Vierbeiniger Roboter lernt, Leitern zu erklimmen

“The proliferation of robots like [Boston Dynamics’ Spot](https://techcrunch.com/2024/09/23/boston-dynamics-spot-can-now-autonomously-unlock-doors/) has showcased the versatility of quadrupeds. These systems have thrived at walking up stairs, traversing small obstacles, and navigating uneven terrain. Ladders, however, still present a big issue.

… The team outfitted the quadruped with specialty end effectors that hook onto ladder rungs. The real secret sauce, however, is reinforcement learning, which helps the system adjust to the peculiarities of different ladders.

The system can correct itself in real-time, adjusting its climb instances in which it has misjudged a run or incorrectly timed a step.”