View Full Version : [Project] CHARLI humanoid

06-29-2010, 02:59 AM
I figure I'd post a little info on the CHARLI humanoid robot here (as I've been working on the project for the past year)

and recent news, it won 3rd place in robocup 2010 adult size competition!

a link to the full writeup (on Tech's website)

if you have any questions I'd be happy to answer!

Also, I apologize for the image size, it was a poster I made for our senior design fair thing... I'd resize it, but I'm not entirely sure how to do thumbnails on this forum yet

http://www.narmd.com/photos/CHARLI/Poster%20V2%20senior%20design%20CHARLI%20helvetica .jpg (Sorry, had to make this a link, image was so large it was breaking our front page)

06-29-2010, 05:40 AM
Great Robot!

One question, where did you get the 2GHz Fit PC2, I can only find 1.6Gz max, did Compulab do a special for you or are you overclocking (which would run quite hot though)?

Another question do the RX-64's work well for the shoulder? My new robot is goiing to have Dynamixel arms, but I was considering EX-106 for the shoulder with RX-64 for the elbow, so seeing your scheme has made me wonder if RX-64 would de the job. Any idea what these arms can pick up?

Anyway, great job!


06-29-2010, 06:29 AM
CHARLI-L is pretty cool, but I'm more interested in what actuators you guys are using on CHARLI-H's legs (as seen in this photo):


Are they custom, or off the shelf?

- Jon

06-29-2010, 09:44 AM
risknc. Do you feel that 8 EX-106's per leg (not counting hip yaw) are adequately powerful enough for a robot of this size? If not, what do you feel would be adequate power?

06-29-2010, 10:38 AM
Wow. That's an excellent humanoid. How did you make the white plastic parts?

06-29-2010, 05:15 PM
> Do you feel that 8 EX-106's per leg (not counting hip yaw) are adequately powerful enough for a robot of this size?

billy, did you notice those 4 springs on each leg? I bet it practically removes most of the weight.
And I thouhgt I was crazy to think of doing something like this on Bioloid.

06-29-2010, 05:19 PM
billy, did you notice those 4 springs on each leg? I bet it practically removes most of the weight.

Gravity defying springs?! :O

Great humanoid! I've chatted with Dr. Hong a bit about this bot and his heavier brother, fantastic MechE work from all of RoMeLa!

06-30-2010, 02:20 AM
as far as the 2ghz FIT PC goes, I pulled those specs from another powerpoint that I believe was incorrect (sorry) it's 1.6ghz and we're definitely not overclocking (really no point with the current workload we're putting on the processor)

The choice to put 64's in the shoulder was based on a payload issue, we wanted to keep the arms as light as possible while still being able to pick up light objects (i.e. cup, tennis ball, ect), if you're looking to pick up something moderately heavy (i.e. 2+lbs) I would look at something like a 106, also, the more weight you add further from the shoulders (more actuators) the less you'll be able to pick up. we actually had a nifty solution for actuation in the hands using some 64's located in the chest specifically to reduce weight in the arms. If you want more strength, I'd recommend some gearing, but that gets complicated, as you'd pry want to move away from servos at that point


Those are mock ups of custom in-house ballscrew/carbon fiber linear actuators ( Derek is workin on that stuff right now actually :P) that's for the heavy version of the robot. I heard him crunching some numbers and it looks like he'll be pulling somewhere in the neighborhood of 1000N per actuator (which is SICK maxon ftw!) The plan is for CHARLI heavy to be able to run/jump and do lots of fantastic things


I didn't design the legs (that was a graduate student), but I do know the 106's in the legs are pushing the limits, again, this is the light version of CHARLI, and is meant to just do really basic ZMP walking, no fancy dynamic running stuff, something this large using servos for actuation is pretty rare, precisely because its hard to get servos with enough torque at that scale. The parallel 4 bar keeps the feet parallel to the ground (to reduce ankle moments) and the springs keep it about neutral standing up, so it actually exerts energy to bend down (counter intuitive, but it works better for not frying the motors), one of the advantages over similar robots at this scale is the cost factor, 1 Harmonic drive costs as much as all of the motors in the knee, and if your robot tumbles with harmonic drives, there goes all of your money :P

but yes, we've set the knees to magic blue smoke while testing


The white plastic parts were designed by an artist, CADed, milled from Cast nylon at full scale, then placed in a vaccum former. 1/16th clear thermoplastic was formed over the molds, removed, trimmed, and painted on the inside. it took us quite a while to develop this manufacturing process, but I think they turned out nice


you're sort of correct about it "removing" the weight (though it increases the load when bending down) I think you could do this on bioloid, but the length of extension would be very small, and you'd need fairly stiff springs, I wouldn't recommend doing it with those plastic parts (maybe if you make some metal replacement brackets with spring mounts) also, I imagine Bioloids aren't quite pushing the limits of their actuators, so something like springs might just end up burning more energy in the long run.


Thanks!, also thanks for thumb nailing the image!

06-30-2010, 10:21 AM
Gorgeous work! I'd love to see a video of it walking.

06-30-2010, 01:01 PM
This is the only video I've seen..
I think it does some walking in it.

07-01-2010, 03:34 AM
Vid's that are online (there are more but they're still floating around on people's computers)

CHARLI intro
CHARLI talking
More CHARLI talking
Dr. Hong's Seven species of robot talk (TEDx NASA 2010)
Dr. Hong on robotics research