Is it possible to replace Robotis actuators by cheaper ones?

Tags: #<Tag:0x00007f5ce713ccc0>



Discussing about building Poppys with some other Fab Labs, some people wonder if it would work replacing actuators you are using by simpler servo motors ?

How did you scale actuators you are using ?


Low cost version?
What motors use poppy?
What motors use poppy?
Help with which servos to use for project


a first quick response (more to come by other forum members): we are currently using hi-quality servo motors since the platform was initially made to explore dynamic semi-passive biped locomotion, which requires at least some minimal dynamic control over compliance, combined with high-torque on some joints, and a good precision. The drawback is indeed cost.

Yet, work is ongoing so as to make the skeleton modular in the sense that the end parts of “bones” is modular and can be easily redesigned to accept simpler and cheaper servomotors.

Integrating these simpler servomotors is not right now a priority in the Flowers team, but we would be delighted if members of the open-source community would explore these modifications so as to allow for example the building of parts of Poppy with these other servomotors,

Best regards,


Thank you Pierre-Yves.

We’ll see if the community can find answers :).




While looking at the BOM, I saw this line:
21 Dynamixel MX-28T   $220/each thus a total of $4620

It seems Herkulex DRS-020X Series are similar:

One can buy them for 118,94 € here:

It is a great difference!

The DRS-010 series is even cheaper (nearly half) but the difference seems to be the angle amplitude. Is it mandatory everywhere to have such an amplitude?



To complete @oudeyer, we are also interested by the physical human-interaction. For this kind of experiments/applications, smooth motions and compliance are really important. Without them we would not be able to explore interaction and motion as we did during the artist residency.

Also, in the initial design, the arms were actuated with AX-12 motors. Compliance was so bad for interaction (from our point of view) that we decided to replace them by MX-28.

Yet it is possible to replace them by another ones. When you will have few models in mind, we can discuss about their integration and potential limitations.


Yes you are right, they look like a good alternative to Robotis ones.


Good catch, thank you Loic :).


DRS-0201 in stock on
There is a discount price of $116.64/each when buying 25 or more.
$35.34/each for DRS-0101 over 25 pieces.

On 105€/DRS-0201 and 33€/DRS-0101.

I imagine using HerculeX servos would imply redesigning structure pieces.

For the difference between 0101 and 0201 are you sure it is amplitude ?
I read a half torque for 0101, and brush dc(01) vs coreless dc(02).


The HerculeX is very interesting motor.
The difference between 0101 and 0201 is the torque ( 12kg/cm-> 0101, 24kg/cm-> 0201 and the gear box (plastic for 0101 and metal for 0201).

If you use HerculeS servo, you must redesign the part.


I’ve found the SpringRC SR-518 knock-off of the AX-18s on ebay (e.g.: ). Has anyone used these servos? They are supposed to be “Dynamixel compatible”. Could this be a cheap option?


It could be but we use TTL communication (3 wires). I have looked a bit over the internet and I only saw RS232 communication for the SpringRC. The electronic board we are designing will not be compatible (but can probably be hacked).

Also is there more powerful servos, i.e. comparable to MX-64 ?


I’ve found some documentation on the SR-518 that includes a lot of great information about their communication protocol. It is half duplex RS485 (4 wire), just like the Dynamixel RX and EX series. So, not great news, but not terrible news, either.

There is also a whole chapter on creating a servo controller with the AVR SCM, including schematics, Eclipse screen shots, and sample c source code.

I tried to attach it, but apparently I’m too new to do such a thing. :frowning:


If anyone is interested, we have been working on a new kind of Servo, that is USB controlled and uses 2.43mA (vs 100mA for MX-28 / 64) when not moving and has the possibility to drop down to 100nA when in standby to preserve the battery.

A later version may well use USB 3.1 which allows 12v at 5A down the USB cable so only one cable would then be required. But we have to wait for the standards board co confirm the new specification in July.

It is likely to be exactly 153% of the size of a MX-28 servo and initial indications are it will be between 40 & 50 Kg/cm of torque. (i.e. print poppy at 153% scale and just add these new servos, so poppy would now be 128.5 cm tall).

This means that if you print out a version of poppy with only MX28 servos at 153%, you are all ready to go (though the internals of the head will also expand)

Further, this servo has a JTAG debug port and open source RTOS on it so people can expand it and add their own routines, as well as it has a 3D accelerometer, current sensor, temperature sensor, can detect if the motor power is lost and in ASCII mode can be controlled simply via any program (even a terminal emulator) and looks like a standard serial port over USB (i.e. a CDC device) and has been tested on OSX, Linux and Windows.

We are looking at a binary mode where a user can use LIBUSB and Python (as an example) to control the servos and react to various thresholds, for example if the servos detect a freefall or sudden knock, it can interrupt the host and interact in whatever manner is required.

We have built the initial prototypes, and are looking at build the new scaled up prototype that I have mentioned above.

The price is going to be much cheaper than a MX-64 or 28 and runs at 11.1v (i.e. 3S LiPO battery)



It seems really promising!
@DouglasPearless: I am a bit curious about how it looks like and what is inside. Do you already have some information available (documents, videos, code, website)?


Have a look here this shows a prototype before we move to the new case design.



Back to the SpringRC servos: The forums software won’t let me upload the user manual I found for the SR-518 (Dynamixel AX-18 clone). For those interested, I’ve uploaded the PDF to Box (a trusted file sharing service). Link is here:


This servos look very interesting for its price. Compatible with dynamixel communication protocol. The problem is the 18kg/cm torque, because, from what i saw, poppy uses 28kg/cm torque.

Good for a mini poppy :wink:


@Matthieu, @All
This is my first post and I have joined today only. I really congratulate entire poppy team for their work. Am a software developer but have some noob questions on servos and mechanical designs.
Actually I agree with the points like MX 28/64 servos offer more torque over weight and precision control to make fluid robotic gaits. But the total cost goes so high that I am searching for a cheap servo.
I want to be involved in this project and contribute some modules. But the price is a big challenge.
Am new to robotics so can some one tell me if I choose Dynamixel AX - 18 servos can we develop for poppy (reshaping frames for AX 18)?
I mean if using AX 18 we can get good gaits, can some one help us with AX 18 compatible frames?

Please pardon me if am posting to wrong box :blush:


We understand this issue but with cheap motions control and physical interactions will be very limited. Also the gait control will be maybe more difficult.

That being said, there are indeed several applications people can explore without high-end motors. Especially when you are not interested by mechanics or control but rather by software or design. Fortunately, the way we designed the robot (modular and easy hackable) makes this kind of modification rather simple. It will of course require some work to adapt the frame but it should be achievable within two weeks of work.

If we create a Poppy compatible with AX-12/AX-18 motors, we will certainly have to reduce its size to about 50cm. A problem will be the head, for the moment the electronics we have is not ready to be embedded in a smaller robot. We have a plan to do that but it won’t be as fast as changing the actuation of Poppy. So for some times the electronic will be quite non-optimized and you will have to plug all your components using usb cables (camera, motors, sensors…).

How to build Poppy?

Thanks @Matthieu that was an awesome response I must say.
One questions on top of that is …
Can we have the frames design optimized for AX 12/18 so that we can start developing on smaller size poppy and later on we will try to fit the electronics on top of … or may be control from out side?