Interesting use of the term 'crew' - traditionally associated with people actually on board a ship or aircraft.
A £1m British Army Watchkeeper drone had to be scrapped after crashing at an airfield in Wales when the ex-RAF officer piloting it disabled the unmanned aerial vehicle's anti-crash systems. Although the official main cause of the accident was given as the automated Vehicle Management System Computer functioning “as designed …
"There was no suggestion in the report that the pilot was culpable for the drone's crash."
First time in history a "pilot" was not held accountable for a crashed vessel under his command after giving specific orders overriding safety protocols. He must be a son of a serving member of Parliament or something. .
As the article explained in length: leaving crash protection on would have resulted in drone doing another fly-around. Probably in endless loop, until growing winds shred it apart or fuel runs out.
And then, after overriding the fly-around protocol, another automagical feature kicked in and pressed the nose down. Right into the tarmac.
So in this situation, about the only way to salvage the drone would have been to disable all "idiot-proofing controls" and land with manual controls. Which it didn't have.
"As the article explained in length: leaving crash protection on would have resulted in drone doing another fly-around."
The _CORRECT_ protocol for ground conditions that your aircraft can't handle is to divert to another landing site. This applies whether the pilot is human of a computer.
The salient point is that the "pilot" is Ex-RAF.
This is a good example of why airlines no longer hire ex-military pilots(*). They consistently ignore safety rules and press on regardless.
It was known more than 5 years ago that US Army drones have a far lower crash rate on landing (it's almost zero) than USAF ones - and that's specifically because disabling the anti-crash systems or attempting to manually land a drone is a disciplinary matter in the US Army
(*) At least not without forcing them to go through civil flight school to get their bad habits knocked out.
> Is that this drone was of a horrible design.
Pretty much my takeaway as well.
> if Master Override is activated and one of the altimeters is malfunctioning, the Watchkeeper opens up its “ground touch” window from 1m sensed altitude to 20m sensed altitude. In other words, the drone might decide it has landed even when it is still 65 feet up.
Clearly whoever designed this was trying to solve a specific issue they predicted might happen, but didn't give enough consideration to what the actual ramifications might be
From the El Reg article, it looks to me like the pilots were trying to work around know problems with the automatic systems.
Hint: If you have an automatic system that you have to regularly override because the automatic system can't cope with a situation, then your automatic system might need revising!
It's worse than that, it was the people who make the sodding thing doing it but will they revise it, nooooo, not unless MoD stump up more money to fix something that should have worked in the first place. Don't get me started on the stupidity of using a LASER altimeter to judge landing height...
"Don't get me started on the stupidity of using a LASER altimeter to judge landing height..."
I think it depends upon how the laser altimeter is implemented. If the system just uses a single laser, in radar mode i.e. just relying upon the timing of returns from a single sensor stream, I can see how it might be degraded in the rain due to scattering and false returns, both from the ground and intervening raindrops. But that's foreseeable, so I'd expect them to use multiple beams in a combined direct/convergence scheme.
Such a scheme would split the initial single beam from the laser generator into five (or more) individual beams that would then be routed along fibre to the 'corners' and 'center' of the aircraft. Each of these sub-beams would then be split into a further sub-pair, with one beam in each sub-pair being used for direct measuring and comparison for consensus voting, whilst the other beam of each sub-pair would be directed to converge with each other, along the lines of the light system used by the Dambusters to ensure that they were at the correct AGL for the bouncing bombs.
Not really rocket science, or terribly expensive to implement.
'I think it depends upon how the laser altimeter is implemented.'
I'm 99.99% sure your idea is better than the one Thales have actually put on the aircraft, if only because there was an almost identical crash at Boscombe Down in similar conditions. Accident report for that one isn't out yet though as far as I know.
"... the stupidity of using a LASER altimeter to judge landing height" especially when it doesn't work too well with wet runways. AFAIK it rains more than a lot in the Aberpoth area - I've been there.
Things were much simpler in my youf, the drone that I worked on (to use the modern parlance) used a parachute when returning to terra firma. The parachute worked well, most of the time it worked well. La plus ca change la plus c'est la meme chose.
If your system like the one here in the States, the contractor built it to MoD (DoD in the States) specifications. It's also possible they tried to talk the MoD into something a bit different and were turned down. Government procurement seldom involves rational thought by the procurement types.
With 4 different equipments to retrieve the altitude, the system could sort the malfunctioning ones is a lot of cases. (3vs1 is trivial, 2vs1+1 might be manageable as well)
And for the landing detection : use Weight on Wheel instead of (or, better, in addition of) a fancy vertical speed detection! As even big commercial airliners get some trouble with air turbulence, no wonder a flimsy tiny drone (especially with the ground effect) get even bigger effects...
'Surely the "ultimate" landing confirmation sensor and far better than assuming you might be on the ground when possibly still 65' up in the air and nosing down to "make sure".'
No, I gather that the the nose down operation was to provide additional friction to improve braking*, not to "make sure" the drone had landed. It was applied after the drone was assumed to have landed.
*... Or more likely it's using the elevator for air resistance, to apply the equivalent of additional flaps. The article isn't clear on that and I don't actually know much on the subject.
Pitot Tube barometric pressure - I thought the pitot was specifically dynamic pressure, and you'd use the difference between it and the static (barometric) pressure to calculate airspeed. Always thought barometric pressure is from a ideways-facing sensor specifically to avoid dynamic pressure. Am I confused, or the author?
> I thought the pitot was specifically dynamic pressure, and you'd use the difference between it and the static (barometric) pressure to calculate airspeed
> Am I confused, or the author?
The author. You are entirely correct.
Also from the article:
"The Watchkeeper has four ways of determining altitude: barometric altitude, from the pitot tube; GPS altitude from its location unit; radar altitude, from the Watchkeeper system's ground radar unit; and its on-board laser altimeters"
- The pitot, as explained above, is used to derive indicated airspeed.
- Barometric altitude (i.e., altitude on a barometric datum) is provided by the altimeters and measured at one or more static ports, so called because they are kept, within practical limits, out of the aircraft's slipstream, typically on the sides of the fuselage.
- GPS provides ellipsoidal altitude, from which geoidal altitude may be derived via a mathematical model. This bears no relation to barometric altitude other than in a very approximative way.
- Radar altimeters, despite their name, do not provide altitude information at all. They provide height information, which in aviation means how far you are off the ground, while altitude is how far you are from a datum, usually an isobaric surface.
- Laser altimeters, I am not familiar with, not being a drone pilot and that system not being present in the civilian types I'm familiar with. If the description from the journalist is correct (and at this point, that's quite an assumption), they appear to be analogous to radar altimeters, although they could also be LIDAR units as found in other autonomous vehicles, which serve a different purpose.
Right - the static port reports the current pressure altitude, which is compared to the dynamic pressure from the pitot tube. By itself, the static port is sufficient for pressure altitude, which needs a correction to account for surface pressure altitude in order to determine an actual altitude measurement.
The author is not confused; without the static port the pitot tube is inconclusive about the dynamic pressure. It's just a contraction of pitot-static tube that makes it unclear.
" Unlike other drones, such as the MQ-9 Reaper, the Watchkeeper cannot be flown manually"
But apparently it can be crashed manually?
" the crew selected Master Override (MO) in the hope they could get it back on the ground as quickly as possible."
Can you get on the ground any faster than crashing?
@AC - 'It was just more "in" the ground than "on"'
Well, no, "no damage was caused, other than to the drone itself", so the very expensive, military-grade concrete runway was completely unscathed by the incident.
Continuing this analysis, "Nobody was injured", so, on the basis that any landing you can walk away from is a good one, this was a good landing. They got the drone down, as fast as possible, without damaging anything else, and walked away.
Many years ago, I recall reading that one of the problems with automation was the insistence of "experts" that they "could do a better job". I'm pretty certain that this resulted in at least one system with a "manual override" that did nothing apart from power a light which said "manual override".
The recent Tesla crash, and future shape of driverless cars are in the mix here.
And a more recent - but very real - phenomenon is the rise of the "I know best in the face of OVERWHELMING evidence to the contrary". (In a recent interview, the comedian Nish Kumar was accused of saying something in a show at the Comedy Store. Apparently the Comedy Store records all performers, and Nish was able to *prove* he did not say what he was accused of. To which the guy complaining said "I know what I know" - i.e. fuck facts).
Well in this case, the highly trained operators knew they knew best - or rather, they knew of limitations and some 'known issues' with the automated system.
The Tesla crash was probably caused by its user not using it as intended - i.e, it was designed to supplement and not replace his control. However, there is a school of thought, as cited by Volvo, that this 'half-way house' approach is potentially dangerous, since human nature is to lose concentration at times.
Biting the hand that feeds IT © 1998–2019