Here is a much more detailed shot of an assembled thumb made from SLS printed components. The fit and action is definitely the smoothest weโve seen so far in printed parts.
It's a firmware development sort of a day here in DexHand-land. Implementing Google Protocol Buffers for USB messaging to make it easy to adapt to different languages and platforms in the future.
Been a while since weโve dropped a video of the DexHand. Hereโs the result of some Saturday morning hacking with a crude initial calibration. Still surprised by how fast this thing moves!
Test sequence, with most of the guts tucked inside the hand. Moving onto thumb control next. There's 4 DOF's in the thumb so I need to figure that out.
Not the most exciting motion path, but the USB control stream to and from the DexHand is working. I am testing a 50hz servo position update loop upstream and 10hz back down at the moment and it all seems healthy. Obviously, the servos don't update that fast, but was curious.
Built up a finger set and finalized all the plastics for the first cobot forearm build. This one is going on a UR3 at the shop for some life cycle testing.
Latest rev of the palm is back together. Just going to go check the endpoints to make sure everything is spinning the right way and then back on the forearm. Really pushing to have a DexHand to play with this weekend!
Sneak preview of the
#cobot
forearm for the DexHand. This unit has a flange adapter to fit on a UR3, but we're planning to make them for any arms that customers might want to test the hand with. What are people using our there these days for arms other than UR?
Testing out the latest prints. It's amazing how colour coordinating the materials can make you look like you mean business. This is the new forearm and palm rev going together. Hoping this is our final spin before retolerancing everything for SLS printing.
Alpha unit mechanical assembly achievement unlocked! All the pieces are in the places, tendons connected, and the motors work one at a time. Going to see if we can get this tuned up!
Some early morning coffee fuelled hacking so that I can get moving on the next rev of the firmware for the hand.
Got lucky - the tweaks that were needed involved far less disassembly than they could have!
I think thatโs the prayer of the
#robot
hand builder. โPlease, god of
Back to the drawing board on a couple of the palm servos. Hoping to print a revision today to get things back together so we can finish up the alpha unit. Sometimes itโs two steps forward, one step back.
Our dark overlord
@iotdesignshop
just open sourced an internal tool called Feetech-tuna. If youโve ever used FD.exe to commission Feetech servos, you might want to have a look.
@saba_khalilnaji
Weโre working on this. What I have heard most frequently from prospective customers:
- Hands are expensive
- Hands are hard to control
- When they break, it takes a long time to get them fixed
- They break a lot
I think the control issues would actually be resolved if we dealt
I don't normally machine a lot of parts at this scale, but it was kind of a fun challenge to build these motor transmission brackets for our upcoming robot arm.
#cnc
#robot
#humanoid
20 Open-Source Robotics Projects ๐ฆพ
- ON GITHUB -
Here are 20 fascinating projects to accelerate your robotics knowledge and unlock valuable learning experiences:
1. ROSbot 2.0: An open-source mobile robot platform
2. Magni: An affordable robotic
@lethic1
Fantastic! That looks great. It's always amazing how simple things look once they are working, but we appreciate all the efforts and tweaks that go into a build like that.
@CubedUnit
The biggest thing is avoiding any surface defects created by SLA or FDM processes. Thereโs a ton of internal chambers and routing for tendons and wiring so it can be quite a chore to make sure everything is cleaned out and fits well after prints.
@rwik_rana
This version of the hand is designed to attach to the end of a cobot arm. Targeting a UR3 in this case so the wrist articulation comes from the arm. We do have a version of the forearm with an articulated wrist as well which provides more natural wrist motion.
@SlimActualX1
@Fennex628
@_nightsweekends
@_buildspace
If you can find some sort of smart servos that will manage current and temperature limits that also fit your form factor, Iโd highly recommend it. On the V2 DexHand we are using Feetech 3032โs and itโs a game changer.
@ramkumarkoppu
Use in prosthetics is a fairly frequent question in our inbound contacts. Itโs not a use case weโre super familiar with, so hard to answer. Our belief is that the regulatory overhead is high and that the DexHand likely has more DOFโs than are controllable. But doesnโt mean it
@th1nkp0l
Thank you! The sleeves are 2mm OD PTFE tubing and the tendons are 80lb Kevlar reinforced fishing braid. The combination allows for pretty smooth motion and good strength.
@kenny__shaw
@lethic1
Weโve been experimenting a bit with the diameter of our servo horns and winches to try find the balance of speed/torque. Toward the lower end of the diameter range we found the issue was not so much being strong enough but actually overcoming our mounting systems for clamping the
@ramkumarkoppu
@RemiCadene
@Ilir_AI
@OrangewoodLabs
@trentshumay
DexHandV2 is fairly agnostic - the hand itself just provides firmware level access to motor controls and some tuning like endpoint adjustment and position/status streaming across USB. Weโll have libraries for driving it via C++ and Python, and a ROS 2 connector. Youโll be able to
@Anthrobo
Thanks so much! Thatโs the goal. Iโve got a bunch of CNC machines downstairs that currently need to be โloaded by handโ but hoping that will be a robot hand soon instead of mine!
@adams10ne
Kind of. We use them for CNC machine tending at our shop, so there are URโs that can be misappropriated for other robotic experiments. Iโd actually really like to try set up a pair for a bimanual rig once everything is working
@QinYuzhe
Incredible work. This is amazing to see this type of learning come to fruition. Is there a hand tracking suite for the RealSense camera already, or is that something that you built?
@ramkumarkoppu
Nothing thus far. Obviously, thereโs lots of support for UR3, but we havenโt developed a simulation model for the hand yet. Itโs on the to-do list.
@MRiwu_u
The 3D printed parts pictured here are largely SLA (resin) printed on a Formlabs Form 2 using Tough 2000 and Tough 1500 resin. The forearm is PLA-CF, but that's just for prototyping. All the parts will be moving to SLS Nylon very soon.
@Teebotjs
The ones in that photo were SLA printed on a Formlabs printer. The resolution on the Form SLA machines is very good, but they do require a lot of supports which is quite a bit of work to clean up. With the new SLS prints, we don't have to worry about support structures.
@ramkumarkoppu
There are actually outer shells that wrap the forearm that can be totally customized/branded. So, definitely something that can be accommodated.
@ramkumarkoppu
Yes. Absolutely. The finger tips and knuckles connect to the long flexors that travel down into the forearm. I didnโt have those connected in this video, but Iโll post the whole thing working soon.
@Fennex628
Thanks Ian! To be honest, Iโm not 100% sure on the torque yet. I have the limits on the servos set at 40% at the moment to avoid catastrophes during testing. That provides a very firm grip on stuff Iโve tried to grab with the hand. No idea what happens if I bump that higher yet.
@kenny__shaw
@lethic1
The bigger issue weโve had with the STS3032 has been that although they support multi turn, the encoder is single turn, and position is not absolute if you cycle the power so you need some sort of homing routine if you want to use the multi turn mode. So, small winches expose
@zshobbs
Also, FWIW, weโre a team of two people thatโs come up with this and built it - trying to figure out if there would be a path that makes it sustainable to keep developing on it. Thatโs maybe important context. The dream would be to make the project self sufficient somehow.
@CKohstall
Weโre taking a run at the dexterous hand challenge. Just getting ready to roll out beta hardware. We welcome any feedback on our project and invite everyone to follow along!
@Adamnathe
The cobot version in the photo has 16 driven DOFโs. We have another version of the forearm for attaching to humanoids that takes it up to 19 with an actuated wrist.
@SlimActualX1
@Fennex628
@_nightsweekends
@_buildspace
No worries! The biggest difference is that conventional servos let out the smoke when they overload, but the smart serial ones will shut down themselves before they blow up. So, that saves a lot of repairs.
@MRiwu_u
Definitely! Right now, our limiting factors are not the resolution of the grasp. They're far more fundamental than that - just basic controls and getting the hardware and firmware working are job
#1
. But, we're definitely eyeing the future and watching all the options keenly.