MAS.714 – Systems and Self, Sherry Turkle and Mitchel Resnik
Childhood Object Paper, due Sept 29 1998

From "Backing up with a tractor and two trailers" to "Controlling highly unstable systems with nonlinear dynamics"

Or: Making a double integration process concrete

Stefan Marti
The Media Laboratory
Massachusetts Institute of Technology
20 Ames Street, Cambridge, MA USA 02139
stefanm@media.mit.edu

 Backing up a tractor is a relatively simple thing. However, if due to a minor deviation, it starts to turn to one or the other side, we all would probably know how to compensate for this: steering to the same side. E.g., if the back part of the vehicle turns to the left, turn the steering wheel to the left side as well. This is intuitively simple to understand. Nevertheless, we all have to learn it to drive a normal car. But if we are maneuvering a vehicle with a trailer, everything is a bit more complicated. When during backing up the trailer escapes to the left side, we most probably do not know intuitively how to compensate for that. And if you think of two trailers, backing up in a controlled way gets very complicated, if not impossible. However, when I was 6, I loved to do this! I tried to back up with a lot of different vehicles, and I got pretty good at it. Playing with them, I think I have developed the ability to understand intuitively such unstable systems and the problems behind it. This is also the reason why I am still fascinated by such processes. For example, right now, I am thinking very hard about an automatic control problem, which is very similar: automatic hovering of a micro helicopter. Helicopters are very unstable, and they have even more degrees of freedom than a tractor with multiple carts. In fact, studying this problem on the background of my early childhood experience made me develop an unusual solution for the automatic hovering problem. The standard approach for such a control problem would include inertial sensors like accelerators and gyroscopes. To correct a position of a helicopter automatically, the sensor, which detects only acceleration, integrates this measurement to velocity, and once again to distance. Since I know how hard this is because errors in a double integration process sum up extremely fast, I have proposed to solve the automatic hovering problem with an absolute position sensor, which doesn't have to integrate the data. I think, the fact that I tried to back up with tractor and trailer and got fascinated by it had an influence on my preferences for looking at highly unstable systems, and made me sensitive for solutions to an automation of this process. Automation of hovering is part of the AI agenda, AI in the narrow sense: extending the capacity of machines to perform functions that would be considered intelligent if performed by people. I don't know if it needs "intelligence," but backing up with multiple trailers is almost as difficult as controlling a helicopter. Just a week ago, a helicopter pilot told me that only a few pilots are able to integrate twice in real time at all! No wonder I had a hard time backing up straight with my toy vehicles... Note 1: My original toy tractor was not out of LEGO bricks, but a wooden toy. My parents never allowed me to use LEGO, because it is made out of plastic, an artificial material! LEGO never was an option, because my parents, and especially their parents, come from an anthroposophical tradition, founded by Rudolf Steiner. Kids are supposed to play with warm and natural materials like wood, and plastic was very bad. There was another reason not to use LEGO: it was all rectangular! The anthroposophical tradition, especially in architecture, prefers angles that are different from 90 degrees. Wooden blocks having leftovers of bark on one side were state of the art, since they broke the regularity of the brick. My parents realized that construction kits are extremely important to me, so they made a compromise and bought me BILOFIX. These are wooden bars, screws, nuts, plates, wheels, and eventually, electric motors. The possibilities are very much the same as with LEGO today. And therefore, my maneuvering experiments were vehicles built out of BILOFIX. Note 2: I have criticized the mechanocentric orientation of Papert, especially in the Introduction and Foreword of Mindstorms. Although I realize that his gears are actually very close to my backing-up experience, I still think this is not the most important thing. It is "only" about intellectual development, but it ignores moral and social development. As an example, I would like to mention another object. Actually, it is the absence of an object. My parents never allowed me to have a toy gun. But all my friends had toy guns! The reason for this decision must have been a pacific thinking, I guess. This "not-having-a-gun" experience influenced my moral thinking definitively much more than any other event in my life. Not being armed in a world of arms made me think of alternatives to armed conflicts, made me think about non-violent conflict solutions, and had a very distinct influence on my attitude towards armies and social responsibilities. Since all Swiss citizens are per definitionem part of the Swiss army, I almost became a criminal by following my primary conviction that arms are not a solution to any conflict. Note 3: I had a pretty hard time finding an appropriate childhood toy experience. I had to call my parents in Switzerland, and we went through my first years to find an object. We were talking for an hour before I remembered my tractor-and-trailer game. Therefore, I conclude, it must not have been obvious to me. Or was I just not aware of it? Do we over-emphasize such events? I don't know.

Send me some comments! Last updated Sept 29 1998.