There appears to be a low cost 3D printable turtle here which may be a starting point:
Something I recall about the early turtles was that their wheels tended to slip a lot resulting in increasingly skewed drawings. This was often made worse by the drag caused by their umbilical cords used for power and data.
With the likes of Wi-Fi and Bluetooth and the advances in battery technology then umbilical cords should no longer necessarily be an issue.
However is there any low cost technology which could perhaps be used for better indoor positional accuracy these days ? Could a turtle ever be made accurate enough to autonomously mark out a floor for tiling or to mark out a sports hall for line marking.
How about an outdoor turtle which could draw in the playground/car park using some sort of chalk or water soluble paint or alternatively to scribe sand on a beach ?
I guess that turtles have taken a bit of a back seat to the modern 'explorer' class of educational robot.
Not sure what the old design used, but would small stepper motors fix the positional issues? Not necessarily a cheap option, of course. Digital compass sensors would at least fix directional issues.
Love the idea of doing this on a beach.
I think the problem was more to do with the wheels slipping on the paper when doing turn after turn on more complex designs, so a stepper motor (or encoder stepper motor) wouldn't necessarily help here. The wheels were probably general purpose so perhaps not optimised for any one given surface.
It was a long time ago but I recall lines not joining up at corners and banana shaped drawings. Back then people were amazed if the program ran to completion without crashing and that the line was continuous. People today however perhaps expect a lot more from technology.
In the old days, small stepper motors (especially without gearboxes) could often miss steps, so did not guarantee rotational accuracy. Hence why some stepper motors had encoders attached to them to measure actual rotation for increased accuracy. However with the likes of a wheel, this does not guarantee travel distance. When you have two wheels then it doesn't guarantee tracking accuracy.
Note that the Honeywell HMC5883L 3-axis digital compass as used in the likes of the Adafruit module is only accurate to 1 - 2 degrees and this would only help with angular measurement and not distance measurement.
2 degrees error on a single turn over a distance of a meter, is something in the region of a 35mm error.
The University of Reading cybernetics department came up with a 3D printed turtle-like robot called ERIC
which features in its robotics MOOC on Futurelearn.
It has a 3 axis accelerometer built-in but I don't think it has a magnetometer.
It is a self-balancing two wheel robot so perhaps not the best suited to line drawing as it tends to rock about as it starts/stops. However it is an example of a low cost design. There was some talk of it becoming commercially available to accompany the course but I'm not sure if this ever happened or not. For the initial course they provided an on-screen ERIC simulator instead.
This Harvard Root turtle-like project may be of interest.
Being able to traverse a whiteboard adds a new dimension.
More info in the Industrial Automation and Robotics blog