scout actuator design [updates]
Updated: Jan 6, 2020
The actuators were the first thing that came with designing the scout robot. The reason is, the size and the strength, the power system, the controllers - everything depends on the actuators themselves. I wanted scout to be the size of Cheetah 3, and I estimated it would be about the same weight or less. So I was looking at an actuator with similar specs.
Looking at the paper of Cheetah 3 (linked above), the actuators on Cheetah 3 are custom 230Nm actuators which spin at about 21 rad/s at max speed. So scout's actuators would need to be at least 230Nm if they were going to be similar sizes, weights, and more importantly if the LEG proportions were going to be similar. The speed was less of an issue but it couldn't be too slow. Scout isn't supposed to jump onto a table, but it can't be a snail.
First, it's always good to look at where we're coming from. So stepping back here are the issues with the chipONE actuator. The chipONE actuators were very simply just HUGE servo motors. They were pretty high torque and very easy to control like any other servo. You just gave it the same signal as any other servo. It's positions were always the same at their respective control command (again, because it's a servo). They were cheap too.
The problem is they never provided enough torque for a robot of scout/chip's size, it only translated to 26Nm of torque. Since the servo was cheap, inside was a brushed motor and a simple potentiometer. It was very easy to burn out or break.
One of the biggest advantages to Cheetah 3's actuators, and Stanford Doggo's actuators is the fact that they're based on brushless motors. I'll make a post on brushed vs. brushless motors but the main difference is brushless motors are more efficient and usually are much more precise and capable in our ability to control them. Scout needs brushless motors too.
But because of the fact that scout is a robot dog, it needs brushless motors built for a robot. Ones that can handle high current, brief periods of stall, and are designed to run as actuators and not necessarily just motors that spin. (Meaning they're designed to hold their position or similar). Which is why I went with a newer FRC motor, the NEO Motor because it was designed for robots, is brushless, and produces 2.6 Nm of torque MAX which is a lot.
To get this up to the required 230Nm of torque for a robot like scout. We'd need a gearbox on the motor with a pretty large gear reduction. The NEO motor is designed to interface directly with versa-planetary gearboxes, and if we use a 100:1 gearbox we can get 2.6*100/1=260Nm of torque from the net actuator. Planetary gears allow large reduction gearboxes to have a relatively small form-factor.
So the actuator became relatively simple. A NEO motor, which by-the-way has a built-in hall-effect encoder, in series with a 100:1 versa-planetary gearbox. If we look at the following document, these are the load ratings for a VersaPlanetary. A 100:1 gear reduction with simple loading is HIGHLY not recommended. But if we see the second page, this gearbox would be rated at roughly 100Nm of torque because the last 10:1 stage would be the limiting factor and it is rated at 100Nm. So there is cause to worry, yes, we're building a 260Nm torque actuator where the failure stage is rated at 100Nm. This was an oversight in design for sure if this was a final product, but because we are simply testing the required torques, if we want to use brushless, and etc, it will be fine for now as a testing unit. WE WILL FIX THIS IN THE FUTURE IT IS A KNOWN MECHANICAL ISSUE. Also, it's important to note, though the actuator is CAPABLE of 260Nm, we doubt it will actually require 260Nm to walk the robot. So we will just state the rating of the actuator as 100Nm. And if all loads are under that, the robot will be okay.
For versa-planetary's to work, there's a motor, a motor-mount (in our case a CIM adaptor allows the NEO to mount to the gearbox directly), two 10:1 stages one after another, and an output-shaft block.
Noticing the fact that the NEO shaft is very slightly larger (by about 1/8") than what the versa gearbox is designed to mount to, I realized it wouldn't fit with the stock gearbox setup. Frankly I realized this after all the parts showed up. You'll note in the diagram above, the red markings show you can fix this with a spacer in-between the motor and the mounting unit, decreasing the overall effective shaft length.
These are pictures of the motor spacers, they're 1/16" each so we need two in-between each motor and the grab mounting unit.
Here is the gearbox with a straight output shaft, so this is really the full actuator in one configuration (which is the straight output shaft one).
Here is the gearbox in the second possible configuration, SAME GEARBOX it just has a 90-degree output shaft (shaft's mounted at 90 degrees). I don't know if I'll use the 90-degree mount at this time but I ordered four in case any motor needs to be mounted at a weird angle or something or we need to consolidate space, or we need to mount the motor and drive in a different direction from where the motor is mounted.
The 90 degree drives were ordered without knowing if they'd be used. So for now we have two possible configurations of the actuator as shown above. BOTH ARE THE SAME OUTPUT THOUGH. 12V, 260Nm, Brushless, Encoder-Enabled, actuators!
CALCULATIONS, AS EXPLAINED ABOVE, WILL RATE THE ACTUATOR AT 100Nm.
Cheers :) ~<3