Event
End-to-End Reinforcement Learning Study of Quadrupedal Robot Jumping in Lunar Gravity
1 reports · 1 sources
Levin Müller and other researchers propose a multi-stage reinforcement learning method that generates a single end-to-end control policy for commanded point-to-point jumping of a quadrupedal robot under lunar gravity. In simulation, the jump landing-point error is smaller than the robot's body size; in hardware transfer tests, the policy successfully completed vertical jumps under lunar-gravity offloading conditions, and the measured torque and power curves were consistent with simulation and within hardware limits. The policy architecture is simple and relies only on onboard sensing compatible with the task. The study also analyzes the torque and power requirements of lunar jumping, as well as performance degradation under extreme anomalies such as control updates being frozen for up to 1 second during flight. The researchers state that it is deployable for real lunar missions.
Report timeline
- End-to-End Reinforcement Learning for Quadrupedal Jumping in Lunar Gravity
The study proposes a multistage reinforcement learning method that produces a single end-to-end control policy for commanded point-to-point jumping by a quadruped robot under lunar gravity. The policy architecture is simple and relies only on mission-compatible onboard sensing; in simulation, landing errors are smaller than the robot's body size. The work also analyzes torque and power demands and characterizes performance degradation under extreme outlier events, including control-update freezes of up to 1 second during flight. In hardware transfer tests, the policy successfully completed a vertical jump under lunar-gravity offloading, and measured torque and power profiles matched the simulation and stayed within hardware limits.
arXiv cs.RO ·