$6,500 for OpenArm v2. Incredibly smooth movement, 4kg of load per arm. Its unclear how far the level of autonomous control has come so far. The video shows mostly tele-operated operation. I have to agree with other comments here that folding laundry is the killer app.
One would expect it to at least have calibration and inverse kinematics solved. That's still a massive pain point for the SO-ARM101 despite it being so common now, since there's no base_link to gripper calibration nor sag compensation and every center offset stacks.
Lerobot banked on having full DL control of each joint so they wouldn't have to do that properly, but now VLAs just use inverse kinematics as a crutch and predict gripper offsets, so we're back to square one.
I'm not super up to date on new releases, the one I'm most familiar with is X-VLA which predicts gripper poses and relative movement deltas for kinematics to move to instead of joints, and trains on forward kinematic gripper poses. I think Intern and Octo have the same approach.
To me it does not seem like progress that, on a stressed and finite planet, humanity might "need" an energy- and resource-intensive machine to complete a task as physically undemanding (and even frivolous) as folding a sheet. But here we are.
I just want my laundry folding robot to run slower than me, and not be able to do karate backflips etc in the event of some undesired scope creep happening one day.
$6,500 for OpenArm v2. Incredibly smooth movement, 4kg of load per arm. Its unclear how far the level of autonomous control has come so far. The video shows mostly tele-operated operation. I have to agree with other comments here that folding laundry is the killer app.
One would expect it to at least have calibration and inverse kinematics solved. That's still a massive pain point for the SO-ARM101 despite it being so common now, since there's no base_link to gripper calibration nor sag compensation and every center offset stacks.
Lerobot banked on having full DL control of each joint so they wouldn't have to do that properly, but now VLAs just use inverse kinematics as a crutch and predict gripper offsets, so we're back to square one.
Hi! do you have any references for VLAs using inverse kinematics? I'm very curious and can't find much
I'm not super up to date on new releases, the one I'm most familiar with is X-VLA which predicts gripper poses and relative movement deltas for kinematics to move to instead of joints, and trains on forward kinematic gripper poses. I think Intern and Octo have the same approach.
Crosses fingers for a general purpose laundry folding bot
We've come a long way from watching 50x speedups [1] to real-time [2] videos of watching robots fold clothes, student-style.
[1] https://youtube.com/watch?v=gy5g33S0Gzo
[2] https://www.pi.website/blog/pi0
Off-topic, but even though it was 16 years ago, the robot looks so beautiful and well-designed. It looks so aesthetic!
To me it does not seem like progress that, on a stressed and finite planet, humanity might "need" an energy- and resource-intensive machine to complete a task as physically undemanding (and even frivolous) as folding a sheet. But here we are.
I just want my laundry folding robot to run slower than me, and not be able to do karate backflips etc in the event of some undesired scope creep happening one day.
House scale Faraday cages and an airlock for fiber optics. Should be sufficient until maybe 2032.
I've got a collection of socks needing darning!
Finally, automatic hand for single man
One step closer to becoming Adam Jensen
Or Smasher, there's a few Adams we're taking steps towards
i asked for this
I think in target-rich environments, the humanoid form is pretty easy to beat.