Mars Failures: Bad luck or Bad Programs? 389
HobbySpacer writes "One European mission is on its way to Mars and two US landers will soon launch. They face tough odds for success. Of 34 Mars missions since the start of the space age, 20 have failed. This article looks at why Mars is so hard. It reports, for example, that a former manager on the Mars Pathfinder project believes that "Software is the number one problem". He says that since the mid-70s "software hasnâ(TM)t gone anywhere. There isnâ(TM)t a project that gets their software done."" Or maybe it has to do with being an incredible distance, on an inhumane climate. Either or.
almost /.dotted (Score:2, Informative)
Why is Mars so hard?
by Jeff Foust
Monday, June 2, 2003
This June will see the beginning of the most ambitious exploration of the Red Planet in a quarter-century. If all goes well, three launch vehiclesâ"one Soyuz and two Deltaâ"will lift off this month, placing four spacecraft on trajectories that will bring them to Mars by this December and January. Those spacecraft include the first European Mars orbiter, Mars Express; Beagle 2, the British lander built with a mix of public and private funding; and NASAâ(TM)s twin Mars Exploration Rovers, perhaps the most advanced Mars spacecraft even built. They will be joined at Mars by Nozomi, a Japanese-built Mars mission launched in 1998 and forced to take the long road to Mars because of thruster problems.
This should be an exciting time for those interested in Mars exploration, and for scientists and activists alike, it is. If these missions are successful, they should offer new insights about what happened to the planetâ(TM)s water and the potential for past or even present life there: some of the most important questions in planetary science and astrobiology today.
The catch is, if these missions are successful. The history of robotic exploration of Mars, stretching back more than four decades, is littered with failed missions and dashed hopes. Some of these failures can be chalked up to the growing pains of early planetary exploration, when a wide variety of spacecraft of all types failed. Others, particularly the 1999 failures of NASAâ(TM)s Mars Climate Orbiter (MCO) and Mars Polar Lander (MPL), are more indicative of management, programmatic, and other problems, rather than purely technical issues. Understanding these problems, and acting to correct them, are critical if current and future missions are to succeed in studying the Red Planet.
The star-crossed history of Martian exploration
Mars has been one of the most popular destinations for missions beyond the Earth. Since 1960 the United States and the former Soviet Union have launched 34 missions to Mars: 15 by the US and 19 by Russia and the former USSR. NASAâ(TM)s success rate is not too bad: nine of those 15 missions, including the Mars Global Surveyor and 2001 Mars Odyssey missions still in progress, can be considered successes. Russiaâ(TM)s luck has not been nearly as good: 14 of its 19 missions failed, and only oneâ"Zond 3â"can be considered a complete success; the remaining four are, at best, partial successes. Overall 20 of the 34 American and Russian Mars missions, or 59 percent, failed.
Four of the seven NASA Mars missions since Vikingâ"Mars Observer, MCO, MPL, and Deep Space 2â"have failed.
Digging into those statistics in greater detail shows some interestingâ"and troublingâ"trends. Many of the failed missions, particularly those launched in the 1960s, were lost because of launch vehicle failures, not because of any fault with the spacecraft itself. Many Russian spacecraft, from the earliest âoeMarsnikâ missions of 1960 to Mars 96, either failed to leave a parking orbit around the Earth or never made it into Earth orbit into the first place. However, in the last 30 years only one mission out of 16 attemptedâ"Mars 96â"was lost due to a launch vehicle malfunction. This can be most likely attributed to the maturity of launch vehicle development, including the use today of vehicles whose designs date back literally decades.
The problem with Mars exploration now appears to be with spacecraft themselves. Four of the seven NASA Mars missions flown since the twin Viking missionsâ"Mars Observer, MCO, MPL, and Deep Space 2â"have failed, all due to spacecraft problems of one manner or another. (MCO is a borderline case, since there was no technical problem with the spacecraft itself, but rather with how ground controllers operated it.) The only other NASA Mars missions to fail, Mariner 3 in 1964 and Mariner 8 in 1971, were each lost due to launch veh
Re:Methodolgies (Score:3, Informative)
http://www.fastcompany.com/online/06/writestuff
That's what they do, and I'm glad I don't.
And as for domain expertise not counting for much, that may be true for some domains, but sure as hell is not for mine (medical informatics).
Re:We landed on the moon with 512 bytes of RAM (Score:4, Informative)
no, for instance the Mars Pathfinder spacecraft had "128 Mbyte mass memory" and used a R6000 computer. While the rover had "0.5 Mbyte RAM mass storage" The R6000 is much less powerful than the original pentium.
http://mars.jpl.nasa.gov/MPF/mpf/fact_sheet.htm
NASA computer technology has for the past decade or two been a few or more years behind the state of the art in consumer electronics. Largely because they have to put the electronics through more testing and only use chips that will withstand possible radiation with low power consumption. Plus add on the years of development of the spacecraft itself... means that your desktop probably (Anyone want to do the math?) has more computing power than all the deep space explorers ever launched, combined.
MUST READ on NASA software!!! (Score:2, Informative)
The problem with most Mars programs is that the code seems to be developed like code everywhere else. Budgets overruns, working late to meet deadlines, and generally living the 'coder life.' This is NOT now critical software needs to be developed, and in fact isn't how most software should be developed.
To those proposing the 'more eyes open source' model, consider this: There's nothing in that model that GUARANTEES formal and complete code review. Something more rigorous is needed for projects like this.
Things to consider... (Score:3, Informative)
Before we continue to crucify programmers, we need to remember how hard it is to really get to Mars, from a purely spacefaring perspective.
From my experiences flying to Mars in Orbiter [orbitersim.com] space flight simulator (FREE!), several problems become apparent:
Mars is a fantastically difficult target to reach for two main reasons. It has very little gravity, and very little atmosphere.
If you shoot for something big, like Jupiter, you find that it is hard not to miss it. It's gravity well is so massive that navigational errors en route are relatively insignificant. Mars doesn't help you very much in this regard. An Earth to Mars flight has to be dead on.
When you get there, you are likely going to want to use the atmosphere to do at least part of the braking maneuver to get into Mars orbit (as most modern probes do). The problem is that Mars has a very thin atmosphere. Think about the sheet of paper analogies with Earth re-entry. Earth's atmosphere goes MUCH farther into space than does Mars'. You have to get dangerously close to the surface (within 50 miles) to effectively aerobrake using Mars' atmosphere. So with Mars, you are more talking about a near-ephemeral gossamer thin 1 cell thick membrane you have to hit the edge of rather than a nice, thick piece of paper.
Re:We landed on the moon with 512 bytes of RAM (Score:5, Informative)
A bit of advice: Leave the covers on, but make sure that you have enough case fans to ensure that the CPU has a constant air current over it. I have the fan on the front of my box blow in and the fan on the back (plus the power supply) blow out. If you leave your case closed, the improved air flow will actually lower the temperature of your CPU and motherboard.
Problem is the statistics are biased (Score:5, Informative)
Take their early record, before Mars 1 got to Mars, they had had a series of attempts. Two, known to the West as Mars1960 A and B reached Earth orbit then disintegrated.
Mars1962 A exploded in orbit at the height of the Cuban Missile Crisis - briefly causing a panic with the Americans thinking a missile attack was underway. Fortunately the computers soon told them that doomsday had been averted.
Next, was a partial success - Mars 1. Which smashed the record for deep-space communications with Earth across a distance of 106 million kilometres. Unfortunately it failed just before reaching Mars.
Mars1962 B exploded in Earth orbit and didn't appear in the Soviet record.
November 1964 saw the launch of Zond 2, a highly advanced probe using ion thrusters to perform stabilisation and orientation tasks. It may have also been the first probe to carry a lander. It died a long and lingering death before sweeping past Mars at only 1400 km altitude. (By this time the US had got their first Mars probe to the planet in working order, Mariner 4 took 22 pictures of the planet from 10 000 km. (Its sister ship, Mariner 3 had failed en-route)).
Neither side went to Mars in the next launch window, but 1969 was a busy year. Three attempts for the Soviet Union, including at least one lander. Mars 1969A exploded in flight as did Mars 1969B. Mars 1969C was removed from the pad after cracks developed in the relatively new Proton rocket design. (Cracking in the Proton was also a major reason for the failure of the Soviet Union to send a manned mission around the Moon during 1969). The US had a twin success with Mariners 5 and 6 flying past Mars.
On to 1971 and a pair of launches for the US, Mariner 8 ended up in the Atlantic, Mariner 9 went on to become one of the most successful missions ever and the first probe to orbit Mars. For the Soviets - mixed results again. Their first mission reached Earth orbit, but went no further and was named Kosmos 419. But then both Mars 2 and 3 left Earth orbit. They each comprised of a lander and an orbiter. The two craft jettisoned the lander before entering Martian orbit - just as the planet entered an intense dust storm with raging winds and almost total blackout.
Mars 2's lander was apparently DOA, it remained silent and does not appear to have returned any data. It was however the first craft to hit (not land on) Mars. Mars 3's lander was more successful. It entered the atmosphere, deployed parachutes and landed on rockets. It deployed its antenna and began to transmit the first picture from the Martian surface. Sadly, just 20 seconds later the transmission stopped. The Soviets said that the lander's parachutes had been caught by the storm and pulled it over.
Mars 2 and Mars 3 orbiters remained on-line and performed experiments on the Martian atmosphere and took photos of the surface. So I would call both missions a partial success and Mars 3 almost a triumph.
The next window was 1973 and the Soviets planned no less than 4 missions to Mars. Mars 4 and Mars 5 would be orbital missions, studying the planet much like Mariner 9, but also serving as telecoms relays for the Mars 6 and Mars 7 heavy landers.
Incredibly, bearing in mind the past track record of the Soviets, all four missions reached Mars in working order. Then everything went wrong. Mars 4's main engine failed and the probe did not enter orbit, it relayed images of the planet as it swept past into solar orbit. Mars 5 was next and was the only unqualified success of the year; it was the first craft to return colour images of Mars.
The two landers then arrived, Mars 7 first, it deployed the lander, but an attitude problem meant that the lander actually missed the planet entirely! Mars 6 was more lucky, the probe entered the Martian atmosphere, took readings all the way down and went dead ab
Re:Small Simple... Solid State (Score:5, Informative)
The GCs read only memory consisted of a series of peg-boards into which the code was wire wrapped (by hand). There were 74,000, 16-bit instructions that could be programmed in this way. There was 4k iron-core memory in the computer. There were two GCs used in Apollo. The CSM one was responsible for leaving earth orbit, mid-course correction(s), entering lunar orbit, etc. The LM GC controlled descent and ascent as well as autopilot functions for lunar orbit docking. The computers ran the programs for these manuevers from ROM, but using astronaut input parameters using the "noun-verb" input methodology.
The software was actually very sophisticated and did not consist of simple control loops - joystick feedback was actually processed to ensure commands kept the spacecraft within limits. The most important parameter was keeping the antennae pointed at the Earth.
AFAIK, there are no space-qualified Intel built '486s. There are space-qualified computer systems with '486s in them, which may seem like semantics, but these systems typically employed multiple '486s, with bus operations and data continually compared to look for differences indicating upsets. This is a point that always confuses people because at one point IBM/NASA indicated the AP101 Block IIs had the same amount of power as a '486 - this seems to be misinterpreted as the AP101s have '486s built into them.
Half a lifetime ago, I helped with some hardware failure analysis for the IBM Orbiter Computer Systems Group (It was an intermittently failing memory board on STS-4) and I have to say that they were the most impressive software group that I have ever been associated with. They learned their skills with the Apollo CSM/LM GCs and Apollo Instrumentation Ring - you just don't make mistakes when the instructions are wire wrapped. The software engineers that worked on the shuttle software didn't have a problem with going with the (relatively) complex AP101s (originally designed for the B-1). Going from wire wrapped ROM to battery backed RAM was seen as a good thing, but it did not mean that the software development process changed in any way.
I'm trying to remember if there were two or three support binders for each module of software in which the requirements were clearly defined, the science and reference information provided, all calculations/constants defined to support the software binder. Coding is always the last thing that is done and only if the support binders are complete and signed off. This process is very expensive, but the software produced is essentially perfect (I believe that there has been one non-safety of flight software error in shuttle history and several hundred thousand lines of code). Complexity isn't the issue.
I think the issue is, is there a software development methodology/process that fits in with NASA's "smaller, better, cheaper" and produces the same quality as the Shuttle/Apollo?
myke
A more detailed comparison (Score:3, Informative)