It was just yesterday that we covered how Atlas can use whole body movement and deep learning to pick up and move heavy items like a fridge. These tasks may look easy to humans but robots need to take into account mass, inertia and other factors. The robot uses its whole body control to achieve range of motion and balance that is superior to humans.
Atlas uses reinforcement learning to learn how to lift a fridge. It practices the moves with fridge variations in simulation. As Boston Dynamics explains:
The policy driving the behavior has learned to adapt to variations like the location of the fridge, its mass, the amount of grip on the ground and with the fridge, or the configuration where the fridge settles in between the torso, arms, and hands.
[HT]
