Guide on making centipede robot

[Sodrohu]

Hi there.

I'm a second year student taking EE engineering. Of late, I've been thinking of making a crawling robot resembling a centipede.

Currently the plan is all in my head. For a start, the first segment of the robot (the head) would have six legs. I plan to construct the head segment first, and test it around, to see whether it's moving according to how I want it to be. After that, if it's good, then i'll start with another segment, which will be connected to the head. Again, I'll test it to see if it moves well, and if it does, I'll add another segment. This continues until I have, say, 6-10 segments, depending on budget.

SO my questions would be:
1. I dunno what kind of sensors I should put in my robot. Currently, I'm just thinking of making it move just like its real-life counterpart, but so far I'm not sure on what kind or sensor it should have.

2. The kind of motor drivers I would be working on. I'm planning on using plain DC motors for this, but I'm not sure on the driver part. I've seen the ones in ikalogic's and David Cook's website, and I'm aiming for an efficient motor driver so maybe I'll using MOSFETS.

Oh, and my budget, being a student, won't be much. Maybe around RM500 or something (That's around 150 dollars for you.)

Anyway later I'll post the planned drawings of the robot.

 

[ericgibbs]

hi,

Google for Cricket the Robot

That site has a spider type robot

**broken link removed**

 

[dknguyen]

Is $150 enough? It always seems to me that the only robots in that range are beam robots or small wheeled robots. Anything else seems to cost a lot

GOod thing is that you can always attach more segments as you get money though. WHat's the scale of this thing?

I play with a lot of centipedes and there's only 3 sensors that are obvious to me (they obviously have way more but to an regular observer). One is contact sensors in the antenna that wave around to see where it's going. Another is pressure when you push down on its back which makes it play dead or go running depending on how often or hard you push down. THe third is when it is upside down, the "ovally round" type of centipdes (rather than the lony snaky spikey looking kind) arch their back so they rock back and forth while flailing their legs which usually gets them upright again.

 

[Sodrohu]

I think $150 will definitely won't be enough. I'm a first-timer, and chances are that I'll ruin a component or two or even an entire circuit, or buy the wrong stuff. That being said, $150 is just an estimate, because I think I can save on tools since I got access to them(I'm part of my uni's robocon team anyway).

I'm actually thinking of making my robot remote-controlled. Yeah, contact sensors sounds real great too. I probably can make it automatically stop when its sensors detect an obstacle.

Anyway, I'm not really sure of the exact scale of this thing, but it's something like this:

 

[dknguyen]

Oh, that's pretty huge. Might be cheaper to make a snake then you don't motors for each leg, just each segment.

THe thing about these kind of projects is if you can get enough money to start off you can get a bunch of identical parts machined which makes it cheap and can get some more elaborate mechanics or just framework going on.

 

[Sodrohu]

Yeah, that's the thing that is troubling me most. So far, I couldn't think of any configurations with that kind of setup that uses one motor for more than just one leg. ANyway, I'll start with making just one segment first, then if the budget is there, I'll make some more.

Just wondering, would it be better if I use servomotors instead of normal DC brushed motors? DC motors would need motor drivers, and maybe an encoder to make sure both motors run at the same speed. Servos might be easier to configure, but I heard they don't last long as rotational motors. Which should I do?

 

[dknguyen]

Yeah, that's the thing that is troubling me most. So far, I couldn't think of any configurations with that kind of setup that uses one motor for more than just one leg. ANyway, I'll start with making just one segment first, then if the budget is there, I'll make some more.

Just wondering, would it be better if I use servomotors instead of normal DC brushed motors? DC motors would need motor drivers, and maybe an encoder to make sure both motors run at the same speed. Servos might be easier to configure, but I heard they don't last long as rotational motors. Which should I do?

Is this thing using legs or wheels? Go RC servomotors. Two RC servos getting the same signal won't necessarily run at the same speeds, they just apply the same duty cycle to the motors (if the potentiometers are identical which they never are). So in that case you'd need to modify some servo motors with a kit or build from scratch with a regular DC motor.

 

[Sodrohu]

Currently I'm planning to have thing thing move on legs, since I want it to mimic real centipedes as much as possible. The legs need not be complicated, just connected to wheels and move in a circular fashion.

I'm tempted to use RC servomotors, but I actually want to use DC motors as I also want to learn how to build drivers and stuff. Does any of you have any schematics for a good motor driver and all the necessary circuits? I've seen the one at Ikalogic's website, but I think that's too way complicated, though I'm tempted to use it..

 

[Sodrohu]

I've been looking around for some stuff, but finally I found this video.

https://www.youtube.com/watch?v=-QcLCv4zHwo

Yep, basically this is what I intend to build.

Now, what I'm stuck with right now is thinking how to replicate the leg movements. If say, the robot is controlled through RC, it will control only the first front legs of the segment, and changing the speeds of either legs(turning the segment left or right) will make the rest of the segments follow suit too, by means of a electronic/mechanical mechanism.

If I'm using two motors for each segment(excluding the first one), then I've to deal with them not turning at the same speed while going straight. Right now I've been thinking of using one motor to drive both legs for each segment, but I'm stuck at figuring out how make the left/right leg slow down while the robot turns left/right.

FYI, the robot can only go forward.

 

[Cotowar]

SO my questions would be:
1. I dunno what kind of sensors I should put in my robot. Currently, I'm just thinking of making it move just like its real-life counterpart, but so far I'm not sure on what kind or sensor it should have.

Well, to be quite honest, your sensor really doesn't affect the motion of the bot. Your sensor tells the "brain" if it's okay to move, not how to move. I would look at the size of the robot vs how dexterous you want it to be. if you want this thing to be the size of your forearm when you're finished, the sensors will be quite different from if this thing looks like a bunch of bicycles. If you are making it big, i would suggest a more powerful, general sensor that would tell it if there were a large object, such as a person, in front of it. You can then also include a small sensor pointing more downward than forward to determine if there are smaller objects in the way.

Just my opinion. Good luck with it.

PS, pm me please once you get the walking part finished. I was planning on making a scorpion robot, and since they move in a similar fashion and I've never done this before, I'd like to know how much Excedrin Migraine to buy...lol

 

[Cotowar]

Yeah, that's the thing that is troubling me most. So far, I couldn't think of any configurations with that kind of setup that uses one motor for more than just one leg. ANyway, I'll start with making just one segment first, then if the budget is there, I'll make some more.

Just wondering, would it be better if I use servomotors instead of normal DC brushed motors? DC motors would need motor drivers, and maybe an encoder to make sure both motors run at the same speed. Servos might be easier to configure, but I heard they don't last long as rotational motors. Which should I do?

Sorry for the double post, but could you possibly do some sort of cam timing system on every other leg? I know it would be ridiculously hard, but it would be pretty frickin sweet to run the whole thing off of only two motors. What I'm talking about is something similar to the piston action in a car engine. The only thing is, if you offset the cam gear a little more, you can get a rolling effect, similar to the way a centipede walks. I hope you understand what I'm talking about.

The upside is the electrical work would be significantly easier. Simply attach a high torque motor to the end of the cam shaft and plug it in. The down side is the mechanical work would be a ***** and a half to calculate out. You would be best served building one setup that moved the one side all at once, and the other all at once, and shifting and testing from there until you got your rolling effect. There would also be no singular control over legs, so you couldn't make the thing step over items (unless you make an even more complex joint system that can support the robot's weight) but it would walk in a straight line easily enough.

You could use a U-joint for the connection between the cam shafts, that way you still have a rigid body about the rotational axis, but can move up and down as well as left and right.

 

[Sodrohu]

Yes! Someone actually replied!

PS, pm me please once you get the walking part finished. I was planning on making a scorpion robot, and since they move in a similar fashion and I've never done this before, I'd like to know how much Excedrin Migraine to buy...lol

Well, I'm gonna have my finals in three weeks' time, and I'm busy with soem other stuff too, so it's gonna be a LONG time before I ever PM you. I plan on making this during the semester break, so safe to say it's gonna be aound a month from now.

Sorry for the double post, but could you possibly do some sort of cam timing system on every other leg? I know it would be ridiculously hard, but it would be pretty frickin sweet to run the whole thing off of only two motors. What I'm talking about is something similar to the piston action in a car engine. The only thing is, if you offset the cam gear a little more, you can get a rolling effect, similar to the way a centipede walks. I hope you understand what I'm talking about.

The upside is the electrical work would be significantly easier. Simply attach a high torque motor to the end of the cam shaft and plug it in. The down side is the mechanical work would be a ***** and a half to calculate out. You would be best served building one setup that moved the one side all at once, and the other all at once, and shifting and testing from there until you got your rolling effect. There would also be no singular control over legs, so you couldn't make the thing step over items (unless you make an even more complex joint system that can support the robot's weight) but it would walk in a straight line easily enough.

Uh...tell you the truth, I do get the idea about using cams to operate the legs, but I don't think I would be able to figure out the workings of such configuration, at leasst for now. Maybe I can do some research more on it, but I prefer to balance evenly the electronics and mechanics of this robot, though this may change in the direction of whichever is easier to do.

My untested plan on the locomotion goes something like ththe picture attached.

I dunno whether my mechanism would work the way it should, but for now that's all I can think of.THe difficult part(as stated in the pic) is to figure out how the metal rod would be able to slow down one leg (since the plan is one set of legs for a segment is driven by a single motor), making the robot turn left or right.

You could use a U-joint for the connection between the cam shafts, that way you still have a rigid body about the rotational axis, but can move up and down as well as left and right

.

can you post some reference sites where I can understand this more?

 

[Cotowar]

can you post some reference sites where I can understand this more?

here is a picture of a simple U-joint. In reality, they are much more complex looking, and are generally used in cars along the drive shaft, and with the front suspension.

**broken link removed**

basically the way it works is you hook the blue crank to one segment, and the red crank to the other. you are then capable of both horizontal and vertical movement, while still allowing for rotational drive. They are really simple to build, but the idea is pretty abstract. You could honestly build it from dowel rods and wire pins if you have a dremmel.

as for the cam shafts, I've been mulling over this design for you and I think you are right in wanting separate motors for each segment. If you didn't, you wouldn't be able to turn, because every other leg would speed up and you wouldn't get the lopsided effect needed to make a turn.

I'd do two motors and 4 legs per segment, with the legs on the right side being 180 degrees out of phase, and the legs on the left side being 180 degrees out of phase, but have the left be not exactly 180 degrees out of phase from the right. that way you will still get the rolling effect of centipede motion, but can speed up one side to make a turn. I'll draw some of it out later today and post the pics for you. Right now I'm in a class learning how to click a mouse and use the internet in Windows Vista...LOL. Dumb community colleges and their required classes...

 

[Cotowar]

heres a side and top view of how to physically make the legs move. Use worm gears, that way you can lock the legs when you aren't walking, and it will let you stop on hills without sliding down.

That's how to get the legs to move, by using a U-joint like the one I posted above to connect the segments together, and using your rod idea, I think you can make the centipede turn. The only thing I'm not sure about is the step length. The step length will be symmetrical with my design, so if you go to turn your outside leg will need to step farther in order to keep it in line, maybe just build it and try from there. The solution is really simple, but I haven't dealt with anything like this in about 8 years, so I would have to actually see it move before I could tell you how to fix it. Hope that helps.

**EDIT**
I'll have to upload the images to imageshack before I can post them here, because the uploader isn't working, the problem is that my laptop battery is about to die and I have to run a few updates, so I'll do it when I get home and can get on my desktop.

 

[Sodrohu]

I've been thinking about this one, but as of now I still couldn't think of a way to change the speed of one leg when using a single motor to power two legs(side by side). Actually, it's not exactly the speed of each leg, just the distance that one leg covers wthat I need to change. Attached are the proposed mechanism that I've been talking about.

Maybe, maybe, if somehow I couldn't figure out how to do a mechanism to vary the walking distance of each leg, then maybe I'll have to just go with independent control for each leg. I'd rather not do this though; I want to keep the phase of both legs(on each segment) the same, so there'll be no missed footing and the walking gait will be constant throughout(just like in real centipede...at least I think so)

Anyone got any ideas on how to do this?

 

[Pommie]

By moving the position of the top pivot up and down you will change the distance the leg moves. This can be achieved by putting a slot in the top of the leg.

Mike.

 

[Sodrohu]

By moving the position of the top pivot up and down you will change the distance the leg moves. This can be achieved by putting a slot in the top of the leg.

Mike.

Uh...It's kinda hard to imagine that. Can you post up a pic showing how it's done?

 

[Pommie]

Here you go.

Mike.

 

[Sodrohu]

Wow, that's depressingly simple. To think that I once sacrificed a night's sleep just to think it over...