This is Stringman. An open source room scale CDPR compatible with LeRobot and designed for picking up laundry. @IlirAliu_
$1235 assembled at https://t.co/UY0T5oTlwZ
I tried to put too much variety in the dataset again. Saw a little success with DiT and thought "this model can handle everything!" So I trained it on 250 eps of all kinds of random tasks. and just hoped for the best.
The model learned to ignore the prompt and generate average movement examples. It also failed to make any cross room traversals.
I'm going to go back to training a single cam grasp only model again because that worked.
This is Stringman. An open source room scale CDPR compatible with LeRobot and designed for picking up laundry. @IlirAliu_
$1235 assembled at https://t.co/UY0T5oTlwZ
@1972parzival@IlirAliu_ Yes, entirely because rooms have four corners. A three wire system would be more stable, but have less usable work area. I'd love to hear any feeback
You can order a Stringman cable robot at https://t.co/Kz4LUEK2QB
I have sold out of the small initial batch of robots and I will have more ready to ship by late June. they're mostly finished but waiting on actuators from Damiao.
I opted to skip the "pre-order" pattern and it's just "order" :) I've never pre-ordered anything myself, but let me know how you feel about pre-orders.
I'll fill orders as fast as I can. it's a solo operation, but the assembly process is pretty streamlined.
@fabiannode55 I mean I can't ignore that. Do you want to buy one? I have one left until late June when I get a shipment of motors.
https://t.co/E0wtARq3Rv
I will do everything in my power to make sure that you enjoy it.
This is Stringman. An open source room scale CDPR compatible with LeRobot and designed for picking up laundry. @IlirAliu_
$1235 assembled at https://t.co/UY0T5oTlwZ
@PeterDiamandis Actually the only expensive thing left is your health insurance. We could eliminate that obstacle. The number of people starting businesses would quadruple
I've talked to tons of potential customers for my product and the only unanimously held opinion is that subscription services suck and they'd rather just own something outright.
And yet talk to anyone with experience in business or investing. And they'll tell you if you don't have a source of recurring revenue you die.
That was the layout of the first prototype roughly. I wanted the lateral lines to stay high, always above head level if possible. But several factors led to me swapping the vertical "winch" element with a 50cm plastic tube.
first, the closer to the ceiling the box gets, the more errors in calibration and timing matter, making it difficult to precisely position it the higher it gets.
Second, in this robot I supply power to the gripper from one of the support lines. That line had to wind around a small diameter spool in the gripper with a smaller slip ring and I became concerned about wire fatigue there.
Fourth, I wanted a wrist DOF. the rigid tube gave me that but also making the rigid tube move up and down would have added complexity, noise, and weight.
finally, I just lived with the wires for a while and realized that having them low sometimes when the robot was operating wasn't a big deal. I generally operate the bot when nobody is in the room and then park it up out of reach on a hook.
The 50 cm vertical tube solved a bunch of these problems, while also still letting it get right next to furniture. Also the exact length needed to be known for accurate swing cancellation, and with the previous version I kept having to zero the winch line to find out how long it was at boot.
A robot arm basically has a rocket equation problem. Each motor closer to the root must support the weight of all the ones above it.
An advantage of cable robots is that the motors don’t have to be on the moving part. In a CDPR you can place all the main motors on the wall, and they can all be the same part with the same torque. You don’t even have to place one in each corner. In theory you could have a cable robot in every room in your house and put all the motors in the basement with the lines passing through PTFE tubes in the walls.
In Stringman, each anchor has a pair of motors, one direct, and one indirect, routed though a ceramic fishing eyelet in another corner. Consolidating the motors into fewer components drives the cost down, but two camera angles still gives a good view of the room.
The spool motors are Damiao direct drive actuators (DM-H6215). And unlike an arm, they’re all sharing the load, so they can be one of the smaller models, and having no reducer keeps the noise to a minimum.
If you want a cable robot in every room in your house by the way, I’m your guy. https://t.co/NISnwnUxVI
If you have a set of m3 bolts you can assemble most of the gripper without motors to see how it works. https://t.co/7ufKy4C3sp
You can also just print Anchor/Anchor - cover top.stl in white to get an idea of how it would look on your wall.
I have one hardware kit left, after that I'll have more when motors arrive from damiao in late june.
https://t.co/mzYZfp3yeS
I'd be happy to help with the build process, you can dm me or ask on discord https://t.co/uS2EebZgZf
Yes, and I'm not totally sure how best to address that concern. I hope you'll believe me when I say it's not a big deal (you've walked in front of fishermen at the shore before right?) But the fear is real and many people are put off by it. All I know how to say is that you get used to it easily and I'm hoping to offer a level of convenience that makes the wires worth the trouble.
@mattarderne It opens up a lot of possibilities and makes it so that anything that moves can potentially move intelligently.
Most of the credit though goes to raspberry pi and Chinese actuator companies for making hardware cheap.
Without the cables, you'd either be down on the floor where all the obstacles are with legs or wheels, or paying 10x more to have tracks on the ceiling. Both are possible but involve more cost.
I think the cables are really the answer to "how can a get a useful reach and payload now for the lowest cost"