The path to hell is always paved with good intentions (The government is your friend.)

The comment by Karl Crary (scroll down to his comment on March 19, 2004) shows that irrespective to the many tests one can imagine, you are never in a condition to test all your variables, especially when it comes to an autonomous vehicle. In effect as Karl says " In the race, DARPA gave them a course not much wider than the vehicle itself, so the error of the GPS was a significant factor. (This was apparently in an attempt to be helpful: the course was narrow enough to miss all major obstacles.) This revealed a bug that had never come out in testing: they did not compensate properly for the error in GPS. As a result, Sandstorm found itself adhering tightly to a course a few meters off from where it was supposed to be. The vehicle correctly identified the obstacle that killed it, but thought that it was not allowed to steer around it." the precaution DARPA took killed the vehicle's ability to do real autonomous decision making. Even if Sandstorm used the services of a DGPS quality inducing a cm difference between its known position and the position given by the GPS, if there was no correction made on the offset between the true position and the one given by the GPS there was no way to eventually make it to the end of the race. When you look at the video you see there is no good reason as to why Sandstorm should have failed in this part of the race....

DARPA Grand Challenge Race Video

Since I could not make it to California for the race, I had a tape made of the satellite feed DARPA advertized. We put a small and larger version in order to even out the load. My bet is everybody is going to hit the 31 MB file....Enjoy.

Learning how to drive.

This article in Bayesian integration in sensorimotor learning tells us that it looks like the central nervous system's idea of learning is refining a probability. So it is all the more interesting that FASTSLAM uses a similar approach for rovers in unknown environment.

Quality is our job Number...errr... make that Number 5

When you read about what happened to the NOAA-N satellite you know something went very wrong.  

More detailed information on this accident can be found in the Earth Science Missions Anomaly Report for GOES/POES Program/POES Project.). According to NasaWatch, the people who were working on this  bird  were also working on a Quality Training Workshop for some ISO re-certification.

The sky is falling, who are you gonna call ?

The reason we got to know about 2004 FH (you know the meteorite large enough to do damage) is because of 2004 AS1....

Murphy's law for autonomous path planning

So let say you design a path planning algorithm for your autonomous robot and think that it should work ok because you have this super camera, laser ranging and think of the environment a little bit like Mars. You have seen examples of races, and feel comfortable your vehicle can recover from a bump. Then you enter the DARPA grand challenge, your vehicle goes fast but then you get a bird in your field of view, how is your algorithm taking care of THAT.

Nature ain't fair


Found this in the news in France. In Dunkirk, a wind turbine fell victim to a 70 mph gust of wind. Wasn't wind supposed to give us free energy not knock us down ?

That's about 6000 kms short

A recently discovered Near-Earth Asteroid makes record-breaking approach to Earth at 43,000 kms. That is 6,000 kms away from the Geostationary belt where GEO satellites are parked. 6,000 kms is about 5 times the size of Texas. The Barringer crater was formed with a meteorite only 30-50 meters wide, 2004 FH is 30 meters wide.

Breaking the rules II

This is a very inspiring article on how Mel Gibson broke most Hollywoods rules.

Better photos of the grand challenge...

can be found here. I especially like the it's always the little details, or the peanut that would instill fear quotes..

So to make a statistics of what the robots could not do, we have:
- vehicle cannot start again after being stopped to let another vehicle through.
- berms
- barb wires
- wrong direction of the tracks
- too cautious because too big
- bad starts (forgetting gyros, locked up brakes, concrete barriers).

As one can see, it looks like the crux of the issues to be modeled is not the steady state 50 mph straight line road, rather the little things.

We need better maps, actually less expensive ones

This ViaMichelin Espace PDA still requires one to have a PDA (400 $) and then you have to pay up to 200$ for a mapviewer. Still way too expensive.

Grand Challenge Details

Here is what really happened to each and every of the competitors of the first grand challenge race organized by DARPA. Interesting ....

7 miles is the boundary of our knowledge

According to the Status Board of the DARPA Grand Challenge, Both the Read Team and the SciAutonics II team made it as far as 7 miles into the race. Two hours before the start, the total course length was finally given: 142 miles. This is a far cry from the original 250-300 miles given to the competitors a year ago. So the state of the art of our knowledge is now known: 7 miles out 142 miles. Out of fifteen teams, only four made it as far as five miles The remaining eleven did not make it past the second mile.

Images of moving proportions

First there is this one, of driverless moving vehicle at the
Grand Challenge Qualification stage. They are moving because one witnesses the removal of the human element. Another moving picture is this one left by Charlie Duke during Apollo 16 on Descartes hills. This one is moving because one added the human element in a humanless world.

Grand Challenge Qualification Stage

So far only vehicle has made it. But it took a second trial. The first one ended up hurting, actually about 250 K$ of hurt according to Red, the head of the CMU Red team. More photos and tracks of the race can be found here.

Movie Game director II

Maybe my idea of having movie-game directors was nota bad idea after all. According to Sony, in the youngster generation, TV losing out to video games.

An Analog Fractional Fourier Transform

This article on electrically controlling the focal length of a lens makes it an ideal tool to provide an analog of the Fractional Fourier Transform. Interesting...

Linking Stellar interferometry and Harmonic Analysis

In Boone's thesis on dynamical configuration of interferometers for the study of galaxies, one can read two articles eventually published in "Astronomy and Astrophysics". For the first time, a software is written describing how pairs of telescopes should be located to obtain good images. The interesting point is that in interferometry, the positioning of the telescopes is directly related to a location in the fourier space or the u-v space as they call it. What I find fascinating is that Boone's program does not seem to take into account what he is looking at. If one does not have a preconceived idea of what one is looking for, this is fine. However, the recent use of new harmonic tools like curvelets to decompose astronomical images tells us that a large reduction of observations in the u-v plane could be achieved. The idea is as follows: each curvelet has a very specific and known shape in the u-v plane (or fourier space, see courses by Jean-luc Stark or papers by David Donoho). Astronomical images are decomposed nearly optimally in this curvelet basis, hence a small amount of evaluation in the u-v plane should enable one to recover the whole fourier transform of the astronomical image being sought.

Brainstorming

I just found a good summary of brainstorming application here. It looks like a lot of people are thinking about these things, yet I really began to be interested in these things by accident even though I badly needed them.

Interferometry and wavelets

When I read about this technique, I cannot but think that they are doing an analog of a decomposition of Haar functions in 2-D. It looks like some other peole are thinking along those lines with more sophisticated functions (fresnelets or curvelets). Food for thoughts...

We need better maps V

Maybe this is the beginning of an answer for a low cost approach to better maps: a disposable Computer . Maybe I should be looking into this ?

This ain't disruptive anything, it's called listening to your customers

Once again the word disruptive has been misused in this Technology Review article entitled: Disruptive Incrementalism. Disruptive technology is really about how to develop a technology for a niche market, make money off of it and then expand it into the very competitive markets. The article describes just the opposite, how does one go directly into the mass market and tweak some minor attributes of the product to make it widely successful (that is until people figure out the rechargeable batteries are not replaceable). Anyway, this is as if I were to claim that our in-house task manager had been disruptively adopted within our little outfit because one day, the developer decided to implement a simple library on top of it. The mere set up of that library opened the door to something not seen before: i.e. the user could now upload document to it and send an http address to people they wanted to show the document to. No more large E-mail attachments, no more setting up of a web site for the purpose of sharing information within a project, no more need for FTPs, no duplication of files in so many different computers, no need to go thru murphy's law a day before presentation (because a hard drive failed)....This was not disruptive, rather in one small increment, it answered many different unfulfilled and most importantly unindentified needs. The developer was listening to its customer. Now this application could really be disruptive in that it could be sold to very small companies. We are trying to figure out what would be the selling point...