ANYmal Parkour paper trains agile navigation for quadrupedal robots

Evidence notes
- The ANYmal Parkour paper proposes a learned hierarchy for walking, jumping, climbing, and crouching through parkour-like obstacles.
- The authors report successful real-world transfer on hardware at speeds up to two meters per second.
- The work gives ANYmal a concrete benchmark for agile locomotion and real-world policy transfer.
Company context
ANYbotics develops autonomous legged robots for industrial inspection in complex and hazardous environments. Its ANYmal platform targets repeatable monitoring, inspection and safety workflows.