For those of you too young to remember, the Strategic Defense Initiative (SDI) was proposed by President Ronald Reagan on March 23, 1983.
As the story goes, Mr. Reagan enjoyed watching old science fiction movies. And he got this wild idea of shooting nuclear missiles out of the sky with laser beams. The idea, which was also called “Star Wars” (note the sci-fi reference), was all but impossible at the time, and made us the laughing stock of the world. It also threatened to militarize space, which would be even worse.
Anyways, it is now 30 years later. And few people even recall the program (I assume the next president, President Bush the elder, quietly swept it under the rug when he came into office). But I have just one question: Is it anymore feasible now, this idea of lasers shooting down missiles?
I know there are many examples of technology that 30 years ago would seem almost impossible. And computer computational skills (if that is relevant) is far more advanced now. So, is a program like SDI more feasible now?
Very relevant, and still just as big a killer as before. Back in 1985 David Parnas made a huge contribution, and very publicly pointed out that the software needs of SDI were utterly infeasible.
David set on a world tour speaking about this issue, and I well remember him visiting us down here in Oz. Even to an audience of geeky computer scientists, who were not all that keen on hearing they couldn’t do stuff, he made a resounding impression. The reality is that essentially nothing has changed since then.
Sadly in software there has been nothing like the progress many people imagine, and although computers are vastly faster, if anything our ability at creating reliable software is unchanged.
Even if the money could be found, the science sorted out, and the space borne x-ray lasers perfected, there is zero chance you could create a software system that was reliable enough to actually put in control of such a system. Most damming:
It must work perfectly the first time it is needed.
It is intrinsically impossible to test.
Back in the 80’s there were high hopes for advanced computational aids to programming that might help. (Automatic program writers, advanced AI systems and so on.) Parnas was very scathing about the prospects of these ever working. He was right. None of them ever eventuated. About all that has changed since then is that we are actually a lot more modest about what we think we can realistically achieve.
Cancelled in 2011 after years of massive budget overruns and technical failures.
Not a strategic ABM interception system.
After eight years of “operational capability”, $40 billion dollars, and numerous test failures, the system finally successfully intercepted a representative target earlier this year in highly orchestrated threat scenario. The system (and particularly the acquisition system) remains incompletely tested and the troubled Exoatmospheric Kill Vehicle (EKV) which actually performs the physical interception has had continuous problems with the sensor package not functioning correctly.
There have been some significant advances in the capability of theater-level interception, and we had a working point defense system against strategic attack in 1975 (albeit, at great cost and suitable only to defend relatively compact installations, resulting in cancellation).
SDI, as envisioned and promoted by Reagan, remains elusive as do many of the enabling technologies (directed energy weapons, highly autonomous early warning and command systems, single stage to orbit launch vehicles, invisible space unicorns) which were required to support it.
The software system includes the generation the “threat indications” - ie process sensors data in real time. You need to test the system end to end. Just testing a portion by feeding it synthetic inputs that would come from another software system in an operational deployment isn’t doing an end to end test.
The problem is that you must assume that the attack will include as many countermeasures to SDI as the attacker can think of - decoys, attacks on sensors systems, attacks on communication systems. Everything is going to be noisy, there will be random and uncontrollable outages. You have no useful way of modelling this, and the complexity is simply astronomical. The system is vastly larger and more complex than first sight might suggest.
Testing of large real time control systems (heck, almost any real time control system) is difficult at best. You will spend more time writing test systems than the actual system being tested. The range of possible inputs a SDI system must cope with would take until the sun goes cold to enumerate. Any idea of proper code coverage is fanciful. The scale of SDI, even for a project mooted in the mid 80’s makes any software system we have now look puny. Just look at abject disasters like the FAA’s advanced automation system to see how things can go wrong.
The fundamental strategic issue hasn’t changed either. A working SDI system might actually be more dangerous than a non-working one. The threat of retaliation has always been a check on military actions. Make it possible for a nation to attack other nations while it’s immune to counter-attack and you greatly raise the temptation to launch attacks. And while I’m as patriotic as the next guy, I don’t believe the United States is immune to this.
It’s still around in the guise of the Missile Defense Agency, renamed under Pres. Bush II from the “Ballistic Missile Defense Organization” (BMDO), which was so renamed under Clinton from the “Strategic Defense Initiative Organization”, founded under Reagan in 1994.
As far as the feasibility goes, insofar as technology has advanced in the intervening 30 years, it’s more feasible now than then, although in terms of the original vision, it’s probably not very feasible even now.
However, a lot of advances in related fields such as high energy physics, supercomputing, advanced materials and engineering were funded by the SDIO/BMDO/MDA as part of their mission.
Testing software simulations doesn’t mean that one has actually tested the system.
The main challenge is that the geometry of any national missile defense test is always hosed. Basically, a target is launched from the Pacific that will land in the ocnean, and an interceptor is launched from California to blow it up. In a real scenario, a missile is launched from North Korea to land in Seattle (or whatever) and an interceptor is launched from Alaska. We just can’t test the real scenario because we can’t shoot missiles at the continental United States from the places we expect the missile to actually come from.
Right - but the assertion (in post #3) was that the software is “intrinsically impossible” to test.
Why couldn’t a test missile be launched from a submarine?
And if the system demonstrated an ability to reliably cope with test missiles arriving from, say, the west (e.g. Hawaii), does that not suggest it might be able to handle those arriving from other directions?
In baseball the batter hits a ball that has to be aimed within a certain path.
If the pitcher could throw the ball anywhere in a 180 degree arc in front of him and the batter would still have to hit it, it would be a much different success rate.
throwing a target into a known trajectory is much different than trying to cover hundreds if not thousands of possibilities.
Exactly! The software is not the system. The system is the software, the hardware on which it executes, the input sensors, the output effectors, the human operators, and all of the external influences. If you want safety and reliability, you must design and test all of these things together. It is expensive.
There’s the problem.
No system engineered in the real world could hope to meet this. If that’s what they were expecting, they were guaranteed to be disappointed.
Israel has it’s Iron Dome. It’s proven reasonably effective against primitive, slow missile /rockets. Thankfully they aren’t being shot at by nuclear ones (yet).
WAG wouldn’t Israel be considered on the forefront of this type of anti-missile technology research? They’ve got the biggest incentive to improve their system. It’s literally a matter of life or death for them.
Norman Spinrad, the very leftist SF writer, did a piece on SDI for LeMonde that lays it mostly on Pournelle. Pournelle has always said that Spinrad got it wrong in every way.
Pournelle was a member of the Citizens’ Advisory Council on National Space Policy, a fancy name for a group of right-wing SF writers who wanted to boost space and fight the cold war.
The Council did write the draft for Reagan’s SDI speech, though Pournelle insists that it was rewritten by Reagan.
Where this puts Pournelle is cloudy. To my knowledge he’s never stated he’s SDI’s father, although he most certainly worked hard to get it into reality. It’s certain that Reagan had already been made aware of the concept through talks with other scientists, but having an advocate group always is the way to get policy made.
The way I read Spinrad, his only mistake was saying that the Council was a trick to get funding for space programs, whereas Pournelle claims that they really wanted to kick commie ass. Who was right politically is better left for a GD thread.
It is even worse than this. An even moderately sophisticated threat will be designed to look like something innoculous; for instance, an ICBM will be designed to look like a space launch vehicle (using the fractional orbital bombardment schem), or a low flying cruise missile will be designed to behave like (and have the radar cross section of) a small jet. Since the system will be designed to discriminate such signals as non-threats by default (lest it attack non-threats) defeating the system is a matter of not looking like a threat, or alternatively, overwhelming the system with apparent threats (decoys and penetration aids) such that it can’t figure out what to hit.
This is why the Ground Based Midcourse Defense (GMD) system is caveated as only protecting against launches from “rogue states”. It’s not that North Korea or Iran are inherently more likely to initiate an attack on the United States and its allies than Putin-era Russia or China, but rather, the system is only capable, even in principle, of effectively intercepting the most primitive attacks. Pen aids, decoys, and other countermeasures are far easier to develop than the methods used to detect and discriminate against them. In essence, GMD is capable of protecting against 'Sixties ear ballistic missiles which is only useful if someone is trying to fly an R-7 at you.
The discussion about having to test the system end-to-end to assess effectiveness is also on point. Much criticism is leavied against many ABM tests because they are “artificial” in the sense of only testing certain portions of the system such as the tracking radar, the booster, or the seeker sensor, without even attempting to conduct an actual intercept. It should be understood that this kind of development testing is crucial to characterizing and improving the performance of individual subsystems and elements. However, when you’ve deployed a system without conducting a successful “threat-like” intercept and avoid doing so, it gives the strong appearance (backed up by analysis of vulernability gaps in testing or performance, as vocal ABM critic Ted Postol has written extensively about) that even if all of the components perform up to specification there is low confidence in effectivenss even under the constrained conditions of a “threat-like” test.
And those constraints (that the target is a specific configuration, coming from a known sector within a specific interval of time) fail to test the system under a true threat scenario with little or no prior warning. For battlefield and theater systems which will operate under some known conditions, e.g. that the general direction of the threat will be known, that they’ll be on alert when an attack is likely to occur, and that anything in the designated zone that doesn’t have an IFF response may be considered a threat. In a strategic system, which has to be both “always on” and “fail safe” from accidentially intercepting a non-threat, building all of the intelligence to autonomously discriminate friend from foe and then testing it in real world conditions to the point that you have high confidence in the effectiveness of the system is essentially impossible and egregiously expensive to conduct. Even the simple, constrained tests (which are often launched either from Meck Island, Kauai, or air dropped from C-17 aircraft on a pallet to replicate a somewhat realistic threat trajectory) are often hundreds of millions of dollars to conduct, and the five target gang-bang test conducted by MDA back in October 2012 (FTI-01) is estimated to have cost on close order of half a billion dollars and didn’t even include ICBM-class targets.
Testing such a system by software simulation only (even a so-called “hardware-in-the-loop” test in which hardware is fed signals that make it appear that there exists a threat) can verify that it will perform as expected to known inputs without spurious behavior, but in no way validates its operations under real world conditions in which a threat may look or behave differently than expected, failure of communication links or control authority may occur, and environmenal conditions or other “fog of war” issues (the “unknown unknowns” in Rumsfeldspeak) may result in unexpected system errors which cannot be vetted by software simulation alone.
Although Jerry Pournell was a booster of the Strategic Defense Initiative (albeit largely as backdoor support for his own hobbyhorse, large scale space development and single-stage-to-orbit launch vehicles) it was really Edward Teller, along with Lowell Wood, both of Lawrence Livermore National Labs, who campaigned for and promoted SDI, bringing it to the attention of the Reagan Adminstration. The concept was not popular with many in Reagan’s cabinet as it posed both a change from the offensive capability buildup–they favored broad scale deployment of the MX (later “Peacekeeper”) MIRV-capable system in a variety of increasingly absurd basing scenarious including “neutrally buoyant capsules” floating in Lake Michigan and silos bored a mile into the Rocky Mountains–and was potentially destabilizing against the relative balance of forces; however, Reagan was enamored with the notion of providing a “protective shield” against nuclear attack (he even offered to provide the technology to the Soviets as concessions to reduce or eliminate nuclear arsenals) and was not sufficiently technically astute to appreciate the implausibilities of bringing the critical technologies to fruition in any foreseeable interval. The most viasble of all proposals–which later became the “Brilliant Pebbles” orbital kinetic interceptor concept–was deemed technically feasible but would cost hundreds of billions of dollars to field and maintain, and would of course violate several treaty obligations. Post-Soviet Union, it was quickly selected for the chopping block in favor of smaller scale projects like THAAD, modernization of Aegis, and what became GMD and the (cancelled) KEI. So it goes.
