Is it possible to replace Robotis actuators by cheaper ones?

Tags: #<Tag:0x00007f5ce81eed70>


Little off topic but we have new servos MX28AT MX64AT with advanced heat factors.


my opinion is there is no futur to design a robot that get its ability to do something based on hardware. In theory, only the software should rules and it should be able to adapt to different hardware. The would be no evolution if we are stuck to the specification of hardware.
We are not built to swim, but we can learn.


I’m not sure you can learn how to fly… even if you try real hard ^^


No, not possible to find stall torque. But it is weaker than ax-12 for sure. This motor is not for strength unfortunately.
On the other side, it is very funny to use it. You can obtain behaviors very special. I made a catapult to launch objects very far with precision.

I do not think this motor can be used for a biped robot. It is good to practice dynamic robotics.

To answer @nosys70 I could not do a catapult with a classical modelism servo. I think there is a world to discover if you work in parallel on hardware and software.


i agree, bust most people working on hardware usually ignore half of the rules.
for example, biomechanics is using very different solution and approach than pure mechanics.
Most of the design i see in robotics are usually pure mechanics (when not bad mechanics).
I doubt my knees knows exactly a which degrees it is bending, but it still works fine.
So if you end up with a servo problem like being not precise enough, you got a bigger problem
than just to find the right servo. and if by any chance you find one that is ok to walk , you could discover soon it does not fit to run or jump or dance or … swim.


So what kind of creature shape you think would be feasible with these motors and opening interesting experiments with dynamics ? Quadrupeds ? Snakes ?


The most interesting creatures to do with these servo are dynamic ones. I imagine :

  • a tripod jumping robot like a kangouroo
  • a precise catapult (this one is very simple and could be very didactic since you can touch a PID controller)
  • a Spring Louded Inverted Pendulum (described in Russ Tedrake course)
  • the rear part of a quadruped
  • an acrobot

I am going to plug the motors at the ankles of Alion to test.

“so many birds, so little time”


I confirm the MX-12W are too weak to be in a “child sized” robot.


Has any progress been made on this? I would love to build a poppy, but the price of parts is too high currently. Has anyone made successful modifications and scaled down?


To my knowledge, yet noone modified the parts to make them adaptable to other servos. But you can have a look at this post:

It shows a part that seems to fit a standard servo into a dynamixel case.

Maybe @new2bots can tell you more.


Some work is involved to use this servo adaptor. First the mounting tabs on the standard servo need to be removed and filed down. The bottom of the servo is removed. Slide the wiring harness through the hole in the adaptor and carfully press in the servo hand pressure is all that’s required but it is a snug fit. The. Replace the 4 screws from the standard servo through the bottom of the adaptor and voila an inexpensive servo that fits dynamixel sesigbed robotics components.


in the same range, some food for though here

on one the best extract:
"the only resemblance these robots have to real humans is external and superficial: they look like us, but their operationalprinciples are so far removed from our own that there is little prospect that their incremental refinement will result in anything better than more of the same.

What is the problem? Skin. This opaque envelope conceals its contents from our perceptual apparatus, and prevents us from paying enough attention to what lies within. We therefore overvalue human-like shape, and undervalue human-like functioning."

and the best proof is the walking mechanics that use no electronics and still walk like an human.

This explain you cannot build a robot by designing an envelope that look like an human, and then trying to mimic the movement by creating a structure that has nothing to do with what is inside the evenloppe. Unfortunately , poppy is build like that, and that explain why it is still not walking, why you need motors so expensive etc…

A good robot should be able to stand up by itself, using basic isoelastic concept.
most of the movement then could be initiated only by a small drive , with a small power.

that is the principle of human body. even an anorexic girl still stand up despite having almost no more muscle. Then if you expect your robot to be a terminator, you can think
to use hydraulic or big motors.


This explain you cannot build a robot by designing an envelope that look like an human, and then trying to mimic the movement by creating a structure that has nothing to do with what is inside the evenloppe. Unfortunately , poppy is build like that…

I disagree! It’s all about dynamics! The proof you suggest, the passive dynamic walkers (PDW), prove exactly the opposite. They might not have motors but their joints are exactly in the same “position” as if they had motors. In theory, if you had a perfect motor by means of no friction, you can install a motor on each joint of a PDW and achieve the same result. Plus that now you can actuate each joint and if you know how to actuate it you can control the walking.

Of course, the way of how to actuate these motors is the key concept and it’s complexity is the number one reason why poppy is still not walking.

I agree though that placing a motor in the knee, ankle, elbow etc is NOT the optimal solution (obviously there is a better one - human being…) but I believe that you can achieve natural human-like walking with this kind of actuation.


the problem is not with joints position. obviously joints are at the place they must be.
if you learn a minimum of biomechanic, you can see that there are a lot of energy management in muscles and tendons. if you look at the way muscles wrap a shoulder, or if you study the ankle, you can see it is not just a problem of placing joints. there is alot of lever effect hiddden in the placement of muscles and the place they are attached. Same for tendons and ligaments.
unfortunately all of this is missing in a robot that implement only joints and motors, and the result is while the development is going further, you can see the problems being more and more about the power of the motor and the way to drive/position them. that the reason why it is becoming more and more difficult because each choice based on a hardware choice is closing the door to other solution.
That is exacltly the opposite of an open project because finally , all the project is tied to a single element (the motor) , and even to a single model with all the limititating specs.


I still don’t understand your point. On one hand you say that

And on the other you refer to biomechanics,

Anyway, even if I don’ t have a minimum knowledge of biomechanics I understand the importance of it in the design of humanoids. My disagreement was on the statement that poppy is still not walking because of its structure. On the contrary, one of their original goal was to study the “impact of morphology in biped locomotion” - hence the unique leg design.


Yes, but for the moment, I think it is the most open project for a humanoïd. Do not forget the project is tied to the servo motor which offers the best performances for the moment to study passive walking (the power is not the most important thing for Poppy). There is no open equivalent for the moment. It is on the good way to be more open and open.

The realistic point of view with artificial muscles has another purpose for me and it is more complex and not as accessible as Poppy. But it is interesting.


Well… since it has motors, it can’t be called “passive” walking anymore! :smile:


I don’t know if you ever touched one of these motors MX-28, but it is possible to make the motor very compliant (as if there were no motor), that is the reason it is very useful the study passive walking since the joint angle can be measured. Only this kind of motor can do this. I though about articulations with clutch and springs but it is very sensitive to tuning, external forces and fatigue.
The perfect motor should be the one which also emulate a spring with sensor… This one does not exist yet (maybe with the brushless motors of Boston Dynamics of Spot…)


Thot what do you mean? That the motor by itself is very compliant or you can program it to make it very compliant?


The motor is compliant when you set motor.compliant=True. If you set it to false, the motor goes to the goal position you want.