A crash course in Mars lander computational science...

louis · 2016-11-24 04:19:47

https://www.theguardian.com/science/201 … s-altitude

Sounds a bit like a schoolkid error? But what do I know!

elderflower · 2016-11-24 04:27:19

Blame the computer!

GW Johnson · 2016-11-24 10:05:48

So, do y'all still want self-driving cars and beer trucks?

GW

SpaceNut · 2016-11-24 10:39:31

Partial reporting of facts with no explanation of what events would cause the issue.

Which points back to english units of measurement, programing and metric units of measurement... Then to the device not working as it did not survive the trip due to not sufficient testing.....

louis · 2016-11-24 16:18:31

Safer than human-driven cars!

GW Johnson wrote:

So, do y'all still want self-driving cars and beer trucks?
GW

SpaceNut · 2016-11-24 20:42:13

Well GPS can be as far off as 500 feet or so and that is a huge deal when trying to land something on mars as you just crashed a very expensive piece of hardware if mars had GPS...
Stick to testing of software, devices and stick to one unit of measurement. Then we will have less crashes on mars.....

GW Johnson · 2016-11-25 16:31:45

"Safer than human-driven cars! smile"

I dunno about that, Louis. I've had a license to drive for 52 years now, and I never hit another one, and I have always avoided being hit by anybody else. Started driving 4-6 years before that under my Dad's instruction. That's quite the long record for any computer to beat.

Seriously, what seems to have happened to Schiaparelli is an error occurring that the programmers didn't anticipate. Not a units-of-measure error (which is prevented by adequate human-to-human communication, no matter what units we are talking about). Some sensor failed or fed bad data to the control system, and that problem grew as it rippled through, very rapidly to a fatal outcome.

These things can only do what they're programmed for. If something unanticipated occurs, bad things are going to happen. "AI" is still an oxymoron, and I don't see that changing anytime soon, even if some robot passes the Turing test.

And THAT is why I will not set foot in a computer-driven car or a computer-flown airplane. I have long thought Airbus made a very serious mistake with computer flight controls that over-rule the human pilot, rather than the other way around. This approach killed a test flight crew near Paris in the first flight demo of Airbus's first 4-engine airliner; many, many years ago.

It happened again a few years ago off Brazil. All three pitot-static ystems froze up and fed bad data to the flight control computer system. The pilots were too unused to hand-flying the airplane to respond correctly to the problem. The result was falling in a stall 6 miles to the Atlantic, pulling nose-up pitch commands to the yoke all the way down. Inexcusable.

GW

Last edited by GW Johnson (2016-11-25 16:38:26)

louis · 2016-11-25 18:44:24

LOL - you should try UK London traffic rather than those wide open, straight Texan highways!

I saw a programme here on BBC re the moon landings. Something similar happened with Apollo 11 - I think the programming was out and that led the lander into an area full of boulders (owing to a speed miscalculation) which had to be manually avoided. Had it been a robot craft, disaster would probably have ensured. But Apollo 11 was nearly 50 years ago...you'd have hoped ESA would have learnt by now.

Anyway - all jet airliners are computer flown now!

GW Johnson wrote:

"Safer than human-driven cars! smile"
I dunno about that, Louis. I've had a license to drive for 52 years now, and I never hit another one, and I have always avoided being hit by anybody else. Started driving 4-6 years before that under my Dad's instruction. That's quite the long record for any computer to beat.
Seriously, what seems to have happened to Schiaparelli is an error occurring that the programmers didn't anticipate. Not a units-of-measure error (which is prevented by adequate human-to-human communication, no matter what units we are talking about). Some sensor failed or fed bad data to the control system, and that problem grew as it rippled through, very rapidly to a fatal outcome.
These things can only do what they're programmed for. If something unanticipated occurs, bad things are going to happen. "AI" is still an oxymoron, and I don't see that changing anytime soon, even if some robot passes the Turing test.
And THAT is why I will not set foot in a computer-driven car or a computer-flown airplane. I have long thought Airbus made a very serious mistake with computer flight controls that over-rule the human pilot, rather than the other way around. This approach killed a test flight crew near Paris in the first flight demo of Airbus's first 4-engine airliner; many, many years ago.
It happened again a few years ago off Brazil. All three pitot-static ystems froze up and fed bad data to the flight control computer system. The pilots were too unused to hand-flying the airplane to respond correctly to the problem. The result was falling in a stall 6 miles to the Atlantic, pulling nose-up pitch commands to the yoke all the way down. Inexcusable.
GW

elderflower · 2016-11-26 05:26:12

Not just airliners. Fighters are also computer controlled these days. The complicated software leads to very extended development times and some losses before it gets sorted out (new Lockheed Lightnings and Saab Gripens for a couple of examples). For a one off Mars lander you can't make multiple units and write a couple of machines off, nor can you intervene in real time.
Where people are involved in space there must be manual override. Manual docking is used from time to time at the ISS when the computer system isn't managing the process properly, and then there was Apollo experience as well.

GW Johnson · 2016-11-26 10:12:52

Elderflower:

Yeah, I know about the fighter problems. There's been good reason to believe (for about 40 years now) that a simpler airplane and a more-experienced pilot will do just as good in combat as these computerized monstrosities. Such as Red Flag 2 at Nellis in the early 80's. The then-new F-15's were beaten soundly by the aggressor squadron flying relatively crude and uncapable F-5's, once equipped with radar warning receivers modified from Radio Shack. No software development at all there.

That convinced me that a Korean War-vintage F-86 equipped with wing stations for missiles, and a cheap radar warning receiver, and flown by an experienced pilot, could take on anything in anybody's air forces today. You need durability, maneuverability, Mach ~1, a mix of guns and missiles, and you need to know when somebody else's radar paints you. All else is BS (to misquote the Red Baron).

The only reason complicated software is needed is for good flight control and handling, out of a two-engine airplane otherwise too large for the mission. So don't do that! And for goodness sake, no more multi-service airplanes! It didn't work with the F-111, why on Earth would it work with the F-35 now? The air force and the navy simply have incompatible requirements. (Marines fly navy airplanes off of navy ships, doing both navy and air force-type missions, just shorter ranges).

The focus for some decades now has been on building these monstrosities instead of useful combat aircraft, because it is far more lucrative for the industry giants. That's the giant corporate welfare state influencing things, because their lobby money owns the government that contracts with them. Corrupt as all hell, and has been building up since WW2. Has spread far beyond defense. Needs to be done away with. Simple as that.

Louis:

I have lived all my life in what became the I-35 corridor, which is now the most dangerous and deadly road in the entire nation. I learned to drive in downtown Dallas, Texas, before there ever were any freeways, and came of age driving those same freeways as they were built, a process taking years. Construction zones are particularly hazardous, being cramped, sinuously twisted, and poorly marked, while no one wants to slow down.

Driving the long straight highways out in west Texas is a real treat. A vacation of sorts, from the urban traffic hazards. Plus, I learned "combat driving" in a small, incredibly-underpowered car: a 36 HP VW beetle, with a top speed under 70 mph, and which took "all day" to top-in at that speed. I must have learned survival skills well: both myself and the little car survived. In fact, I still have the little thing in mothballs out here on the farm (and it still runs fine). It's 56 years old now. I'm 10 years older.

As for the airliners, I dislike having pilots unused to hand-flying their airplanes under all circumstances of the flight. It's a serious training lack that has already killed a lot of people. The ones from Boeing up through B-767 still let pilots over-ride the computer anywhere, anytime. I don't know about the B-777 or B-787. But I certainly hope the pilots still have ultimate authority. I just wish they did more hand-flying for the practice.

GW

Last edited by GW Johnson (2016-11-26 10:26:50)

Terraformer · 2016-11-27 11:57:58

Veering further off topic, but...

That convinced me that a Korean War-vintage F-86 equipped with wing stations for missiles, and a cheap radar warning receiver, and flown by an experienced pilot, could take on anything in anybody's air forces today. You need durability, maneuverability, Mach ~1, a mix of guns and missiles, and you need to know when somebody else's radar paints you. All else is BS (to misquote the Red Baron).

I wonder if part of the problem is that the air forces are out of practice? When was the last time there was a real aerial battle? It seems most of the time, they're flying bombing missions and air support against less capable militarises, where a supersonic interceptor isn't well suited to the role...

GW Johnson · 2016-11-27 17:04:44

Hi Terraformer:

You are exactly right about "lack of practice". Typical air force front line pilots in the west have maybe 2-3 years before they rotate out of actual flight duty. It's worse in the Russian east, they don't fly as much as western pilots, even during those 2-3 years.

The kind of thing I refer to as "experienced pilot" is one with 15+ years flying experience in the appropriate aircraft types. These are quite rare. You find them in the US reserves and national guard, and to some extent, in the fighter training schools.

Guys like that (or gals, but there are very precious few of those, unfortunately) can do things at well under 5 gees, what most line pilots cannot do maneuvering at up to 9 gees. They do not need high-gee airplanes. Nor do they need supersonic cruise, or even high dash speeds. They need only the things I listed before: maneuverability to around 6-7 gees, a mix of guns and missiles, combat speeds near Mach 1, and a radar detector to know when they are being painted by a hostile radar.

The Israeli combat air force is but one example of the skills I am talking about. Our "aggressor squadron" training pilots are another such group, at least to some extent.

Stealth is quite overblown as a combat advantage for experienced pilots like that. For one thing, stealth is very frequency dependent. What looks like an insect at battlefield 100-MHz frequencies looks like an ocean liner at kilohertz search-track frequencies. The difference is only local pixel resolution: a couple of miles at KHz, several meters at 100-MHz.

So what? Visual range is a few miles. See what I mean?

GW

Last edited by GW Johnson (2016-11-27 17:07:38)

elderflower · 2016-11-30 11:50:31

I rather think that the most recent proper air battle was over the Falklands, where Harriers faced Mirages and A4s. The A4s were outclassed as you would expect, but so were the Mirages, despite having supersonic capability. The long range didn't help them when they came to combat range as they didn't have a lot of spare fuel. This is something that almost any agressor is going to face.
Sharkey Ward's book is a good read and instructive.

SpaceNut · 2016-11-30 21:34:36

I think earth aircraft is a bit off topic but yes it does come down to instrumentation and sensors.

I look at it this way since we have had 2 crash the third one should be a charm of a landing....

Why is the cycle of landers and probes to mars going at a snails pace....We should be seeing way more than 1 every launch cycle...

GW Johnson · 2016-12-01 10:24:55

Whoops, sorry. We wandered far afield from the computer control difficulties experienced fatally by the probe. A quote of myself from post 7 above:

"Seriously, what seems to have happened to Schiaparelli is an error occurring that the programmers didn't anticipate. Not a units-of-measure error (which is prevented by adequate human-to-human communication, no matter what units we are talking about). Some sensor failed or fed bad data to the control system, and that problem grew as it rippled through, very rapidly to a fatal outcome.

These things can only do what they're programmed for. If something unanticipated occurs, bad things are going to happen. "AI" is still an oxymoron, and I don't see that changing anytime soon, even if some robot passes the Turing test."

Between things like this probe crash, and my bad experiences with desktops, laptops, and software, I am amazed that as many of the space probes have worked as well as they have. That success seems predicated upon being able to anticipate everything that might go wrong. If you don't successfully do that, then what you overlooked will go wrong, that's Murphy's Law.

I think folks trust their lives to computers entirely too much. How well one might emulate human behavior or responses is irrelevant. They cannot think and they cannot learn the way we do. They can only execute their programs, even if they are empowered to adjust them (a pale imitation of learning).

As for the paucity of probes sent each opposition, I think you can blame that on the corruption of politicians by big lobby monies of outfits uninterested in science or space, only their bottom lines. I believe it is in fact why we in the US were given two such distasteful candidates to vote against. Real change is less profit for the giants who bought the jobs of all these officials. And the US is not the only country that suffers from this.

GW

Last edited by GW Johnson (2016-12-01 10:27:17)

SpaceNut · 2016-12-01 19:58:17

Having grown up with computers, programming and troubleshooting hardware as well as the firmware I understand all to well when it comes to firmware and data in that garbage in gets mostly garbage out..... The same holds true of a quality control program as its only as good as the testing that is put in to a product as to how well it will make it in real life situations.....

New Mars Forums

Announcement

#1 2016-11-24 04:19:47

A crash course in Mars lander computational science...

#2 2016-11-24 04:27:19

Re: A crash course in Mars lander computational science...

#3 2016-11-24 10:05:48

Re: A crash course in Mars lander computational science...

#4 2016-11-24 10:39:31

Re: A crash course in Mars lander computational science...

#5 2016-11-24 16:18:31

Re: A crash course in Mars lander computational science...

#6 2016-11-24 20:42:13

Re: A crash course in Mars lander computational science...

#7 2016-11-25 16:31:45

Re: A crash course in Mars lander computational science...

#8 2016-11-25 18:44:24

Re: A crash course in Mars lander computational science...

#9 2016-11-26 05:26:12

Re: A crash course in Mars lander computational science...

#10 2016-11-26 10:12:52

Re: A crash course in Mars lander computational science...

#11 2016-11-27 11:57:58

Re: A crash course in Mars lander computational science...

#12 2016-11-27 17:04:44

Re: A crash course in Mars lander computational science...

#13 2016-11-30 11:50:31

Re: A crash course in Mars lander computational science...

#14 2016-11-30 21:34:36

Re: A crash course in Mars lander computational science...

#15 2016-12-01 10:24:55

Re: A crash course in Mars lander computational science...

#16 2016-12-01 19:58:17

Re: A crash course in Mars lander computational science...

Board footer