Drossel - Humanoid Robot Project

(Latest post update: 2/1/2018)
(Check latest post for update)

I'm working on my second arduino robotics project and I thought I'd make a thread here encase anyone finds it interesting as I progress along with it.

The design I chose is a humanoid robot called Drossel from some small Disney short films.

Reference Photo of a drossel figure.

This is the latest render of my design i'm drawing up in blender.


(The black object you see floating above the left elbow and inside the legs are 1/1 scale mg996r servo's.

(Update: I'm replacing the mg996r's mainly because of a large deadband but also a lack in power).

The leg connections are almost completed though because of the shape and size of the servo's I had to adjust the height from 24 inches to 31.5 inches for the whole robot. I believe that will make it too tall for the servo's that i'm using which are MG996R servo's specced around 10kg of torque. I did get a pair of legs walking using MG90S micro servo's that were 8 inches tall but just barely and they weren't supporting the weight of the battery pack. Those servo's are only specced for 2kg of stall torque supposedly but are likely less.

So, as of now i'm designing the inner workings of the hip/waist. The robot torso needs to sway left/right but I also would like the ability for it to bend forward to help pick up an object on the floor. It will be interesting coming up with a working design. The arms/head shouldn't pose any complex problems except for fitting the servo's inside the arm joints without distorting the aesthetics of the robot.

The only problem I foresee at the moment is needing to replace the leg servo's with higher torque ones. I'm building from the bottom up and using my past project to compare I should get an idea of what kind of servo's i'll need when I hit that roadblock.

Nice design! I'm working on a humanoid robot too. Maybe you could give me some suggestions?

This is my 2nd robot so I feel like I can't give any good pointers atm. Though if you plan to make a handful of humanoids I feel like having a cheap but decent 3d printer would come in handy. This way you can either print off replacement parts for certain 3d printable robots or you can design one yourself using various 3d programs like Blender 3D which I use.

Also, if someone was new like me I would probably suggest buying chinese servo's instead of expensive brand name ones until you get an idea how various motors perform. This way you don't blow a lot of money on a bunch of servo's that aren't up to the task because you lacked experience with various kinds.

There's probably a bit more I could suggest but i'm to tired to come up with more lol.

No new update on the robot except that I worked out how the servo's connect to the frame after a handful of 3d test prints. I should be able to start printing the legs this week.

I just made it simple and hacked a Meccanoid G15. It's powered by an 86Duino One, which has the following specs.

-300MHz 586-level CPU
-128MB DDR3 RAM
-2MB Flash
-microSD Card Slot (supports up to 32GB)
-accelerometer/gyroscope

I'm still working on the software to control it. Using a 3D printer to print parts is great, but 3D printers are very expensive.

“3D printers are very expensive”
Not so anymore.
You can get a usable printer for <$250:

I bought one recently, the Electron 3D - Sixth Generation Prusa i3, onsale for $233 at the time.
We have it assembled, but think the stepper wires are not correct, we’re looking into re-doing those but haven’t figured out quite what to change them to yet.

OK, so they cost hundreds of dollars instead of thousands of dollars. It’s way cheaper, but still a big investment.

You also have the Tiko 3D printer, which (IIRC) has an aim of getting things down to $100.00 for a printer. I am a backer of that kickstarter, and hopefully my printers (I backed it for two of 'em) will be shipping fairly soonish (they just got the first 100 units in and are testing them before final thumbs-up - unfortunately, I am not a part of that round).

There have also been other $100.00 and sub-$100.00 3D printers out there; basically if you make the build volume small enough (and/or get the scale of production up high enough) it is more than possible. Though I think for any useful build volume, $250.00 (approx) is about the minimum you can expect. Plus, the market is saturated now with printer companies and products - and the potential market is one where pretty much anyone who wants or needs one, has one.

it still isn't at the point where it is considered something for ordinary users (which is keeping the demand flat) - and I doubt it ever will, because people want to use these for more than printing pre-made designs - they want to dump their own designs. However, working in 3D - despite advances in various software - is still not a very intuitive process for most people.

Printed off the foot. Going to print off the shin piece once I got everything correctly sized to fit together since I think my servo 3d model is abit off.

I think my x axis stalled from barely overheating a couple times during the print (had this issue with long prints in the past). I re-installed a fan back onto the motor and I think that should take care of it.

Nice work on that foot!

Thanks! Working on the shin now.

Update:


Front


Side
(I left out the middle shin panel because of an error somewhere that keeps it from sitting flush with the top and bottom shin part. Also, it was supposed to be 1 piece but i was tired and it ended up as 2 pieces lol.)

I'm still working out kinks and stuff so it's going to be reprinting again after adjustments and if i'm satisfied ill mirror and reprint the other leg.

Lets talk torque. Its measured in Nm not kg. You mention this thing will be about 1m tall, so
to balance or manipulate that long a lever arm there will be a lot of torque needed. What is
the estimated total mass?

(Lets assume 500g, 0.9m long, uniformly distributed mass, to hold that out horizontally takes
about 3 Nm, which is a lot. To balance from 10 degrees to the vertical would only be about 0.5 Nm
which would be within the MG996R's nominal ability. All assuming 500g mass of course. I suspect
the backlash / play in a servo is going to be an issue. Note that stall torque rating for a servo is not
a sustainable rating, the servo will overheat and probably melt if held at that torque for a prolonged period.

MarkT:
Lets talk torque. Its measured in Nm not kg. You mention this thing will be about 1m tall, so
to balance or manipulate that long a lever arm there will be a lot of torque needed. What is
the estimated total mass?

(Lets assume 500g, 0.9m long, uniformly distributed mass, to hold that out horizontally takes
about 3 Nm, which is a lot. To balance from 10 degrees to the vertical would only be about 0.5 Nm
which would be within the MG996R’s nominal ability. All assuming 500g mass of course. I suspect
the backlash / play in a servo is going to be an issue. Note that stall torque rating for a servo is not
a sustainable rating, the servo will overheat and probably melt if held at that torque for a prolonged period.

The entire android atm is made out to be 33 inches tall or 860 cm. Each leg is 43 cm or around 17 inches. It got 1.5 inches taller after I revised the chest but I don’t see it getting much taller now thankfully.

I could probably give a good estimate of the final weight but i’m about to print the other leg and waist sections so I can give an accurate value soonish. I’m assuming i’m going to get some stalling problems in the most bottom motor’s in the foot since this is where I had trouble in my 8 inch micro servo legs. I was able to get around it though and had it doing a walk cycle from a start position for a couple minutes no problem which is enough for me to learn from and play with.

This project is mostly a learning experience of forces at play(along with 3d printing large connecting pieces that aren’t warped at all).

im also working on this type of robotics and android applications, as far as the algorithm im been influence alot by AI movies specially recent ones like Chappie Ex Machina and Uncanny Ghost in the shell and so on. so Consciousness programming is one thing NOT being worked on by anyone in the field of Cognitive Sciences Neural Sciences or AI, so i know enough to start something practicle after all what you need for this type of applications is not just programming skills but imagination.

EX CORP DOT WEE BLY DOT COM your are all welcome to reach out of hit up my facebook page as well i did started a blog for this type of conversations on my website so feel free.

Update: Got both the legs and hip connected and calibrated the motors. They balance alright though they aren't quite rigid enough yet because the 3d pieces I printed that connect to the servo's don't completely secure them and can slide around a bit. The plus side is i can slide them on and off relatively quickly to tweak things so its fine for now.

image removed, redesigning robot

I haven't tested the motors with everything all hooked up but judging from the performance of a single leg i don't think there will be any issues. At least not until the upper torso is made adding more weight.

This look absolutely great! Looking forward to seeing some updates about this one.

danphobic:
This look absolutely great! Looking forward to seeing some updates about this one.

Thanks! I'm waiting on some new servo's that should be 2-3x stronger then these current ones and perform better all around so it'll be about a week+ before anything new is posted.

Good news is the upper half of the android is almost drawn out and im just waiting to make exact measurements on the new servo's before I start printing again.

Ok a double post but since my last reply was in March I think its alright.

Anyways, project reboot! I’ve been redesigning the robot from the bottom up to be sturdier with more functionality and i’m close to obtaining 50% completion. I just finished designing and printing the left leg and hip section. Once i’m happy with how that leg performs ill mirror duplicate it,print the right leg, and hopefully have a 2 legged robot that can take some steps.

There is some weak points and places it needs a redesign but it does function with the correct movements. I think i’ll wait till the other leg is completed before posting a photo of the 3d printed counterpart.

I’ve improved the knee joint so it functions much better now and applied the improvements to others. Getting closer to the point where I can build another leg and have it attempt its first step.

Update on progress:

The entire leg except 1 motor attachment has been printed and assembled. I would have that part done now but i'm short one motor so it will have to wait till my next batch arrives. The leg from the knee down is very sturdy but the thigh to hip section has weak areas that need to be reinforced later.

The leg should be far enough along now that I can test some compound movements and upload the results. I can't use this plastic breadboard with the amount of amps ill be pulling so I need to go shopping and rig something up first before I can do any tests.