Centipede

UserPreferences

The Centipede is a segmented insect-like robot. The segments are constructed such that they are removable and "stackable". That is, the robot can be made almost any length you like.

I have been encountering some problems in the design definition stage. The big problem is that of knowing the turn angle. I would like to use a potentiometer for it but all the ones I have looked at have a high friction on the wiper. The robot is light weight so this is unacceptable. If you know of a small, low-friction potentiometer, please drop me a line.

UPDATE (23MAR04): Some people that I have contact with through work are getting me the name of a manufacturer that makes 360 deg. pots. That is, the pot can turn infinite angles and has very low stiction. I'll post contact info for the manufacturer as soon as I find out. It is exactly (almost... it is a bit large) what I have been looking for!

Failing that, I decided that a flex sensor (http://www.spectrasymbol.com/bend.html) would be a good choice. They only sense flexing in one direction so I would need two of them back to back. I figure they would have to be about one inch long. That means custom.

UPDATE: (23Sep2004) I have found a part that will do the trick! It is used on the nose-vane of the Predator UAV for attitude sensing. The part is made by the Dynamation Transducers Corporation.

I will use the pots for a prototype so can I can get the motion control algorithms and circuitry worked out. The proto can also be large so the high stiction value of the potentiometers is less of an issue. However, the final goal is to have segments running off pager motors (if I can find a pot small enough).

I plan on using the Motorola Nitron series of HC08 microcontrollers. I am eyeing the 68HC908QT2 or the 68HC908QT4. The difference between the parts is the size of the code space: the QT2 has about 1.5KB and the QT4 has about 4KB. The devices are based on Flash technology, have a four channel eight-bit ADC, six I/O pins (some are muxed with the ADC), are eight pin packages, and have an internal clock.

The pins are assigned thusly (assuming using flex sensors):

Left motor control 
Right motor control 
Left turn angle sensor 
Right turn angle sensor 
GPI for "all stop" 
Vss 
Vdd 
spare

Assuming I can find a good potentiometer:

Left motor control 
Right motor control 
Reverse motor control 
Turn angle sensor 
GPI for "all stop" 
GPI for "reverse" 
Vss 
Vdd

I am considering making the brain of the Centipede using a Cypress PSoC. I became aquainted with this part when I was at Dell and it is capable of some very complex processing. It has analog and digital blocks. As a general-purpose processor, it does not work so well, but for embedded applications, it is terrific!

Nuts and volts! I got screwed!

What do you know about that?! In the October 2003 issue of Nuts and Volts there is a centepede robot on the front cover!

No kudos to me in the article, though. Parallel-divergent evolution? It could happen. Just remember where you heard it from first! ;-) My design is cooler anyway (mainly because it's mine) since it is autonomous and Mike Keesling's RoboPede is remote controlled. Mine is stackable; his is a "head controls body" type architecture. Sadly, little progress has been made on mine as my having been unemployed prevented me having the funding needed for these things. When it comes down to motors and microchips versus babyfood, the babyfood wins every time.

But since I have a job again, I can start craking on it.

main page: MarkHinkle