TWO SATELLITES COLLIDE IN ORBIT

My guess is that it will be the timing that's all important,
at almost 10 km/s,

Jan

J. J. Lodder wrote:

'Twas many moons ago that I worked on radar, but one of the places I did
work was at Eglin AFB, on the FPS-85 spacetrack radar built and used for
just that purpose. I believe it's still in use, and might still be the
primary way of tracking space objects. (See for example
http://www.fas.org/spp/military/program/track/an-fps-85.htm .) Software
processing has advanced tremendously since I was there, but the
positional resolution is still fundamentally limited by the beamwidth
and pointing accuracy. The way you tell where an object is, is to move
your antenna pattern around and see where you get the strongest return.
So the beamwidth as well as the pointing accuracy are the determining
factors in fixing a position. The beamwidth of that system is awesome,
but I wouldn't be surprised if it and the accuracy add up to enough
error to make it impossible to predict a position within a few meters.
Add ionospheric distortion to this, and I don't see any way a collision
could be predicted with any reliability.

Roy Lewallen

wrote:


A very good article on this is available at:
http://www.thespacereview.com/article/1314/1

--Rolf

wrote (with clarity & insight):



Excellent, thank you. This is the "big answer" I was hoping to find.

The main limitation on how many satellites can be screened by the US
military is trained personnel, with computational resources and sensor
network capacity being secondary and tertiary limitations. The
computers do all the heavy calculations, but it is up to the human
analyst to determine if the answer makes sense and is correct.
Intimate knowledge of the numerical techniques behind this process is
required, along with considerable experience with both conjunction
analysis specifically and space surveillance in general.

With rare exceptions, active duty military personnel do not develop
the skills to perform this task. At most, they will be in a particular
location for a three- or four-year tour and within that tour they are
likely to have multiple jobs. The military promotions system demands
incremental moves within a tour to various leadership, instructor, or
evaluator positions to demonstrate suitability for promotion. Staying
in the same job for an entire tour is considered evidence of someone
who doesn't have "the right stuff" and is a potential death knell for
any career. There is also a strong prejudice against analysts in the
officer ranks. ...

Suppose there was effective online (InterNet-based) training at
this sort of thing. Suppose that lots of otherwise unemployed
people were enticed (by offer of regular employment) into taking
the training then working at the job, so that instead of a few
highly paid military personnel who are shunned if they keep at the
job long enough to become expert, we have millions of random people
doing the work, with a way to determine who is doing the best work?
Would that fix the problem?

Many years ago (circa 1970) I invented a near-optimal algorithm for
sequencing flashcard-drill for maximum miagration of learning
through short-term and medium-term memory into long-term memory.
More recently, circa 1995, I developed software for coaching
short-answer questions. Coupled with the near-optimal sequencing
algorithm, this provides a very effective way to learn the correct
answers to questions. Other kinds of exercises, similar to those
developed during our CAI-Calculus project (1986-91) at Stanford
IMSSS, would be useful for converting memorized facts into
practical know-how. I believe my technology would be suitable for
training newcomers at this kind of task, to get them to the level
where they can be useful at judging the predictions of collision
probabilities given by the orbital models.

Several years ago, somebody else invented a new kind of futures
market, analagous to commodity markets (predicting price of hog
bellies or frozen orange juice a few months in advance), but used
for scientific predictions, such as the probabilty that somebody
will invent a solution to a particular problem within so many
months or years. By having people bet money on their predictions,
winning if their predictions are correct but losing money if their
predictions are incorrect, there's incentive to make good
predictions, and the market itself shows more weight to those who
make good predictions hence win bets hence build up their assets
hence overwealm the losers in establishing bit/ask thresholds.

In recent months, I've been planning to build a new kind of
economy, over the InterNet, a "cooperative" whereby people barter
their services. Part of this "cooperative" will be bidding on
contracts for services. All contracts are bid on amount of *time*
they will require to complete, rather than on real money. The
lowest bidder (least time to complete task) gets the contract, but
gets paid *only* if the contract in fact is completed before the
deadline (time started plus bid time-interval to complete equals
deadline time to end). Another part of this "cooperative" will be
predicting the probability of various things, such as whether
completed work was correct or not. People who make good predictions
will build up their account and be able to bid on larger contracts
or make larger wagers on probabilities. People who make poor
predictions will exhaust their small account and need to start from
scratch again, and again, and again, never having much effect on
the public bid/ask thresholds.

As of tonight, upon reading the above complaint about difficulty
keeping good personnel at the risk-analysis task long enough to
become expert, I feel my existing software for computer-assisted
instruction (flashcard drill and practical exercise drill), adapted
to risk-analysis training, followed by my proposed probability
futures market tied into my "cooperative" economy, adapted to
making wagers on the probability of various collisions, ought to
"fix" the problem, except that not everyone currently unemployed
would be willing to do lots of work assessing satellite-collision
risk, with no pay except credit towards computer services, no
actual *monetary* income to pay the rent. My proposed system plus
some government funding should *really* fix the problem, offering
tens or hundreds of thousands of unemployed people a chance to earn
some money, hence an incentive to *try* learning this skill,
resulting in thousands or tens of thousands of people who become
good enough to be of value at this kind of task.

for more info: tinyurl.com/urlgo
pass the quick Turing test to prove you aren't a botnet,
then enter 'neweco' for the new "collective" economy,
or 'uh3t' for my general contact info.