DIYing a Robot Actuator, for Better or Worse

The DIY promise sure is appealing: a fun project, a beautiful bespoke result, and a cost that is just a fraction of what the store wants. Of course, it never works out that way. The project ends up being a nightmare, the results aren’t as good as you expect, and the cost ends up being much higher than if you just purchased the thing. But Brandon Lai is young and I admire that he learned that lesson without giving up when he decided to DIY a robot actuator.

Lai wants to build a robot and is aware of the frustrating truth, which is that robots are expensive. The actuators are a major line item on a robot’s bill of materials and Lai thought he could save some money by building them himself. And he was smart enough to know that he should work from an existing design — MIT’s Cheetah, in this case.

The standard Cheetah actuator design crams a planetary gearbox inside of a BLDC (brushless direct current) motor, yielding a lot of torque in a compact package. Lai decided to take the same approach, but with a cycloidal gearbox instead of a planetary gearbox. His reasoning was that it would have less backlash and would put less stress on the 3D-printed parts.

To build it, he started with an off-the-shelf stator ring. Then, by hand, he wound the copper wire around that. It, plus the 3D-printed cycloidal drive parts, went inside a CNC-milled frame/housing from PCBWay. Lai even attempted to design his on STM32-based motor controller, but a flaw caused it to blow an IC in the drive circuit. So, he had to abandon the custom controller for the duration of his testing.

The actuator worked, but it didn’t quite live up to Lai’s hopes. It wasn’t pushing with as much torque as he had calculated (though that could have been due to limited current from the power supply) and had significant backlash. More importantly, the total cost — not including the controller — was about $400.

For comparison, the CubeMars AK60-39 costs $448.90, exceeds Lai’s targeted torque by a good amount, has almost no backlash, and has a robust integrated controller. That’s the harsh reality: mass-production is really hard to beat.

But though Lai admits his original goal was unrealistic, he isn’t throwing in the towel. He has already gone back to the drawing board to start developing a new design and that persistence is exactly how a person learns the skills to make the DIY approach feasible.