The NITLEMITE Robotics Project

In my "spare" time, I'm working on the NITLEMITE robot - yes, the one where I'm playing with LEGO bricks. The purpose of NITLEMITE is to develop a nitinol-powered walking robot similar to the Stiquito robot of Indiana University and the SCORPIO robot of The University of Toledo, using LEGO parts for the body as in the robots from MIT's Mobile Robots Lab. The use of LEGOs allow for easier prototyping and greater expansion of walking robots.

Progress report

10 August 1994:

GATOR test circuit operational. GATOR uses a 7555 timer to generate a square wave of about 0.5 Hz; this can be varied by a potentiometer. Two monostable multivibrators (MM) are used to establish a two-phased control signal: one MM triggers on the low to high transition of the square wave, and the other MM triggers on the high to low transistion of the square wave. The output signals for each phase pass through its own set of DIP switches. By turning on a DIP switch, one connects that phase signal to the corresponding leg circuit. From this, one can set up a pattern in the DIP switches for leg testing and gait experimentation.

19 August 1994:

The prototype NITLEMITE robot took its first steps today. Walking should improve with the development of a better foot design.

8 September 1994:

The prototype robot is still "shuffling", and improvements don't look promising. However, it has served as a useful foundation. I'm now prototyping a new design, which is a streamlined version of the first prototype, but still using some of the basic premises developed. Initial "playing around" with the new leg design looks promising. The true test will be how well it works with nitinol wire attached to it, and how well it works when sets of legs are grouped together. Eventually I'll get around to bringing in my camera; once that happens I'll post some photos of both designs here and give some construction details.

15 September 1994:

The streamlined version looks very good. One leg was hooked up with nitinol - and the leg moved the robot forward. One thing to note is that if you use 1/8"-o.d. Al tubing, it'll make a tight fit into the hole inside the studs of top of a Technic brick element. By threading the nitinol inside the tubing, then placing a 3/32"-o.d. Al tubing section inside the 1/8" tubing, the nitinol will be securely fastened. Still haven't brought in the camera; hopefully I'll do that soon. :-)

21 September 1994:

Still no photo, but I do have a diagram of the Nitlemite leg. The leg is formed by a horizontal component made from a 1x6 LEGO plate and a vertical conponent made from 0.036"-dia. music wire. A hole is drilled in one of the LEGO studs and a notch is cut into the side of the plate. The music wire is bent so that there's a short horizontal component that the plate rests upon. The music wire is glued into place for stability. The key modification is the use of a "spring wire", made from 0.025" music wire, that is separate from the actual leg. The spring wire is constantly pushing against the aluminum crimp on the leg wire, providing an electrical contact. The nitinol wire can be activated via connections to the crimp on the body and to the body end of the spring wire. When the nitinol is activated, the leg pivots on the turntable and pulls back on the spring wire. Once current is removed from the nitinol wire, the leg is returned back to its original position by the spring wire. The spring wire is physically prevented from pulling the leg all of the way back by the leg arrestor LEGO stud. The crimps are made as mentioned in the note for 15 September; the leg crimp has a 3/32" component that's longer than the 1/8" component. The 3/32" portion is crimped to hold the nitinol in place. During construction, if the nitinol has some slack, it can be pulled up by twisting the leg crimp prior to the actual crimping. Please note that the body is also made from LEGO bricks - I'm just too lazy to do all that drawing at this time! :-)

17 November 1994:

The photo of the partially assembled robot, with two of the four legs wired, is now in this document. Work will continue on a sporadic basis; other things are taking up most (if not more than all!) of my time. I do have a BASIC STAMP in my possession; one of these days I'll put together an interface and see how well it works...

14 April 1995:

Virtually no progress to report, save thoughts on what can be investigated for future work. The major problem with the NITLEMITE and SCORPIO robots is the power supply. With the nitinol wires eating 180 mA each time you activate them, it won't take long to drain a battery. One solution is using solar cells. Edmund Scientific (and probably a few other places) sells hi-output solar cells. In full sun conditions they will put out about 500 mA / 0.55 V. Unfortunately, these cells cost about $15 each. What could be done is to dedicate solar cells to the nitinol wires when walking, then when the battery for the circuitry starts to go, the robot can "sleep" while using the solar cells to re-charge the battery. With a good design one could limit the number of solar cells needed.

I find the use of solar cells to be interesting as one could deal with more behavioral mechanisms with such a robot. What should the robot do when a room is dark? If there is a reliable period of light followed by dark, what would be the best way to organize one's "day"? Could a robot detect the place where the light is the most intense and refer to it as "home" for purposes of "sleep" (to recharge it's system)? What information could a small robot gather about its environment, and how could it be communicated to others?

A related design item would be to develop nitinol powered "tongs" so that the robots could interact with the environment instead of just maneuvering through it.

I've mentioned the Basic STAMP earlier - it's small size and ready accessibility would make for a good on-board controller for the masses - all that's needed is the interface for the nitinol drivers. The STAMP is made from a PIC microcontroller; a design based on a high-end member of the PIC family would make for a better controller for those who like assembler.

Developing controllers based on low power designs would help to expand the amount of time that the robot could operate on a single battery charge. Some of the 3V versions of the 6811 would do nicely here, and versions of the PIC can operate at 2.5V and go into sleep mode. Limiting the peak current demands would also be of interest. This would require research into seeing if a single 180 mA current source could be multiplexed to the legs requiring movement: can a pulse train work in place of a continuous current supply?

Return to Nitinol Robotics Home Page

John K. Estell - 14 April 1995