Mechanics and Construction > Mechanics and Construction

Simple tripod walker

(1/2) > >>

Gopher:
I've decided to try my hand at robotics again for the first time in years; I've decided to try to build a simple 6-legged walker, doing a basic tripod gait. I've come up with what I think is a pretty slick and simple way to produce the desired leg motion, and getting it to move forwards/backwards shouldn't be too big an obstacle (as long as I'm not too picky about how straight or fast it walks, anyway).

My rough design so far includes three identical segments, each with it's own motor controlling two legs. I'm building first versions of those segments now, using motors scavenged from some of my old zipzaps, and my first version will just connect those three segments and provide common, constant power and control segments - so it'll just walk, constantly, while powered, testing the basic walking mechanics out.

Once that's finished, the next step is adding the ability to turn. I've got an idea about this, but I'm not certain it will work, which (finally) is what I'm here asking about. My idea is to have joints between the segments, and just "bend" it left or right, either with weak electromagnets or possibly using muscle wire. I'm fairly sure this will work, but I haven't actually tried it, and I don't think I've ever seen one done this way, either. Anyone sure about the viability of this approach, or have an alternative suggestion?

I appreciate any help; advice, links, anything is fine. My ultimate goal, many versions down the road, is to make a robo-roach; as small as possible, with very simple hard-coded logic. I don't have much ME experience, but I have done some EE including designing simple control circuits, and I don't anticipate much trouble designing the brain (though implementation is another matter)

For the simplest turning method, you should use differential drive. This states that if the left side walks forward, and the right side walks backwards, the body of the robot will rotate clockwise about the center axis. This wont be the most efficient method because of feet dragging, but it will definitely work with zero change to your design.

Dont use muscle wire - highly inefficient, slow, weak, etc. Muscle wire works best for micro-sized applications only.

You should upload pics of your design, as I am very interested - I have been toying with a design for a few years now to imitate the roach, too.

Gopher:
I'm concidering muscle wire precicely because I want to make it as tiny. The prototype is going to be about four inches long and an inch and two inches wide, but in later versions I intend to make them as small as I possibly can. The walking movement you describe sounds workable...  It will require a little modification of my design, because the prototype uses one motor to control each left/right pair of legs (one for front, middle, and back), but I don't think the foot-dragging will be too big a problem. It would remove the need to have the body twisting, though, so in theory I could devise a way of sharing motors between legs on different segments, but this is not a trivial goal...

I don't have any easy way to digitize my designs, which live on paper and on my whiteboard at the moment; I'll get a friend to take some digi pics when I get a chance, so I can post them.  The mechanics I have come up with for the legs is very simple, though. The legs will be made of steel wire. It will run through a hole towards the edge in a wheel, going out and then bent to point down. It will be secured to the wheel by washers on either side, rather than being secured to the wheel (which would force the leg to actually spin). So the leg cannot wobble in the hole, I'll use a short metal sleeve attached to the hole in the wheel, or possibly just put a couple of wheels together. A strut will point up from the leg just before it starts it's downward bend, and be loosely run through a fixed hole above the leg's center position. This should create a nice circular movement at the feet, without requiring me to fabricate anything but the simplest of parts. I'm pretty sure this will work, though even if it does I might go with some completely different approach in later versions.

I've never used muscle wire, but I want to get some to experiment with; I can imagine using four strands to create a similar circular movement in the feet by simply cycling the current around them in a circle. This would remove the bulky motors, making more space for the brain. I expect ultimately the control circuitry is going to dictate how small it can be more than the mechanics, though, motors or not.  I'll worry about that more later, though, for now I have to get the Roachbot mk1 put together, which will just be hard-wired to walk foward constantly.

The ultimate goal, which may or not be acheivable in the intended manner, is to make an autonomous roach. It would have a solar panel, not for direct power but to recharge it's battery (it's OK if it takes hours of sitting to draw enough juice to scurry around a room for a minute) It would scurry around, using a pair of photoresistors to find a bright spot to sit, and going into an idle state when it finds a spot sunny enough to charge. It would also contain a small mic, and would respond to noise by going into a hide state, reversing it's logic and searching for a dark spot instead of light. It would also have antenna, it would use to detect and follow walls and edges. Ultimately I picture it being a novelty kind of "executive desk toy" - a robot that lives on the desk, occassionally coming out to charge under the desk lamp and then scuttling under the monitor when someone approaches.  I know all of these things are individually acheivable, but I don't know if it's within my own ability (or budget!) to put all these things together into a single microrobot without making it much larger than I like. But I'm starting ultra-simple, and I'll add sensors and logic one at a time in a progressive series of new versions, and that only after I settle on a working design for the chassis. I'm also prepared, albeit reluctantly, to give up on having it be a walker if I can't get it to work out - I am aware of the amount of complexity this alone introduces - but I have to take a few stabs at it first.

dunk:
hi Gopher,
http://www.xs4all.nl/~sbolt/e-spider.html
sounds like it might provide you with some ideas.

dunk.

Gopher:
Wow, that's pretty slick; I can easily see that design scaling down pretty well, too. One thing, I intend my roach to be more of a spastic walker than a deliberate one, trading control for speed. An advantage of the leg movement my mechanism will produce (if it works, that is) is that the legs will actually lean forwards before contacting the ground and push off backwards before rising again. I hope that this will increase the forward thrust without increasing the amount of vertical force; this is important, because the faster the legs go, the more tendency there will be to "bounce" and the less controllable it is. I'm thinking about abandoning any pretense of a tripod gait, because I doubt my ability to maintain synchronization without a proper microcontroller. Instead, the legs on each side will share a motor and be synced mechanically to push off in series. This won't be graceful, but it should function just as well at high foot rpms.

I don't expect any version of the roachbot to deal with slopes or obstacles, I just want to come as close as possible to scurrying. If my leg designs don't work out, I might try something more like the spider design next. My only absolutely inflexible goal for this robot is for it to scurry like a cockroach; if I can't find a small, low-power leg system that produced enough speed (I really want at least 10 body lengths per second!!) then I'll abandon legs in favor of wheels, but I was inspired originally by some videos I saw of bots built at Stanford.

Thanks for the link! Definitely gave me some ideas