PDA

View Full Version : [Question(s)] Linear Actuator



webgeek
11-29-2009, 03:29 PM
I'm trying to build a 4-leg robot that can haul a payload of a couple pounds and was wondering if people had used linear actuators (smaller than the ones sold by Trossen) in the past for smallish robots like this. I know people prefer servos, but spending >$100 per servo is more than I can justify if there is any other option. It seems I could design the robot around actuators as the amount of movement in each joint is minimal.

I was thinking ball screws or ACME-lead screw type systems coupled to a stepper motor as the most logical option. I have my own CNC mill so I can manufacture the actuators from scratch if need be. I was just wondering if others have done this successfully.

One possible design is something like this but dramatically smaller:
http://buggies.builtforfun.co.uk/Sim/linear-act.html

Thanks!

Mike

lnxfergy
11-29-2009, 04:06 PM
Got a link to a particular linear actuator? I've generally not seen smaller linear actuators that are powerful and cheap (until you cross a certain threshold, the servos of similar strength tend to be less expensive, especially when you add in the cost of the controller that often isn't included with the linear actuator).

-Fergs

webgeek
11-29-2009, 08:54 PM
I'm actually expecting to have to make the whole thing myself. My plan is to use a coarse (albeit small) threaded rod with a nut on it - kinda like a lead screw mechanism. The rod is fixed on one end to a stepper motor. When the stepper moves, the rod rotates and the nut will move up and down the rod. The idea is that a "sled" will be attached to the nut and the sled is what will be used to provide the actual force. Of course, I'll have to make some mechanics to handle holding the sled so it moves in a linear way and mounting the motor to the whole mechanism. Sorry, I'm explaining it poorly and I don't have a diagram yet. If you are familiar with the basic lead screw idea this should make at least a little sense.

This should behave like a worm drive in that there is far too much friction for it to run backwards. Additionally the torque of the motor is increased dramatically with the large reduction at the expense of speed. It seems like this will work OK in the end but I was wondering if anyone had done something like it before.

What got me thinking down this path was when I realized I could likely design joints that don't require a circular motion to move them AND that the force could be applied at an angle to the joint itself - basically getting away from the idea that the joint rotates around the servo. Not terribly original thinking but I think it might work for my current idea. This would mean I need holding power, but not necessarily a ton of speed and I can use mechanical leverage to trade off speed vs. torque. Thanks!

Mike

MikeG
11-29-2009, 09:18 PM
In my experience quads need a fair amount of joint speed to keep from tipping. There's generally a point in the stride where the bot is unbalanced.

You can pick up a RC servo for $12.00.

webgeek
11-29-2009, 09:20 PM
Yeah, I was concerned about that too - three feet on the ground isn't enough at some parts of the gate. This brings up my next idea - what about a large gyroscope to make up the difference temporarily?

Mike

lnxfergy
11-29-2009, 09:28 PM
Depending on your requirements, you can usually stabilize a quad by avoiding poor placement of the feet. (draw a support polygon/triangle using the legs on the ground, keep the COG inside it, in academia/research, the shortest distance from COG to an edge of the support polygon is the "stability margin", a number of stability margin approximations exist)

-Fergs

webgeek
11-29-2009, 11:17 PM
Thanks Fergs - in my head I can see what you mean though I'm having trouble visualizing how the gait will work. Guess that's something I get to play with to make work. Fun stuff :)

Mike

MikeG
11-30-2009, 07:50 AM
Another way to say this is your quad has to take small steps and employ a ripple gait.

lnxfergy
11-30-2009, 08:15 AM
Another way to say this is your quad has to take small steps and employ a ripple gait.

I don't think that's necessarily true. The ripple gait would be necessary if you can't move the legs fast enough not to fall (if you can move the legs quick enough, you may be able to move diagonal legs at the same time, or nearly the same time but slightly staggered, regardless of how fast you can move, there's gonna be a little bit of 'falling').

But as for small steps, even if your steps are slow, if you play with the geometry of where your feet are, and move the COG a bit during the walk, you can probably use most, if not all, of the available stride length. I'd actually think that longer strides are easier with a slower bot, since it's tough to swing your COG side to side on a fast bot (without having the action knock you over on your side).

-Fergs