A quadruped robot has learned to walk across slippery, uneven terrain entirely through simulation, without any human-designed gaits or manual tuning. The system relies on deep reinforcement learning ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible results