Doing some quick sketch ups and brainstorming.
The parameters that were given to me was to achieve a minimum stroke length of 6", a stroke speed of 6"/sec, and using the Dynamixel MX-106 motor.
The Dynamixel MX-106 motor. The given specs of this motor was the most restrictive factor, due to it’s low RPM and average torque rating.
http://www.trossenrobotics.com/p/mx-106t-dynamixel-robot-servo.aspx
For the structural components of the actuator, I decided to go with carbon fiber tubes that would be cut to size.
Since the MX-106 servo has a low RPM, and the ballscrew for the linear translation would cause further reduction, a gearbox was called for, but had to be used in reverse as a speed increaser. This 2-stage, 20-1 ratio, interplanetary gearbox was chosen for it would give us the desired output RPM, and would allow for a compact/low profile design by nesting it inline with the CF tube.
Isometric CAD rendering of the actuator design in it’s extended position. The primary housing was made clear to get an internal view of the ballscrew.
Isometric cross-sectional view.
Side orthogonal cross-sectional view.
Drawings for all the parts that were to be machined manually. All parts except 3 (the linear encoder/rear mounting assembly) was made on manual machines.
A poorly shot image of me squaring the end of the CF tube on a lathe. I had to wear masks and gloves in order to avoid inhaling the carbon/epoxy dust and getting splinters.
Was granted special permission to use this huge beast to turn the hardened steel ballscrew.
Even with a large/ rigid lathe, the fact of cutting hardened steel with pretty intense interrupted cuts from the threading causes tool & part deflection, which can be seen in the bumps along the turned shaft portion.
Just a quick glimpse of the progress thus far with the parts I have.
CAM was done on one of my favorite CAM software; HSM Works.
Using the Tormach PCNC1100 to machine the parts with complex features since it would take too long on manual machining.
The three CNCed parts are finished and assembled: The motor mount, the rear floating mount, and the front actuator mount.
The gearbox we chose had face.back plates that retained the gears and doubled as bearing mounts. Since we want the gearbox tucked within the housing, I turned the square faceplates into a cylinder of 1.5" to pressfit into the housing tube.
All the gears being greased with some nice Moly grease
The output gear was pressed fit and bonded to the ballscrew, then fitted into a bearing, and seated in the gearbox. Attached to the MX106 motor, you can see the input adapter i machined to transmit power from the servo horn to the gearbox through a double-d shaft.
Motor mounted, gearbox tucked in, and mounts attached to test fits before final assembly.
Since I forgot to take a shot of all the parts before partial assembly, here is one of them in some major sub components.
I turned some aluminum end caps/rings for the CF tubes to protect the edges and for a aesthetics. The polished aluminum paired great with the black of the carbon fiber.
Ballscrew and ballscrew nut. This is my first time implementing a ballscrew into a design, and I found the ball bearing recycling system on the ballscrew nut extremely amusing.
ballscrew nut assembled to the actuating tube.
Post final assembly: retracted position.
Post final assembly: extended position.
A more frontal view of the actuator
Linear Series Elastic Actuator (LSEA) Project
In robotics applications, it is usually desirable to implement rigid actuators due to the desirability of precision and repeatability. While those qualities are paramount for a high performance traditional robot, the requirements are changing as we develop robots not to just work FOR humans, but work along side WITH humans. As robots interact more with people and the environment created by us, they require the capability to “feel” and a certain degree of mechanical flexibility/compliance. Series Elastic Actuators (SEA) address both those requirements, and are used in many modern humanoid robots. Through the implementation of a mechanically elastic component in series with the load and drive, the actuator is equipped with ability for force feedback/control, as well as shock absorbing capabilities.
The SEA that I designed was a linear actuator variation that is intended to be used as a demonstration/study unit for a low power/compact robotics application for a robotics lab in South Korean, as well as an exploratory prototype actuator for a new humanoid prototype in our lab at UCLA.
The main goal of the design was to use a selection of pre-existing hardware in the lab inventory, as well as to make it as compact as possible with the given performance parameters. In order to achieve the goal, the actuator components were all designed to fit axially, where the interplanetary gearbox was nested directly inside the actuator housing, and the series elastic component clamped around outside of the housing. A linear encoder (transducer) was attached to the elastic component in order to measure spring compression, which can return a force value. To get a higher performance efficiency,
Working on this project was a big learning experience for me, as I read several research papers to learn about what a SEA was and the various factors that affect it’s performance, such as actuating precision, stiction factor, bandwidth, and efficiency. It was interesting to see the applications the technology was capable of being applied to, and it was an exciting opportunity to brush up on my design & manufacturing skills.
Some problems that I came across in the project was the implementation of the interplanetary gearbox, and some friction/binding experienced in the elastic component. The interplanetary gearbox had a little bit more backlash than desired. A potential fix for that would be exploring other transmission designs, such as a harmonic drive.
Role |
Designer, Manufacturer |
For |
RoMeLa (Robotics & Mechatronics Laboratory) |
Date |
August 30th, 2016 |