will end up being cancelled, but we'll still end up paying for them.

      Who writes these contracts, and why have they still got jobs?

      1. phuzz Silver badge

        Re: So someone in Britain......

        "Who writes these contracts, and why have they still got jobs?"

        Well, they wrote the contract when they worked for the MoD, and now they have a job with the company that won the contract...

  19. imanidiot Silver badge
    Facepalm

    In other words

    Even with the recommendations and problems already identified in the earlier crash, the UK government decided not to sue Thales for incompetence, Thales is still incompetent and the programming for the the thing is still F*^#ing stupid!

    1. SkippyBing

      Re: In other words

      I'd have to check the dates but I think the second crash happened prior to the inquiry into the first one being completed. Of course this doesn't mean MoD actually have a way of forcing Thales to rewrite the software. Well I mean they have guns and stuff but I think that's frowned on.

  20. hammarbtyp

    Different TLA

    Strange in my day the Weight on Wheels sensor(W.O.W) was called Weight on Ground.

    No idea why they changed it :)

  21. Wiltshire
    Black Helicopters

    Something's strange here. Drones have been seen being tested on grass strips elsewhere across Salisbury Plain. So, are there are other makes of drones that are more reliable on rough ground? Or perhaps that's only in good weather as well?

  22. Norman Nescio Silver badge

    More on GPS autoland

    The way model aircraft use GPS to allow autoland takes advantage of the fact that their flight durations are short. If you measure your GPS location at a fixed location, then measure it again a few minutes later, it will be different, but not that different. The location wobbles about over time, partly as a function of weather in the ionosphere making signal path lengths from the GPS satellites vary, partly due to the very, very, slight irregularity in the clocks in the satellites (and for many other reasons). What you will find that over a 24 hour period, if you plot all your position fixes, 95% of them will be within a circle of small radius that will very likely also have your real position within that circle. So any particular location fix is likely to be wrong, as is the next one, but they will be wrong by roughly the same amount as the wobble or drift usually takes place over relatively long time periods. In the link here: ( http://gpsworld.com/gpsgnss-accuracy-lies-damn-lies-and-statistics-1134/ ) Figure 4 shows a plot of 10,000 GPS fixes taken over 3 hours as a scatter plot, and you can see how successive plots don't differ much from each other, but over time, they wander about quite a lot.

    This means a model aircraft can take a GPS reading at take-off and 10 minutes later, when it is landing, the wobble or drift won't have been that much, so you can rely upon the GPS position you are currently measuring to be 'near-enough' the take-off location for landing purposes.

    Large military UAVs will typically have flight durations that are greater than 10 minutes. According to Wikipedia, the General Atomics MQ-9 Reaper can have 14 hours endurance, and that is long enough for GPS position errors to add up, so knowing what the GPS position of your take-off point was 14 hours ago doesn't help much, because in the 'average' worst case, the current fix could be 40 metres off vertically from the one at take-off; and with 14 hours endurance, there is no guarantee the mission requires you to land where you took off, either.

    The FAA are looking at using GPS for autoland on commercial airliners.

    http://aviationweek.com/commercial-aviation/faa-targets-2018-gps-based-autoland-capability

    However, that uses ground-based transmissions to augment the basic GPS signals. This uses ground-based GPS receivers that have well-defined known fixed locations so that the current GPS errors can be determined and sent to the aircraft to correct its own GPS reading.

    I imagine that military UAVs might have a requirement to be able to operate in 'radio silence', as you don't really want a SAM locking onto its radar altimeter transmissions; or indeed a missile locking onto ground transmissions at its landing point. That does make operating in IMC difficult.

    I agree with other posters that a strain sensor rather than a WoW switch makes sense.

    1. Anonymous Coward
      Anonymous Coward

      Re: More on GPS autoland

      I don't think - but am happy to be corrected - that rc auto land systems use GPS height - I doubt it ever has the accuracy. The system I'm familiar with more commonly get's its altitude reference either from a barometer or from a range sensor, either ultrasonic or optical, or a combination of both. Barometers obviously drift , but as the drone is landing close to the operator, there's no real issue in providing it with an accurate barometric reference, if that's the way you want to go. They use a GPS to get the plane into the correct horizontal position, and then trigger landing flare actions on the basis of height and rate of descent.

      The irony is that the specific failure that caused this crash is highlighted in the wiki page of a commonly available model autopilot, that you could pick up off ebay for £20 or so:

      Also note that if you have a longer range rangefinder then it is a very good idea to set the minimum range of the rangefinder well above zero. For example, the PulsedLight Lidar has a typical range of over 40 meters, and when it gets false readings it tends to read ranges of less than 1 meter. Setting RNGFND_MIN_CM to 150 will discard any rangefinder readings below 1.5 meters, and will greatly improve the robustness of the Lidar for landing.

      1. Norman Nescio Silver badge

        Re: More on GPS autoland

        Thanks for that - upvoted accordingly. I don't operate drones of any kind, and the manual I could find online made no reference to Lidar or Barometry.

        1. Anonymous Coward
          Anonymous Coward

          Re: More on GPS autoland

          If you want to read more on how this stuff is done I can recommend the Ardupilot website. You do take a risk - the toys and technology are so seductive that your bank balance may be damaged, unless you are on Santa's 'nice' list.

  23. anonymous boring coward Silver badge

    "Its vertical acceleration reduced enough for the onboard logic to decide that it was now straight and level and therefore on the ground"

    WTF!!!???

  24. Anonymous Coward
    Anonymous Coward

    So it's not rated for an instrument landing.

    Time to divert to another airport / base then.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like