small motors or linear actuators

"Mike"

There were a couple of other posts on linear actuators this week, take a look because there are some very nice replies.

Recently I've bought a couple of linear actuators for my project from Firgelli Auto, but these are too much for your application (30 pounds of force). They have an affiliated company (firgelly) that is just starting to commercialize micro linear actuators, and they claim that one of the applications is for robotic legs (small robots). Search google for firgelly micro (or mini) linear actuator.

Cheers

Padu

Hi - I saw those discussions, but unless I missed something, they were all about much larger devices. I'm looking for small! Each leg segment on my robot is only 6cm long - so the smaller the better!!

I just took a look at the Firgelli ones - and they don't seem like a very good fit for me. Specifically, their force is a little low (18N max) and their stroke is incredibly small (2cm). Maybe if they had a larger stroke I could compensate for their low force with leverage - but geeze - 2cm isn't alot to work with.

Squiggle motors seem possible as well

- but they suffer from having even less torque than the Firgelli ones. I also have no idea what the pricing on those buggers is, and calls and voice mails to them have gone completely unanswered.

-Mike

How about something like this?:

Michael

More expensive servos provides better bearings.

• * *

In case when total weight of all servos is significant part of total robot weight real criterion the main criterion is a factor

max torque / servo weight

rather than simple max torque because decreasing servo weight is equivalent of some torque increasing. Such factor for Futaba servos vary from 23 in to 100 in, and value for your Hitec HS-81MG is

42 in*oz / 0.67 oz ~= 54

So you can some increase quality by using for example Futaba S3102 (torque 64 in*oz vs current 42 in*oz, weight 0.70 oz vs 0.67).

• * *

Linear actuator from geared DC motor + screw provides much more better characteristics than servos in case when speed is not an issue, but this variant requires much more complicated mechanics (for example, bearings for screw, coupling screw with gear output shaft ad so on); you need also add encoder (as well as process encoder output).

MicroMo

provides for example such kind linear actuator (servo + gearhead + screw+ nut):

weight: 0.15 oz max force: 150 oz max torque (for joint rotation range 60 degree): 70 in*oz

You also must add a lot of miniature mechanical parts to embed this toy into your robot (as well as encoder).

MictoMo provides top quality/reliability/long life, but it is very expensive way.

Nick

Hi Mike, it's not clear why you would want to use "linear actuators" for a walker. Most people just use servos. Any screw arrangement will be quite slow. Do you have a picture of your bot?

I use Hitec HS322-HD servos on my walkers now, and although they don't have ball-bearings, they do have hardened plastic gears and they are

*very* cheap, less then half the HS81 price. The crust crawler, which is very heavy uses the same HS322's, although it's difficult to imagine they have enough torque to move a 4# walker, plus 7.5# load, as advertised.

I have a very different design walker, and the legs use of form of pseudo-linear actuation ... it actually will carry a 3# load.

- dan michaels ===============

Hi Dan,

Your crawler is very nice, and HexCrawler from crustcrawler.com is nice, but both uses 2 degree of freedom (and two servo) per leg. This implied that such creature can't seat down/rise/tilt, can't climb stairs... Actually such kind of crawlers don't have significant advantages over tracked bots with 2 motors only. Mike' robot with 3 degree of freedom per leg can seat down/rise, can climb stairs - of course, if software is smart enough and robot equipped by powerful actuators. Legs of such robot can be used also as manipulators or as probes for environment exploration (must be equipped by tactile sensors, of course).

Nick

Hi Dan - Like I said in my OP - I am only interested in finding a better solution. Linear actuator or motor - I don't care.

The servos you're using are much, much too large. Note that the HS-81MG is a MICRO servo. Price is also not a major concern here - I have recently gotten funding for my work. I want to move away from servos though - there is just not enough control when using a servo.

-Mike

Hi Nick - The Futaba looks like a decent servo. However - I'm looking to get away from servos. I am just so completely sick of working around the controllers built into them. I suppose I could just rip out the electronics and just use the motor + gears - but that is more of a hack than I would like to do. I would really prefer to have full control over the control loops governing movement.

The MicroMo parts look like they have potential. I will look into them.

Thanks!

-Mike

I missed any link to Mike's robot, so I don't know what he's doing specifically.

Also, I'm not sure that going from 2DOF to 3DOF legs is the trick to climbing stairs. It might be more involved in having a flexible backbone, to better deal with the transitions top and bottom of the stairs.

Hi Dan - like Nick said, it's a 6 legged robot with 3DOF per leg. Total length is about 20cm. Leg segments are about about 6cm long. Servos are positioned in such a way that there is essentially a ball joint at the base of the robot (ie a hip) and a single joint (a knee)

6cm away from the hip.

-Mike

So, it's gonna be far too small to climb stairs in any event. Maybe can scramble over a book, etc. My stiff-legged walker design does not have enough ground clearance to go over obstacles, but I have thought about adding hinged segments so the back could flex.

It has about a maximum of about 12cm of ground clearance - thus the mechanics allow it to go over small stairs.

But this is completely irrelevant and off topic. Stairs are not a goal for my robot.

-Mike

In case 2DOF legs you are much more restricted in selection of next point for particular leg; maybe it is possible to realize stairs climbing using 2DOF legs, but it is much less tricky by using 3DOF legs. The more DOF - the less restrictions - the more abilities; stairs is only one example from the long list.

By the way, 2DOF implies gliding of foot over floor at least on turning; 3DOF provides possibility to avoid gliding. Sometime it is useful and always is more efficient (less energy dissipated).

It is possible also to use "passive" DOF, i.e. not powered and not controlled directly (implicitly) but i don't know real samples.

• * *

Mike hexapod:

Nick

There is defnitely a problem with the usual cantilevered leg designs, ie where the legs rotate around a point on the body, in that the distance from the foot to the body varies by a huge amount when a step is taken. So, either the feet have to slip on the floor, or else the legs have to undergo some very complicated kinematics. The longer the step, the worse the problem.

Inverse kinematics are incredibly simple for a well designed 3DOF leg. Anybody that passed trig in highschool shouldn't have to think too hard to figure it out. This is the inverse kinematics solution for my robot's legs:

double hsquare = x * x + y * y + z * z; double h = sqrt(hsquare); theta[0] = M_PI - acos((leg2length * leg2length + leg3length * leg3length - hsquare)/(2.0f * leg2length * leg3length)); theta[1] = asin(z/h) - acos((leg2length * leg2length + hsquare - leg3length * leg3length )/(2.0f * leg2length * h)); theta[2] = ((x > 0) || (y > 0)) ? atan2(y,x) : 0;

(essentially it's just an application of the law of cosines). If your leg segments are the same length it's even simpler.

These calculations take, as I recall, about an eighth of a ms to calculate on a 20 mips 8bit uC. I calculate it after each set of PWM pulses is sent to the servos.

-Mike

Do you specifically deal with the fact that the distance between the foot and the side of the body varies by a huge amount during the course of a step?

What do you mean? Using inverse kinematics takes care of any such problem, unless I'm not understanding what you're saying.

-Mike

I originally planned on optimizing the inverse kinematic equations - but the floating point emulation is plenty fast enough for me. I mean like I said, it takes about an eighth of a ms to calculate out all three angles. Being that I'm running the servos at 80Hz, that leaves me with over 12 milliseconds to do anything else that I need to do.

-Mike

Hi Mike,

It is possible even more decrease theta[i] calculation time (probably in a few times) using polynomial approximation of a trigonometric functions instead of call functions from math library; errors may be about 1 degree by using approximation (polynomial regression for really used angle range, not Taylor series) of 4..5 degree (requires

4-5 multiplying and 5-6 adding only by using Gorner's scheme).

Nick