探索废墟中的机械之美

Parkour in the Wild Loarnins cerandeeniiio。

Nikita Rudin，Junzhe He，Joshua Aurand and Marco Hutter。

Robotics Systems Lab，ETH Zurich &NVIDIA。

Learning a General and Extensible agile Locomotion。

Policy Using Multi-expert Distillation and RL Fine-tuning。

We present a framework for agile locomotion of legged robots by combining multi-expert distillation with reinforcement learning fine-tuning。

first training specific expert policies are trained to develop specialized locomotion skills。

These policies are then distilled into a unified foundation policy via the DAgger algorithm.

The distilled policy is subsequently fine-tuned。

using RL on the broader training set，including real-world 3D scans。

The framework allows further adaptation to new terrains through repeated fine-tuning。

The fine-tuned policy performs all expert skills indoors.

And generalizes to unseen scenarios outdoors to slippage and disturbances。