March 13, 2024Science RoboticsOpen Access

ANYmal parkour: Learning agile navigation for quadrupedal robots

View Full Paper

Ask AI

Bookmark

Authors

DHDavid HoellerGeorgia Institute of Technology NRNikita RudinETH Zurich DSDhionis SakoETH Zurich

Discussion

Member takes

Implication

Experimental evaluation demonstrates agile parkour navigation in quadrupedal robots via hierarchical learning, indicating robust zero-shot simulation-to-real transfer.

Key Points

Quadrupedal robots achieve autonomous parkour traversal over dynamic obstacle courses using a fully learned hierarchical formulation and agile locomotion skills.
Physical hardware tests achieved obstacle crossing speeds of up to 2 meters per second using perception module representations learned strictly from simulated data.
Simulation-trained navigation policies enable robust scene understanding and real-world deployment without requiring expert demonstrations or prior environmental maps.

Cite This Study

Hoeller et al. (2024) studied this question.

synapsesocial.com/papers/68e74210b6db6435876bbec7 https://doi.org/10.1126/scirobotics.adi7566

View Full Paper

Ask AI

Bookmark

Implication

Experimental evaluation demonstrates agile parkour navigation in quadrupedal robots via hierarchical learning, indicating robust zero-shot simulation-to-real transfer.

ANYmal parkour: Learning agile navigation for quadrupedal robots

Discussion

Member takes

Implication

Key Points

Cite This Study

ANYmal parkour: Learning agile navigation for quadrupedal robots

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider