REPORT : An Inexpensive Turtle
BUILDING AN INEXPENSIVE TURTLE
by
Michael Folk
Mathematics Department
Drake University
Des Moines, IA 50311
Two years ago two colleagues and I set about developing a microcomputer controlled programmable robot for teaching1 . The goals of our project were (1) to build a lowcost (less than $75 in parts) device ( "turtle" ) which could be programmed by children to perform " intelligent actions " as a robot does, (2) to build a low-cost ($400 or less) and portable " brain" for the robot so that it could easily be used in schools , and (3) to develop primitive software which would enable a user to program the robot (more sophisticated software was to follow) .
The turtle/robot system was first proposed in 1950 by Grey Walter, who built a machine which looked like a mechanical tortoise and had electronics built into it which enabled it to "behave" remarkably like a simple animal . In 1969, many generations of tortoises after Walter's first effort, Seymour Papert of the MIT Artificial Intelligence Lab and Wallace Fuerzeig of Bolt, Beranek and Newman jointly headed a project to build a " turtle " much like Walter's tortoise, but with the additional quality that it was controlled by a computer program . The MIT turtle was to be used, together with a simple but powerful language called LOGO, to teach computer science to elementary school children .
I was fortunate to spend almost a year, off and on, using the MIT turtle to teach mathematics and computer science to elementary school children . It proved to be an effective device for motivating children to study mathematics and general problem – solving, but I felt it had three major drawbacks as an educational device : (1) It was expensive to build (about $1200 at the time) . (2) It drew its " intelligence" from a large computer, requiring either that it be located near the computer or that it be connected to the computer by phone . (3) It posed enormous problems in teacher training, as few elementary teachers could be expected to eagerly accept working with the system . All three drawbacks would contribute to inhibit the widespread use (or even trial) of the system in realistic school environments .
In 1975 we undertook the task of trying to solve the first two problems . The results have been promising, though by no means conclusive .
The turtle was built for less that $50 in parts–two-thirds of the target cost , and perhaps a twentieth of the cost of the model . Of course, the $50 cost does not include the costs in human time used to design, find materials, and build the turtle . (For instance, the gears alone, which cost $1 .47 each, were found only after examining scores of possible candidates from industrial gears to those used in children 's toys . A dozen man-days were probably expended just in finding suitable gears . ) Nevertheless, we believe we have shown that a turtle can be built which could be produced cheaply enough for most elementary schools to afford a few of them .
The second problem, that of freeing the system of the need for connection to a large computer, has not been completely solved, but major progress has been made . We were able to put together a microcomputer-based system small enough to fit into an attache case . We actually built not one, but three different systems before we felt we had a truely workable one . The first system cost about $400 in parts . The second and third "generations " were built with parts from the initial system plus generous donations from Motorola, Inc ., the company whose M6800 microprocessor we were using .
The final version is not only smaller than the original one, but more reliable, with more memory and capable of being built for about half the cost .
It is not enough to evaluate the project merely on the basis of cost . Just as crucial is the question of how well the system does what it is supposed to do . The system we have built is about as accurate as the one it was designed to replace . Our primary measure of accuracy is how well the turtle can draw figures on a paper on the floor . Our turtle drawings generally show an error of about 2%, as do those of the MIT turtle I used previously .
A much more important shortcoming of our system is that it does not support the sophisticated software of the MIT system . We did not intend to put the full blown software on our micro system–we were going to put it on our large CDC 6400 system , and we were going to put a skeletal " turtle-language " on the micro system . We have put the turtle language on the micro-system, but we decided about mid-way in the project that, with additional memory, we might also be able to put the sophisticated software on the micro-system . Our next major project will be to try to do this . Without it, we do not feel the system will be adequate for use by children .
1Partial funding for this research was provided by the Drake University Research Council .