
21 | robotics @hkuniversity | prev @miracleplus_hq meche @ sz innox cs @ uq
Hong Kong
Joined August 2022
- Tweets427
- Following717
- Followers679
- Likes940
I'm building a <$5k wheeled humanoid.
progress this week:
- designed 3 wheel holonomic drive and ordered sheet metal parts within 48 hours
- same day redesign and order for sheet metal, belt driven so101
- set prelim actuation points for tendon driven arm that uses Dyneema DM20 (hopefully zero creep), almost 4 DOF actuation at or proximal to shoulder adduction
- calculated almost 8x payload at rated actuator torque (omitted added transmission weight though)
Setting up US banking as an international founder is cooked.
@mercury - app got shot down because no EIN despite announcing pre-EIN banking through stripe atlas
@rhobusiness - blocked by not having a US residential address
@brexHQ - same address blocker
who can help set us up asap?
4
5
828
interesting...
We built a bipedal robot for about $2,500.
A real, mostly 3D-printed robot you can build, repair, simulate, train, and control.
Today we’re releasing LeRobot Humanoid: an open robot-learning platform with hardware, runtime, identification tools, and training environments.
Blog post: huggingface.co/blog/VirgileB…
Repo: github.com/Virgileboat/lerob…
1
977
12 hour evening-sunrise lock in:
> started at 7pm and now it's 7am, still going
> did clean up logistical work for robotics hackathon I hosted last weekend
> speedran 14 page additive manufacturing final report and final exam preparation for thursday
> did calculations and design of waist pitch joint for my <$5k modular retail robot
> requested quote for polycarbonate shell, bought more robot parts (waist pitch actuator, belts, pulleys, bearings, buck/boost converters, cable connectors)
> heading back to campus later to return wrong linear rail blocks, pick up and print more robot parts
4
18
1,435
juicy technical details for the waist pitch joint:
> back of the napkin math, extreme edge case (torso bent over parallel to ground) 265N point load at shoulder adduction joint, 700mm from waist pitch joint)
> 185Nm required, target 1.5rpm for reasonable speed
> same trading time for cheaper hardware philosophy was applied
> DM4340 selected with 24:1 gear ratio gets us 1.5rpm and 216Nm nominal torque
transmission:
> single stage belt drive is impractical, considered buying gear reducers on taobao, found out they were all 500-1000rmb, bulky and inelegant to integrate, briefly considered designing own but definitely not cheaper
> final selection was a 2 stage belt drive (6:1 and 4:1), cheap, light, low backlash and easy to integrate
> since 2 stage belt drive and only 2 shafts, slotted mount tensioning didn't make sense, eccentric idler pulleys were chosen instead both for tensioning and improving wrap around drive pulley
safety:
> the main concern was safety given any failure on waist pitch can be really dangerous
> worm gears are self locking and convenient to package here but rejected due to abysmal efficiency, so then a normally engaged, power-off brake on the output shaft was selected such that, even if belts slip, snap or power is lost, the brake can mechanically arrest the joint and prevent a fall
> this means the brake has to hold around 185Nm in the edge case, friction brakes for 200Nm were around 1000rmb, were bulky and heavy
> selected solenoid driven hardened steel pin and dog ring for a cheap and strong form locking brake, albeit with the disadvantage of discrete locking positions
360
the goat keeps shipping
CounterStrike-1K is almost here.
> 1K hours of multi-pov, pro gameplay
> Video (720p@32fps, h.264) + Audio
> action-annotated + rich game metadata
> Optimized for training (GOP=32+WDS)
Cleaning the data now - whoever flags the most render bugs gets 50$
d2z2oba0ltmswi.cloudfront.ne…
2
493
We’ve postponed the sprint to May 15–17 so a new sponsor can join us. Applications are still open until then!
the boys stay cooking
After building 25+ XLeRobots for hackathons, we redesigned almost every structural part except the arms and head.
And we’re open sourcing the hardware upgrade behind our XLeRobot build.
It’s the actual internal print file we use for customer builds and hackathons.
We made it:
Easier to print - complete .3mf with tested on Bambu P1S/P2S/H2S/H2D.
Easier to build - square side-loaded nut traps that don't fall out during disassembly, snap-on driver board covers (no tools), dedicated USB hub mounts for clean cable routing, quick-release power bank mount, etc.
More stable base - bearing-supported omni wheel modules so the axle is loaded from both sides, replacing the stock LeKiwi chassis that wobbles under full XLeRobot weight.
Thanks to:
@IsaacSin12 @ThomasSchicksal @xu280589 @QILIU9203 @VectorWang2 @LeRobotHF
GitHub repo in reply. Technical specs in thread.
Here is what we updated:
Jed Hoo retweeted
Some updates on the CS:GO 100K hours dataset :)
So far:
> Idempotent, fault-tolerant (w/ retries) CS2 gameplay rendering on @awscloud infra (EC2, SQS, DDB, S3)
> Video(720p@30fps) + Audio + *GT Keyboard/Mouse
> End-to-end benchmarking script w/ in-depth CloudWatch logging
Jed Hoo retweeted
Apr 13: It's time for another round of live robot demos, warm beers and cold pizza at HK's largest robotics lab.
luma.com/bh8lmcr7
Jed Hoo retweeted
meet bopi, my backpack pet
built on top of @livekit's esp32 sdk, this little kewtie pie can spawn occasionally and talk you out of boredom with cute faces
unlike hardware product launches recently, you can buy and build your bopi today
i also have a completely offline version where bopi can go into deep sleep and survive for a few days on your backpack
source code and instructions below
RT @IsaacSin12: ONLY 8 hours away!
MakerMods Physical AI × OpenClaw Hackathon Shenzhen, March 28–29
350 hackers. 25 real robots. live VLA…
decisions:
- holonomic drive design was pretty trivial, omni wheels instead of mecanum to save one actuator at the expense of speed
- moved actuators upwards and added belt transmission for extra ride height, sacrificed direct drive simplicity for generalisability going over cable covers and carpets
- I think I could have moved the actuators diagonal rather than vertical for lower COM but improvement is incremental
- even though I moved the wrist and elbow flexion servos down in the sheet metal so101, I got bottlenecked by the amount of distal mass I was saving, more interesting transmission wouldn't make sense
- decided to work on the tendon driven arm that actually uses decent actuators, more design freedom for real gains
2
12
1,513
















