Digit ran mobile manipulation for 12 hours straight
Chris Paxton says the Agility Digit demo he ran at IROS went about 12 hours, autonomous, in a place that was not in its training data. Elsewhere, Yifan Hou says most of Atlas's new hand motions could not be teleoperated even if you tried.
Shown this week
Paxton says Digit ran mobile manipulation for 12 hours at IROS
Chris Paxton says the end-to-end whole body mobile manipulation demo on an Agility Robotics Digit ran for about 12 hours at IROS this week. He says the footage is 1x speed, autonomous, and in a previously unseen location that was not in the training data.
video is 1x speed, autonomous, in a previously-unseen location (not in training data).
Yifan Hou: most Atlas hand motions are impossible to teleoperate today
Yifan Hou responded to Boston Dynamics' new Atlas hands by saying most of the motions shown could not be teleoperated (a human controlling it remotely) with current tools. Alberto Rodriguez announced the hands yesterday, saying they were designed with high fidelity simulation in mind to enable sim-to-real RL, trained in simulation and run on the real robot.
Most motions here are impossible to teleoperate today.
Teleop and other arguments
RL versus IK fight turns into a fight about whose model is cruder
The argument over whether RL should replace inverse kinematics (solving joint angles to reach a target) narrowed to one question: every method is only as good as its model. Roei Herzig said RL is only as accurate as the simulator and that IK and MPC still win for precise tracking on a fixed-base 6DoF arm.
IK assumes you rigidly go where commanded. Stiction, backlash and flex violate that.